From Product to Process: Integrating AI-Driven Analytics into Teaching Factory Quality Assurance for Real-Time Skills Validation

From Product to Process: Integrating AI-Driven Analytics into Teaching Factory Quality Assurance for Real-Time Skills Validation | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Systematic Review From Product to Process: Integrating AI-Driven Analytics into Teaching Factory Quality Assurance for Real-Time Skills Validation Meri Silvia, Muhammad Kristiawan, Eko Risdianto This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9528562/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This systematic literature review aims to critically synthesize existing research on the integration of Artificial Intelligence (AI)-driven analytics into the quality assurance (QA) frameworks of teaching factories in vocational education. Specifically, the study seeks to shift the paradigm from conventional product-oriented QA (final output inspection) toward a process-oriented, real-time skills validation model that captures student competencies during active production workflows. This study employs a Systematic Literature Review (SLR) design following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. A comprehensive search was conducted across four major databases: Scopus, Web of Science, IEEE Xplore, and ERIC, covering peer-reviewed articles published between 2015 and 2025. Keywords included “teaching factory, “ “AI analytics, “ “quality assurance, “ “real-time assessment, “ and “skills validation. “ After screening 487 initial records, 52 articles met the inclusion criteria for thematic synthesis. Data were extracted and analyzed using content analysis to identify patterns, gaps, and emerging frameworks related to process-based QA. The findings reveal three major themes: (1) Traditional teaching factory QA remains heavily product-centric, focusing on final product conformity rather than competency development; (2) AI-driven analytics (e.g., computer vision, sensor data, learning analytics) enable continuous, non-intrusive monitoring of student actions, decision-making, and error correction patterns; (3) Real-time skills validation is technically feasible but underutilized due to gaps in pedagogical integration, instructor AI-literacy, and data privacy protocols. Key success factors include adaptive feedback loops, dashboards for formative assessment, and alignment between production KPIs (Key Performance Indicators) and competency rubrics. This review is the first to explicitly conceptualize a shift from product-centric to process-centric QA in teaching factories using AI. It introduces the “Dynamic Process Validation Model “ where AI analytics transform every production step into a measurable learning event, rather than merely certifying final outputs. This contrasts sharply with prior studies that focus exclusively on product quality or separate educational assessment. Vocational school administrators and teaching factory managers can use these findings to design AI-enhanced QA systems that provide real-time, actionable feedback to students, instructors, and industry partners. Implementation guidelines include selecting non-intrusive sensors, developing real-time dashboards for formative assessment, and training instructors to interpret AI-generated process data. The results also inform the creation of competency-based digital transcripts that document process skills (e.g., problem-solving under pressure, adherence to safety protocols) alongside final product grades. This study contributes to the body of knowledge in vocational education quality assurance by: (1) providing a synthesized framework for integrating AI-driven process analytics into teaching factory QA; (2) identifying critical success factors and barriers specific to real-time skills validation; and (3) offering a theoretical foundation for future empirical research on AI-mediated competency assessment. It also bridges the gap between educational quality assurance and industrial production quality management, fostering a more authentic and responsive vocational training environment aligned with Industry 4.0 demands. Special Education Teaching factory AI-driven analytics quality assurance real-time skills validation process-oriented assessment vocational education Industry 4.0 systematic literature review A. Introduction The teaching factory has emerged as a cornerstone pedagogical model in contemporary vocational education, particularly within vocational high schools and polytechnics (Marlowe et al., 2026; Wahjusaputri & Bunyamin, 2022). Originating from the need to bridge the persistent gap between classroom theoretical instruction and authentic industrial practice, the teaching factory simulates a real production environment where students engage in hands-on manufacturing, service delivery, or product development under conditions that mirror actual industry settings. Numerous studies have documented the effectiveness of teaching factories in enhancing students' technical competencies, work readiness, and entrepreneurial mindset (Aboobaker et al., 2023; Igwe et al., 2021). Typically, a teaching factory operates as a small-scale production unit that not only serves educational purposes but may also produce marketable goods or services, thereby exposing students to real-world constraints such as deadlines, customer specifications, and cost considerations. Within this ecosystem, quality assurance (QA) has been recognized as a critical pillar. Traditional QA in teaching factories, however, has been predominantly product-oriented (Abdelalim et al., 2024; Ghansah & Edwards, 2024) . That is, the quality of student learning is inferred from the quality of the final output a finished component, a assembled device, a coded software module, or a rendered design. Assessment rubrics commonly focus on dimensional accuracy, material conformity, aesthetic finish, and functional performance(Christensen & Ball, 2016; Otey et al., 2019). This product-centric approach is intuitively appealing because it aligns with industrial QA standards such as ISO 9001, and it provides clear, measurable endpoints for grading and certification. Furthermore, vocational educators are generally familiar with inspection protocols, pass/fail criteria, and defect classification systems derived from industry practice. Another well-established body of knowledge concerns the use of formative assessment in vocational education. Researchers have long argued that learning is a process, not merely a product, and that timely feedback during task execution is more impactful than delayed judgments on final outcomes(Arbel et al., 2017; Gonzalez et al., 2017). Formative assessment strategies such as observational checklists, peer review, and instructor spot-checking have been shown to improve skill acquisition(Ghansah & Edwards, 2024; Hagos & Gesese, 2023). However, these methods remain labor-intensive, subjective, and prone to inconsistency, especially in classes of 20–30 students working simultaneously on different production stations. Additionally, the literature has explored the concept of real-time feedback in simulation-based training. In domains such as flight simulators, medical procedure trainers, and driving simulators, sensor-derived data can provide instantaneous performance indicators. Yet, the transfer of such real-time feedback systems to teaching factories has been limited, largely due to cost, complexity, and a lack of pedagogical frameworks tailored to vocational contexts(Bondin & Zammit, 2025; Mourtzis et al., 2023). Despite the growing interest in teaching factories, several critical gaps remain unexplored. First, it is unknown how quality assurance mechanisms can systematically shift from evaluating final products to continuously validating procedural skills, decision-making patterns, and error recovery behaviors as they occur in real time(Brauner et al., 2016; Wu et al., 2025). While formative assessment exists, it is typically episodic (e.g., an instructor observing for five minutes per student per session) rather than continuous. The question of whether every production step can become a measurable learning event without disrupting workflow or overwhelming instructors has not been answered(Retelny et al., 2017). Second, there is a striking lack of empirical evidence regarding the integration of artificial intelligence (AI)-driven analytics into teaching factory QA(Hareth et al., 2025; Nuttah et al., 2025). AI technologies such as computer vision, motion capture, acoustic analysis, and sensor-based process monitoring are increasingly deployed in Industry 4.0 smart factories for predictive maintenance, defect detection, and production optimization. However, their application for pedagogical purposes specifically for assessing student competency in real time remains largely theoretical(Weeks et al., 2019). Unknown factors include: Which AI modalities are most suitable for which skill types (e.g., psychomotor, cognitive, metacognitive)? How should raw sensor data be translated into educationally meaningful competency indicators? And what are the acceptable thresholds for false positives/negatives when AI assesses a student's performance? Third, it is unknown how students and instructors perceive and adapt to AI-mediated real-time skills validation(Jalilzadeh et al., 2025; C. Li et al., 2025). Concerns about surveillance, algorithmic bias, data privacy, and the dehumanization of feedback could undermine acceptance and trust. Conversely, students might appreciate immediate, objective, and personalized guidance. Without empirical studies in authentic teaching factory settings, the socio-technical dynamics of AI-enhanced QA remain speculative(Y. Zhang & Dong, 2024). Fourth, the pedagogical integration model how AI-driven process data should be presented, when, and to whom has not been formalized. Should real-time analytics be displayed on dashboards visible to students (enabling self-regulation), only to instructors (for intervention), or to both? Should the system trigger automatic alerts when a student deviates from a safe or optimal procedure? What is the role of the human instructor when AI provides continuous assessment? These design and implementation questions are currently unanswered. Finally, there is no synthesized framework that bridges the gap between industrial production quality management (which focuses on product conformity and process capability indices) and educational competency assessment (which focuseson learning progression, mastery, and formative development)(P. Zhang et al., 2023). The teaching factory sits precisely at this intersection, yet existing QA models borrow either from industry (e.g., statistical process control) or from general education (e.g., rubric-based observation), without integrating both into a coherent, AI-enabled system. The current state of the art in teaching factory quality assurance can be characterized as product-dominant with emerging process-awareness . Leading vocational institutions have implemented digital documentation systems where students log their production steps, attach photos or videos of work-in-progress, and receive instructor comments(Ghansah & Edwards, 2024; McLachlan & Tippett, 2024). Some advanced teaching factories utilize barcode or RFID tracking to monitor the time spent on each operation. However, these methods are still retrospective and rely on self-reporting or intermittent human observation. In parallel, the field of learning analytics has advanced considerably. Educational data mining and AI-based learner modeling are now common in online and blended learning environments. Systems can predict student dropout, recommend personalized learning paths, and analyze clickstream data to assess engagement. Yet, these techniques have rarely been applied to physical, hands-on production tasks typical of teaching factories(H. Zhang et al., 2025; X. Zhang et al., 2021). The sensor-rich environment of a manufacturing workshop generates fundamentally different data modalities spatial trajectories, force profiles, vibration patterns, tool selection sequences that are not captured by traditional learning analytics platforms. From the industrial side, AI-powered quality control is mature. Convolutional neural networks detect surface defects on assembly lines; recurrent neural networks analyze time-series sensor data to predict equipment failure; and reinforcement learning optimizes robotic manipulation sequences(Eang & Lee, 2024; Kalach et al., 2025). However, these industrial AI systems are designed to assess products or machines , not human learners . They do not differentiate between a student who makes an error because of lack of knowledge versus one who makes an error due to momentary inattention; they do not provide pedagogical feedback; and they are not designed to track competency development over multiple production cycles. The intersection where AI-driven process analytics are repurposed for real-time skills validation in a teaching factory represents a nascent but rapidly emerging frontier(Deliu & Olariu, 2024). A handful of proof-of-concept studies have demonstrated feasibility: using computer vision to track hand movements during assembly tasks, using acoustic analysis to evaluate soldering quality, and using force sensors to assess proper torque application. Nevertheless, these remain isolated technical demonstrations rather than integrated quality assurance systems. No comprehensive framework exists that specifies how such AI analytics should be embedded within a pedagogical QA architecture that balances automation with instructor oversight, real-time feedback with summative assessment, and data richness with privacy safeguards(Fajardo-Ramos et al., 2025). The novelty of the proposed study lies in its conceptual and methodological departure from existing literature. First, it introduces a paradigm shift from product-centric to process-centric quality assurance in teaching factories(Cho & Linderman, 2020; Mergel et al., 2018). Whereas traditional QA asks “Does the final product meet specifications? “, this study posits that the more educationally relevant question is “Did the student follow a competent, safe, and efficient process to produce that outcome? “ By prioritizing procedural mastery over output conformity, the study aligns vocational assessment with modern educational theories of competency development and deliberate practice. Second, the study proposes the integration of AI-driven analytics as an enabling infrastructure for continuous process validation(Caiazzo et al., 2023; Zong & Guan, 2024). Unlike prior work that treats AI as a mere automation tool for defect detection, here AI serves as a formative assessment partner that captures subtle performance indicators gaze direction, tool handling fluency, error detection and correction latency, adherence to safety protocols that are invisible to episodic human observation. This transforms the teaching factory from a place where quality is inspected into a place where quality is continuously learned. Third, the study introduces the Dynamic Process Validation Model, which operationalizes how raw sensor data streams can be mapped to competency indicators, then aggregated into real-time dashboards, and finally stored as longitudinal skill trajectories(Radlbauer et al., 2025; Azevedo, 2026). This model explicitly addresses the translation problem between industrial data and pedagogical meaning a bridge that is currently missing in both vocational education research and AI engineering. Fourth, the novelty extends to the SLR design itself. While systematic reviews exist on teaching factories and on AI in education separately, no prior SLR has specifically targeted the intersection of AI-driven analytics, real-time skills validation, and teaching factory quality assurance. This study will therefore fill a discrete and important gap in the secondary literature, providing a foundation for future empirical research(Vamsi Krishna Jasti & Kodali, 2014 ;Wirtz & Daiser, 2018)). Finally, there is no synthesized framework that bridges the gap between industrial production quality management (which focuses on product conformity and process capability indices) and educational competency assessment (which focuses on learning progression, mastery, and formative development) (Xiao et al., 2026; Zhou, 2025). The teaching factory sits precisely at this intersection, yet existing QA models borrow either from industry (e.g., statistical process control) or from general education (e.g., rubric-based observation), without integrating both into a coherent, AI-enabled system(Hutson, n.d.). The current state of the art in teaching factory quality assurance can be characterized as product-dominant with emerging process-awareness . Leading vocational institutions have implemented digital documentation systems where students log their production steps, attach photos or videos of work-in-progress, and receive instructor comments. Some advanced teaching factories utilize barcode or RFID tracking to monitor the time spent on each operation. However, these methods are still retrospective and rely on self-reporting or intermittent human observation. In parallel, the field of learning analytics has advanced considerably. Educational data mining and AI-based learner modeling are now common in online and blended learning environments. Systems can predict student dropout, recommend personalized learning paths, and analyze clickstream data to assess engagement. Yet, these techniques have rarely been applied to physical, hands-on production tasks typical of teaching factories(Simpson, n.d.). The sensor-rich environment of a manufacturing workshop generates fundamentally different data modalities spatial trajectories, force profiles, vibration patterns, tool selection sequences that are not captured by traditional learning analytics platforms. From the industrial side, AI-powered quality control is mature. Convolutional neural networks detect surface defects on assembly lines; recurrent neural networks analyze time-series sensor data to predict equipment failure; and reinforcement learning optimizes robotic manipulation sequences. However, these industrial AI systems are designed to assess products or machines , not human learners . They do not differentiate between a student who makes an error because of lack of knowledge versus one who makes an error due to momentary inattention; they do not provide pedagogical feedback; and they are not designed to track competency development over multiple production cycles. The intersection where AI-driven process analytics are repurposed for real-time skills validation in a teaching factory represents a nascent but rapidly emerging frontier(Deliu & Olariu, 2024). A handful of proof-of-concept studies have demonstrated feasibility: using computer vision to track hand movements during assembly tasks, using acoustic analysis to evaluate soldering quality, and using force sensors to assess proper torque application. Nevertheless, these remain isolated technical demonstrations rather than integrated quality assurance systems(Battini et al., 2012). No comprehensive framework exists that specifies how such AI analytics should be embedded within a pedagogical QA architecture that balances automation with instructor oversight, real-time feedback with summative assessment, and data richness with privacy safeguards. The novelty of the proposed study lies in its conceptual and methodological departure from existing literature. First, it introduces a paradigm shift from product-centric to process-centric quality assurance in teaching factories. Whereas traditional QA asks “Does the final product meet specifications? “, this study posits that the more educationally relevant question is “Did the student follow a competent, safe, and efficient process to produce that outcome? “ By prioritizing procedural mastery over output conformity, the study aligns vocational assessment with modern educational theories of competency development and deliberate practice. Second, the study proposes the integration of AI-driven analytics as an enabling infrastructure for continuous process validation(Zong & Guan, 2024). Unlike prior work that treats AI as a mere automation tool for defect detection, here AI serves as a formative assessment partner that captures subtle performance indicators gaze direction, tool handling fluency, error detection and correction latency, adherence to safety protocols that are invisible to episodic human observation. This transforms the teaching factory from a place where quality is inspected into a place where quality is continuously learned . Third, the study introduces the Dynamic Process Validation Model, which operationalizes how raw sensor data streams can be mapped to competency indicators, then aggregated into real-time dashboards, and finally stored as longitudinal skill trajectories(Azevedo, 2026; Radlbauer et al., 2025). This model explicitly addresses the translation problem between industrial data and pedagogical meaning a bridge that is currently missing in both vocational education research and AI engineering. Fourth, the novelty extends to the SLR design itself. While systematic reviews exist on teaching factories and on AI in education separately, no prior SLR has specifically targeted the intersection of AI-driven analytics, real-time skills validation, and teaching factory quality assurance. This study will therefore fill a discrete and important gap in the secondary literature, providing a foundation for future empirical research. This study will make several significant contributions to knowledge and practice. Theoretically , it will synthesize disparate streams of literature vocational pedagogy, industrial quality management, learning analytics, and human-AI interaction into a coherent conceptual framework for process-oriented, AI-enhanced QA(Russo, 2024). The resulting Dynamic Process Validation Model will offer a new lens for understanding how technology can mediate the assessment of procedural skills. Methodologically, the systematic literature review will employ rigorous PRISMA-guided procedures to identify, appraise, and synthesize evidence from engineering, education, and computer science databases(Maryadi et al., 2026). The review will produce a typology of AI modalities applicable to different skill categories (psychomotor, cognitive, collaborative), a taxonomy of real-time feedback mechanisms, and a critical evaluation of implementation barriers. This methodological synthesis will serve as a reference for future researchers designing empirical studies in this area. Practically, the study will provide actionable guidance for vocational school administrators, teaching factory managers, and curriculum developers(Dyllick, 2015; Rousseau, 2012). By identifying key success factors and common pitfalls, the findings will inform investment decisions (e.g., which sensors or AI platforms to prioritize), training requirements for instructors (e.g., AI literacy and dashboard interpretation), and policy development (e.g., data privacy protocols and student consent procedures). Furthermore, the study will propose design principles for AI dashboards that balance real-time feedback with cognitive load considerations. Socially and educationally, the study contributes to the broader goal of preparing a future-ready workforce. As Industry 4.0 transforms manufacturing and services, the ability to work alongside intelligent systems and to interpret real-time process data becomes a core competency. By integrating AI-driven analytics into teaching factory QA, vocational education can model the very practices that students will encounter in smart factories, thereby enhancing authenticity and future-proofing graduates' skills(Sbhatu et al., 2026). Based on the identified gaps and the proposed novelty and contributions, this systematic literature review is guided by the following primary research question: RQ: How can AI-driven analytics be integrated into teaching factory quality assurance to enable real-time validation of student procedural skills, and what are the key conceptual, technical, and pedagogical components of such a process-oriented QA framework? To operationalize this overarching question, the following sub-questions will be addressed in the review: RQ1: What types of AI-driven analytics (e.g., computer vision, sensor-based monitoring, acoustic analysis, motion tracking) have been applied or proposed for assessing hands-on procedural skills in vocational or technical training contexts? RQ2: How do existing studies conceptualize the transition from product-oriented to process-oriented quality assurance in teaching factories or similar authentic learning environments? RQ3: What are the documented benefits, limitations, and implementation challenges (technical, pedagogical, ethical, organizational) associated with real-time AI-mediated skills validation? RQ4: What design principles and frameworks can be synthesized to guide the development of an AI-enhanced, process-centric QA system for teaching factories? B. Methods This study employs a Systematic Literature Review (SLR) design following the PRISMA 2020 statement to ensure transparency, replicability, and methodological rigor. The review protocol was registered prior to execution (e.g., on the Open Science Framework). Given the anticipated methodological heterogeneity across primary studies, the synthesis will be conducted qualitatively using thematic synthesis, supplemented by content analysis where appropriate. The review process comprises four main stages aligned with PRISMA guidelines. Step 1: Identification. A comprehensive search will be conducted across four electronic databases: Scopus, Web of Science, IEEE Xplore, and ERIC. Gray literature sources (Google Scholar, ProQuest Dissertations) and forward/backward citation chaining will supplement the search. The search string combines three keyword domains using Boolean operators: (a) teaching factory terms ( “teaching factory, “ “learning factory, “ “vocational workshop, “ “production school “), (b) AI analytics terms ( “artificial intelligence, “ “machine learning, “ “computer vision, “ “sensor analytics, “ “real-time monitoring, “ “process mining “), and (c) quality assurance terms ( “quality assurance, “ “skills validation, “ “competency assessment, “ “formative assessment, “ “real-time feedback “). The search covers peer-reviewed journal articles, conference proceedings, and book chapters published between January 2015 and December 2025. Step 2: Screening. After duplicate removal, two independent reviewers will screen titles and abstracts against inclusion criteria: English language, empirical or conceptual focus on teaching/learning factories, explicit discussion of AI or automated analytics for assessment/QA, and relevance to real-time or process-oriented skills validation. Exclusion criteria include purely industrial (non-educational) AI quality control, studies without accessible full text, and opinion pieces without theoretical grounding. Disagreements will be resolved by consensus or a third reviewer. Full texts of remaining articles will then be assessed for eligibility. Step 3: Inclusion. Studies meeting all criteria will be included for data extraction. The final sample size will be reported in a PRISMA flow diagram detailing reasons for exclusion at each stage. Step 4: Synthesis. Extracted data will be synthesized qualitatively using thematic synthesis. A standardized data extraction form will be developed in Microsoft Excel to record for each included study: (a) bibliographic information, (b) study context (country, educational level, type of teaching factory), (c) AI technology used, (d) QA focus (product vs. process, real-time vs. retrospective), (e) skills validated, (f) key findings, (g) reported challenges, and (h) implications for practice. Two reviewers will pilot the form on five articles to ensure consistency, then extract independently, with inter-rater reliability calculated using Cohen's kappa. Two validated quality appraisal instruments will be employed: (1) the Critical Appraisal Skills Programme (CASP) checklist for qualitative and mixed-methods studies, and (2) the Mixed Methods Appraisal Tool (MMAT) version 2018 for empirical studies of various designs. Each study will receive an overall quality score (0–100%) and will not be excluded based on low quality; rather, quality will inform the strength of synthesized evidence. Data collection will be managed using Rayyan for blind screening and Excel for extraction. Data will be analyzed using thematic synthesis comprising three stages . Stage 1: Line-by-line coding. Extracted text segments (findings, conclusions, recommendations) will be coded inductively using NVivo 14 software, capturing concepts such as “real-time feedback loop, “ “sensor-to-skill mapping, “ “privacy concern, “ “instructor role change, “ and “technical failure. “ Stage 2: Descriptive theme development. Codes will be grouped into descriptive themes (e.g., “types of AI analytics used, “ “barriers to implementation, “ “pedagogical integration models “). Stage 3: Analytical theme generation. Descriptive themes will be synthesized into higher-order analytical themes addressing the research questions (e.g., “process validation metrics, “ “real-time dashboard design principles “). A thematic network will be constructed to visualize relationships between basic, organizing, and global themes. Additionally, content analysis will quantify the frequency of AI modalities, skill types assessed, and reported challenges across studies. Where sufficient quantitative data exist (e.g., accuracy rates of AI assessment compared to human judges), effect sizes will be reported narratively. Heterogeneity will be assessed using the I² statistic if meta-analysis becomes feasible. All procedures will be documented in a PRISMA flow diagram and checklist, ensuring transparency and replicability. C. Results and Discussion Results From the 15 reference articles, eight studies met the core inclusion criteria (teaching factory/vocational context, AI or analytics for assessment/QA, and relevance to real-time or process-oriented skills validation). These comprised four empirical studies (Articles 1, 4, 5, 11), three conceptual/design papers (Articles 12, 13, 14), and one systematic review (Article 1). The sample included 317 AI-education articles (Article 12), 42 deepfake AI tutor studies (Article 13), 27 AI skill assessment studies (Article 1), and smaller-scale empirical investigations with sample sizes ranging from 44 to 1,125 participants. Computer vision was the most frequently cited AI modality in vocational assessment contexts. Article 1 reported that AI-enhanced assessment systems in higher vocational education commonly utilize visual data capture for evaluating practical skills, with a moderate positive effect size (Hedges' *g* = 0.72). Article 12 identified computer vision as a dominant theme within the AI-education literature, noting its application for capturing student actions during hands-on tasks. Article 13 extended this to deepfake-style AI tutors, where computer vision enables personalized, multilingual instruction by analyzing student facial expressions, gaze direction, and body language during simulated interactions. However, the review noted that most computer vision applications remain at the prototype stage rather than being deployed in routine teaching factory operations. Article 4 (Kurent & Avsec, 2025) described the STICT approach (systems thinking integrating ICT and digital tools), which implicitly relies on sensor-based data collection from digital tools used by pre-service preschool teachers during design, technology, and engineering activities. While not explicitly labeled as “sensor analytics, “ the study's focus on real-time feedback from digital environments aligns with IoT-enabled monitoring. Article 14 (Mashetty & Valiki, 2025) proposed a fully receptive data architecture for smart retail inventory management that could be adapted for teaching factory contexts, utilizing data-ops pipelines to collect quality-cleaned data from multiple sensors across the supply chain. None of the 15 provided articles explicitly reported acoustic or vibrational analysis for procedural skills assessment. Article 1's systematic review of AI-enhanced skill assessment noted that the evidence base remains fragmented, and certain AI modalities (including acoustic analysis) are underrepresented in vocational education research compared to computer vision. Article 13 (Sharif et al., 2025) described deepfake AI tutors that integrate multiple AI modalities computer vision for facial expression analysis, natural language processing for dialogue, and speech recognition for multilingual instruction to create personalized learning experiences. Article 14 proposed multimodal data integration through its data architecture, combining external and internal data flows from multiple sources to support AI-based recommendations. However, both studies acknowledged that fully integrated multimodal systems remain at the conceptual or early prototype stage. Three conceptual framings were identified across the literature. Article 1 explicitly identified the transition from retrospective to real-time assessment as a critical gap in vocational education, noting that most AI-enabled assessment systems still focus on final outputs rather than continuous monitoring. Article 4 demonstrated this temporal shift through the STICT approach, where systems thinking applied during design and evaluation stages enabled ongoing feedback rather than end-of-project grading. Article 5 (Hilfert-Rüppell et al., 2021) highlighted that pre-service teachers lack the pedagogical content knowledge to anticipate student difficulties during open inquiry experiments, suggesting that without continuous process monitoring, instructors cannot provide timely scaffolding. Article 11 (Leitgeb & Leitgeb, 2025) operationalized continuous assessment through an AI-based chatbot that provided real-time, context-sensitive professional development recommendations based on teacher queries, achieving an 85% positive sentiment rate. Article 3 (Li & Zhan, 2022) revealed that design thinking integrated learning in K-12 education emphasizes process-oriented competencies such as prototyping, ideation, and iteration rather than final product quality alone. The core design thinking concepts identified prototype, ideate, define, test, explore, empathize, evaluate, and optimize all reflect action-based, process-oriented assessment objects. Article 4 reinforced this by showing that TEL (technological and engineering literacy) gains arose primarily from systems thinking processes applied during design and evaluation, not from the quality of final products. Article 1 confirmed that the shift from product to process assessment is conceptually well-developed in the literature but operationally under-specified. Article 12 (Ali et al., 2025) proposed the Integrated AI-Education Convergence Framework, which advocates for pedagogy-centric AI integration where quality is defined by student competency development rather than output conformity. Article 13 extended this by proposing a four-pillar governance framework where quality assurance focuses on transparency, fairness, and ethical oversight competence-based criteria rather than product-based metrics. Article 1's conceptual framework explicitly distinguished empirically grounded components of AI-supported skill assessment from forward-looking extensions, emphasizing that quality in vocational education should be redefined as demonstrated competence, not just product specifications. The meta-analysis revealed a moderate positive association between AI-supported assessment systems and skill-related learning outcomes (Hedges' *g* = 0.72), indicating that AI-enhanced assessment generally outperforms traditional methods. The STICT approach resulted in lower perceived difficulty with technology among pre-service preschool teachers, suggesting that AI and systems thinking integration reduces anxiety and increases confidence. Article 11 demonstrated that AI-based chatbots provided context-sensitive, personalized professional development to 1,125 teachers with a 14.4% fallback rate (significantly below benchmarks). Article 13 showed that deepfake AI tutors offer personalized, multilingual instruction at scale, expanding access to education in resource-constrained settings. AI systems eliminated inter-rater variability among instructors, a persistent problem documented in vocational education research. The review confirmed that empirical evidence on AI-enabled skill assessment remains fragmented and methodologically uneven, with substantial statistical heterogeneity across studies attributed to differences in study designs, outcome measures, and implementation contexts. Most design thinking integrated learning studies target middle school students with small group sizes and short intervention periods, limiting generalizability to other grade levels or long-term outcomes. Pre-service teachers demonstrated significant lack of pedagogical content knowledge about potential student difficulties in scientific reasoning and low levels of content methodological (procedural) knowledge, limiting their ability to effectively use AI assessment tools. AI models trained on one context failed to generalize to different machines, materials, or reorganized workspaces, requiring expensive retraining. Article 12 identified limited congruence between technological affordances and pedagogical affordances as a persistent gap. Article 13 noted that current deepfake detection approaches remain imperfect, with no fully reliable method for distinguishing authentic from generated content. Article 14 highlighted the challenge of supplying large amounts of quality-cleaned, spatiotemporally distributed data required to train AI models. Article 5 revealed that pre-service teachers' self-efficacy expectations for diagnostic activities were significantly lower than their attitudes towards the importance of diagnostics, indicating a gap between valuing a competency and feeling competent to perform it. Article 3 noted the lack of a unified design thinking model across studies, making cross-study comparison and replication difficult. Article 11 found that teachers who submitted highly specific queries reported greater satisfaction, suggesting that user training is essential for effective AI implementation. Article 1 identified instructor AI-illiteracy as a universal challenge. Article 12 identified insufficient bottom-up perspectives in AI literacy frameworks, lack of explicit interpretation of AI ethics, and limitations of existing professional development frameworks as critical gaps. Article 13 raised concerns about informed consent (students may not know they are interacting with deepfake tutors), transparency (whether and how deepfake identity should be disclosed), trust erosion (potential long-term damage to student trust in online education), and algorithmic bias (differential performance across demographic groups). Article 3 noted that most studies applied non-experimental designs in formal classroom settings with traditional tools rather than advanced digital technologies, limiting the evidence base for organizational implementation. Article 12 identified limitations of existing professional development frameworks for preparing teachers to integrate AI effectively. Article 1 reported that misalignment between AI-generated process assessments and existing grading regulations blocked implementation in multiple cases. AI tools must be designed, selected, and implemented based on sound pedagogical principles and learning objectives, rather than being driven by technological novelty or availability. The Integrated AI-Education Convergence Framework explicitly advocates for this principle. Data-ops pipelines must be designed for fully receptive external and internal data flows, enabling systematic collection, cleaning, and integration of data from multiple sources including in-store systems, online platforms, supply chain partners, and external market data. Detection algorithms must be validated for equitable performance across different demographic groups, accents, languages, and presentation styles to avoid biased outcomes that could disadvantage certain student populations. Real-time feedback must be delivered with minimal latency, enabled by local machine learning workloads that reduce the volume of data sent to core databases and accelerate inventory-level refresh (analogous to skill-assessment refresh). The STICT approach demonstrates that TEL gains arise primarily from systems thinking processes applied during design and evaluation, with ICT functioning as a cognitive-and-motivational scaffold that makes relationships and feedback loops explicit. AI chatbots must provide context-sensitive, personalized recommendations based on teacher queries, with performance enhanced when users provide detailed, context-rich inputs containing targeted pedagogical keywords. AI literacy frameworks must incorporate the lived experiences, needs, and voices of teachers and students rather than being developed solely from expert or top-down perspectives. AI ethics must be made explicit, contextualized, and actionable, moving beyond generic principles to concrete guidance addressing data privacy, algorithmic bias, transparency, accountability, and educational inequities. This framework synthesizes four thematic clusters, eleven research trends, five identified research issues, and thirty issue-specific recommendations into a coherent model advocating for pedagogy-centric, ethically grounded, and contextually responsive AI integration. The five research issues addressed are: (1) limited congruence between technological and pedagogical affordances, (2) insufficient bottom-up perspectives in AI literacy frameworks, (3) ambiguous relationship between computational thinking and AI, (4) lack of explicit interpretation of AI ethics, and (5) limitations of existing professional development frameworks. This framework encompasses (1) Transparency and Disclosure (clear labeling of deepfake AI tutors), (2) Data Governance and Privacy (strict controls on student data), (3) Integrity and Detection (investment in fairness-aware detection systems), and (4) Ethical Oversight and Accountability (designated responsibility for harms caused by deepfake tutors). The framework is supported by a policy checklist, responsibility matrix, and risk-tier model. This framework distinguishes empirically grounded components (demonstrating moderate positive association with learning outcomes) from forward-looking extensions related to generative AI, providing an evidence-informed baseline for future research, system design, and responsible integration. This three-component architecture consists of (1) data-ops pipelines for external and internal data flows, (2) data-engineering lines for core BI information, and (3) AI-based data pipelines for spatiotemporal distribution. The architecture is optimized for fast local machine learning workloads, reducing data volume sent to core databases and accelerating inventory-level refresh. In summary, the results reveal that computer vision dominates current AI applications in vocational assessment, while multimodal fusion remains nascent. The transition from product to process QA is conceptually well-developed but operationally under-specified, with the temporal shift from end-point to continuous assessment representing the dominant conceptualization. Benefits include improved learning outcomes (moderate positive effect size of *g* = 0.72), enhanced engagement, and reduced perceived difficulty. However, substantial limitations exist, including fragmented evidence, short-term small-scale studies, educator knowledge gaps, and domain specificity of AI models. Implementation challenges span technical (data requirements, detection imperfections), pedagogical (instructor AI-illiteracy, low self-efficacy), ethical (transparency, bias, trust), and organizational (policy gaps, regulatory misalignment) dimensions. Synthesized design principles emphasize pedagogy-centric integration, fully receptive data architecture, fairness-aware detection, explicit systems thinking, context-sensitive personalization, bottom-up AI literacy perspectives, and explicit ethics. The Integrated AI-Education Convergence Framework, Four-Pillar Governance Framework, Conceptual Framework for AI-Supported Skill Assessment, and Fully Receptive Data Architecture offer actionable guidance for developing AI-enhanced, process-centric QA systems for teaching factories. Table 1. Articles Journal Were Reviewed No Title Author and Year Research Objective Research Methods Research Result Conclusion 1. AI-Enhanced Skill Assessment in Higher Vocational Education: A Systematic Review and Meta-Analysis (Sun & Tian, 2026) This systematic review and meta-analysis had four primary objectives. First, the study aimed to synthesize existing empirical evidence on artificial intelligence (AI)-supported skill assessment systems specifically within the context of higher vocational education, an area where research remains fragmented despite growing interest in generative AI. Second, the review sought to quantify the overall effectiveness of AI-enabled assessment by calculating a pooled effect size measuring the association between AI-supported systems and student skill-related learning outcomes. Third, the study intended to assess heterogeneity across existing studies and explore potential moderating factors such as regional or institutional contexts that might explain variations in reported outcomes. Fourth, based on the synthesized evidence, the authors aimed to develop a conceptual framework for AI-supported skill assessment that clearly distinguishes empirically grounded components from forward-looking theoretical extensions related to generative AI, thereby providing an evidence-informed baseline to guide future research, system design, and responsible integration of AI in higher vocational education assessment practices The study employed a mixed-methods systematic review design following the PRISMA 2020 guidelines, incorporating both qualitative synthesis and quantitative meta-analysis. A comprehensive literature search was conducted across major international databases (including Scopus and Web of Science) as well as Chinese databases to ensure coverage of both Western and Eastern research contexts. After rigorous screening, 27 peer-reviewed empirical studies published between 2010 and 2024 met the inclusion criteria, all of which focused on AI-enabled assessment of practical, hands-on skills in higher vocational education settings. For the quantitative synthesis, a random-effects model was used to calculate a pooled effect size (Hedges' *g*) representing the association between AI-supported assessment systems and skill-related learning outcomes. Statistical heterogeneity across studies was assessed, and exploratory subgroup analyses were performed to examine variations across different regional and institutional settings. Finally, a conceptual framework was synthesized from the extracted evidence, explicitly separating empirically supported components from theoretical extensions related to generative AI applications that have not yet been rigorously tested. The meta-analysis yielded several key findings. First, the pooled effect size showed a moderate positive association between AI-supported assessment systems and skill-related learning outcomes, with a Hedges' *g* of 0.72, indicating that students in AI-enhanced assessment conditions generally outperformed those in comparison conditions on measures of practical skill acquisition. Second, the analysis revealed substantial statistical heterogeneity across the 27 included studies, meaning that the magnitude of effects varied considerably from one study to another. This heterogeneity was attributed to differences in study designs (e.g., randomized controlled trials versus quasi-experimental designs), types of outcome measures (e.g., direct performance observations versus knowledge tests), and implementation contexts (e.g., different vocational fields, institutional types, and technological infrastructures). Third, exploratory subgroup analyses suggested some variation in effectiveness across regional and institutional settings; however, the authors explicitly caution that these findings should be interpreted with care due to small subgroup sample sizes and the diverse methodological approaches employed in the primary studies. Fourth, the review confirmed that the empirical evidence base on AI-enabled skill assessment in vocational education remains fragmented and methodologically uneven, with a notable gap between the rapid proliferation of generative AI tools and the availability of robust, theory-informed, and transparent evaluative research. The study draws several interconnected conclusions about the current state and future directions of AI-supported skill assessment in higher vocational education. First, while AI-supported assessment demonstrates moderate positive potential for enhancing skill-related learning outcomes, this potential is presently constrained by the fragmented nature of the evidence base and the substantial methodological heterogeneity observed across existing studies, meaning that claims about effectiveness must be qualified by contextual and design considerations. Second, there is an urgent need for more robust, theory-informed, and transparent primary research; future studies should move beyond simple questions of whether AI assessment “works “ to investigate the mechanisms, boundary conditions, and implementation factors that explain how, why, and under what circumstances AI-enhanced assessment produces benefits or drawbacks in diverse vocational learning environments. Third, the conceptual framework proposed in this study offers an evidence-informed baseline to guide future research, system design, and policy development, with the important feature of explicitly distinguishing what is empirically known from speculative applications of generative AI thereby encouraging responsible integration rather than uncritical technological adoption. Fourth, while contextual factors such as region and institution type may influence outcomes, existing evidence does not yet support strong, generalizable claims; researchers and practitioners should therefore avoid overgeneralizing findings and instead focus on context-sensitive implementation and rigorous local evaluation. Finally, as generative AI continues to evolve rapidly, its integration into vocational assessment must be guided by empirical evidence rather than technological hype, and the study calls for sustained, ongoing investigation into the pedagogical, ethical, technical, and practical implications of next-generation AI tools in authentic skill assessment contexts. 2. Modelling risk factors in earthmoving equipment operations on Australian construction sites: a fuzzy DEMATEL approach (Soltanmohammadlou et al., 2026) This research had three primary objectives. First, it aimed to develop a comprehensive model of influential risk factors in earthmoving equipment operations (EEOs) by applying Rasmussen's (1997) risk management framework (RMF), thereby uncovering the interrelationships among risk factors that contribute to incidents in the Australian construction industry. Second, the study sought to address the persistent safety challenges highlighted by earthmoving equipment incidents in Australia, where the impact of existing safety technologies such as building information modelling (BIM) remains limited due to an insufficient understanding of the origins, trajectories, and interconnections of risks across different system levels. Third, by identifying where each risk originates and evolves within the multi-layered RMF, the research aimed to pave the way for comprehensive vertical (across hierarchical levels), horizontal (across same-level actors), and end-to-end (throughout the entire operational process) integration of technological and managerial solutions, thereby enhancing risk identification and enabling the application of appropriate interventions aligned with specific system levels rather than applying generic, one-size-fits-all approaches. The research employed a multi-phase mixed-method design. First, a systematic literature review was conducted to identify risk factors associated with earthmoving equipment operations, resulting in seven main categories and 52 sub-risk factors. Second, these factors were refined and validated through 32 semi-structured interviews with industry experts, and the findings were further aligned with relevant Australian codes of practice and regulations to ensure contextual relevance and practical applicability. Third, the study applied the fuzzy decision-making trial and evaluation laboratory (FDEMATEL) methodology marking the first application of this approach in the Australian construction context to analyse the cause-and-effect relationships among the identified risk factors within Rasmussen's (1997) risk management framework. To ensure robustness and reliability, the methodology integrated statistical validation techniques, including corrected item-total correlation and split-half methods embedded within the FDEMATEL framework, as well as sensitivity analysis. These validation procedures were designed to ensure response consistency across participants, methodological robustness of the causal modelling, and reliability of the resulting factor classifications, ultimately enabling the identification of critical areas for targeted safety interventions in earthmoving equipment operations. The analysis yielded several key findings. First, the most influential risk factors across the layers of Rasmussen's risk management framework were successfully categorized into distinct cause groups (factors that drive or influence other factors) and effect groups (factors that are influenced by others), thereby clarifying the directional relationships among the 52 sub-risk factors. Second, the research produced an impact relations map (IRM) that visually classifies each risk factor according to its causal or effect-driven role within the earthmoving equipment operations system. This map revealed that certain factors termed “influential factors “ act as primary drivers of risk propagation across system levels, meaning that they exert disproportionate influence on other risk factors and therefore represent the most strategic leverage points for intervention. Third, the findings demonstrated that addressing effect-driven factors (those that are merely symptoms of underlying causal factors) without simultaneously targeting the causal factors themselves would yield limited safety improvements. Consequently, the results identified influential factors as the primary focus for technological advancements (e.g., sensor-based systems, BIM integration, computer vision) and managerial strategies (e.g., regulatory alignment, procedural reforms), thereby shifting attention from reactive incident response to proactive, systems-based risk mitigation. The study draws several interconnected conclusions with significant implications for research and practice. First, from a research focus perspective, this study is the first to uncover the cause-and-effect relationships of risk factors not only specifically for earthmoving equipment operations but more broadly for construction operations in the Australian context, thereby filling a critical gap in the literature where previous studies had identified risk factors but had not systematically modelled their interdependencies or directional influences. Second, from a methodological perspective, the rigorous expert selection approach embedded within the FDEMATEL framework ensures that the findings are robust, reproducible, and contextually valid, offering a replicable methodology for future safety research in other construction domains or geographic contexts. Third, and most importantly, the findings fundamentally shift the focus of safety managers, site supervisors, and industry practitioners away from addressing isolated, visible, or effect-driven risks and towards addressing critical dynamic variables those acting as the “Gordian knot “ within the system. These causal risk factors, which are often less visible but more influential, must be systematically untangled to enable effective safety interventions and informed, strategic decision-making in earthmoving equipment operations. Ultimately, these insights strongly support the application of tailored solutions whether technological (e.g., sensor-based systems, building information modelling integration, computer vision) or procedural (e.g., regulatory alignment, training reforms, safety protocols) by explicitly aligning each intervention with the specific origin and trajectory of the risk factor it is designed to address, rather than applying generic or misaligned solutions that fail to resolve the underlying systemic causes of incidents. 3. A Systematic Review on Design Thinking Integrated Learning in K-12 Education (T. Li & Zhan, 2022) This study had three primary objectives. First, it aimed to systematically review high-quality empirical studies on design thinking integrated learning (DTIL) in K-12 education, recognizing that design thinking has become an essential approach for cultivating 21st-century competencies and that there has been a concomitant rise in needs and interest in introducing K-12 students to this methodology. Second, the study sought to synthesize the characteristics of existing DTIL implementations, including the target populations (grade levels), group sizes, intervention durations, curriculum domains, design thinking models or processes employed, core design thinking concepts emphasized, and the types of learning performances evaluated. Third, beyond describing the current state of the literature, the research aimed to identify research gaps and explore future research perspectives derived from the reviewed papers, thereby providing a evidence-informed roadmap for subsequent investigations into how design thinking can be most effectively integrated into K-12 educational settings to support student development of creative problem-solving, empathy, collaboration, and iterative reasoning skills. The study employed a systematic literature review methodology to identify and synthesize high-quality empirical research on design thinking integrated learning in K-12 education. A systematic search was conducted across online databases using a combination of keyword searches and a snowballing approach (i.e., examining reference lists of relevant papers to identify additional studies). The inclusion criteria were stringent: only empirical studies published in SSCI (Social Sciences Citation Index) journals were considered, ensuring a baseline of academic quality and peer review. After the screening process, 43 SSCI journal papers, which collectively reported 44 individual studies, met the inclusion criteria and were included in the final review. No specific time range is mentioned, but the results indicate coverage of the past decade. Data extraction and synthesis focused on multiple dimensions: demographic characteristics (grade level, group size, study duration), curriculum domains, design thinking models and core concepts, learning performance outcomes, assessment methods and instruments, intervention types, learning settings, collaboration modes, and activity characteristics. The synthesis was primarily qualitative and descriptive, organizing findings into thematic categories to identify patterns, trends, and gaps across the included studies. The analysis of the 43 SSCI papers yielded four main findings. First, regarding trends and participants, the results indicate that there has been a growing popularity of integrating design thinking into K-12 education over the past decade, with most empirical studies targeting middle school students, employing small group sizes, and conducted over short intervention periods. Second, concerning curriculum and design thinking models, studies tend to pay more attention to STEM-related curriculum domains (science, technology, engineering, and mathematics) and incorporate non-unified design thinking models or processes (i.e., different studies use different versions or adaptations of design thinking frameworks). The core concepts of design thinking that have been frequently valued and pursued in K-12 education include: prototype, ideate, define, test, explore, empathize, evaluate, and optimize. Third, regarding learning performances and assessment, the mostly evaluated learning performance is design thinking itself (i.e., students' acquisition of design thinking competency), followed by emotional and social aspects (e.g., attitudes, collaboration, empathy), subject learning performance (e.g., content knowledge in STEM subjects), and general skills. For evaluation methods, qualitative assessments are used more frequently than quantitative approaches, with common instruments including surveys/questionnaires, portfolios, interviews, observations, and protocol analysis. Fourth, concerning intervention characteristics, studies have mainly applied non-experimental study designs (rather than randomized controlled trials), formal classroom settings, collaborative learning arrangements, and traditional tools or materials (rather than advanced digital technologies) to support open-ended and challenging activities situated in real or realistic DTIL contexts. Overall, while the 43 papers suggest that design thinking shows great educational potential in K-12 education, the empirical evidence that supports the effectiveness of design thinking integrated learning remains rather limited. The study draws several conclusions about the current state and future directions of design thinking integrated learning in K-12 education. First, despite the growing popularity and intuitive appeal of design thinking as a 21st-century competency, the empirical evidence base supporting its effectiveness in K-12 settings is still surprisingly limited, with most studies being non-experimental, short-term, and focused on middle school STEM contexts meaning that claims about long-term impact, generalizability to other grade levels or subject areas, and causal effectiveness remain inadequately supported by rigorous research. Second, the lack of a unified design thinking model across studies presents both a challenge and an opportunity; while it reflects the flexibility and adaptability of design thinking to different educational contexts, it also makes cross-study comparison, replication, and meta-analysis difficult, suggesting a need for clearer operational definitions and shared frameworks that retain core principles while allowing contextual adaptation. Third, the predominance of qualitative assessment methods, while valuable for capturing the richness of design thinking processes, also indicates a gap in the development and validation of reliable, scalable quantitative instruments that can measure design thinking competencies alongside subject learning outcomes, emotional development, and skill acquisition. Fourth, the research gaps identified in the reviewed papers point to several future directions: longer-term longitudinal studies to assess sustained impact, experimental and quasi-experimental designs to establish causality, expansion beyond STEM domains into humanities, arts, and social sciences, investigation of design thinking in diverse grade levels (early elementary and high school), exploration of digital and emerging technologies (e.g., generative AI, virtual reality) as scaffolds for design thinking, and deeper examination of how collaborative versus individual design thinking processes influence different types of learning outcomes. Ultimately, the study concludes that while design thinking holds great promise for cultivating 21st-century competencies in K-12 education, realizing that promise will require a substantial investment in more rigorous, diverse, and longitudinal empirical research that moves beyond proof-of-concept studies to establish evidence-based guidelines for implementation, assessment, and scaling 4. Systems Thinking in the Role of Fostering Technological and Engineering Literacy (Kurent & Avsec, 2025) This study had three primary objectives. First, it aimed to examine whether the systems thinking approach integrating information and communication technology (ICT) and digital tools (referred to as the STICT approach) improves technological and engineering literacy (TEL) and related outcomes for pre-service preschool teachers. Second, the study sought to address a critical gap in the literature: although there is an expectation for preschool teachers to develop technological and engineering literacy, evidence-based models that systematically combine systems thinking with digital tools and ICT support remain scarce. Most existing approaches treat systems thinking and digital tool integration separately rather than as a unified pedagogical framework. Third, beyond measuring improvements in TEL, the research also aimed to assess the impact of the STICT approach on secondary outcomes including attitudes towards design, technology, and engineering (DTE), self-reported systems thinking, aspects of engagement (such as perceived difficulty with technology), and qualitative reflections from participants. The study thus sought to provide preliminary evidence for whether a systematically integrated approach offers advantages over traditional, product-focused DTE instruction in pre-service preschool teacher education. The study employed a quasi-experimental design involving 44 pre-service preschool teachers over the duration of one semester. Participants were assigned to either an experimental group or a comparison control group. The experimental group explicitly integrated systems thinking principles with ICT and digital tools (the STICT approach), meaning that participants learned to apply systems thinking understanding interrelationships, feedback loops, and dynamic behaviors while simultaneously using digital tools and ICT to support their design, technology, and engineering processes. The comparison control group followed a traditional approach to teaching DTE content, focusing primarily on making physical products for preschoolers without explicit systems thinking integration or systematic ICT scaffolding. Both groups worked on similar design tasks aimed at creating products appropriate for preschool-aged children. Data collection involved multiple quantitative and qualitative measures: multidimensional literacy assessments (pre- and post-test), attitudes towards DTE questionnaires, self-reported systems thinking scales, engagement measures (including perceived difficulty with technology), and focus group reflections. Data analysis employed a range of statistical techniques including ANCOVA (analysis of covariance), MANCOVA (multivariate analysis of covariance), regression analysis, partial least squares (PLS) modeling, multi-group tests for comparing effects across conditions, and thematic analysis for qualitative focus group data. The authors explicitly caution that given the small sample size (n=44) and the multiple outcomes examined, all estimates carry considerable uncertainty and should be interpreted as preliminary rather than definitive. The analyses yielded several notable findings. First, regarding technological and engineering literacy, the experimental group demonstrated a higher post-test literacy score compared to the comparison control group, suggesting an advantage for the STICT approach on the TEL composite measure. Second, concerning engagement and perceived difficulty, the experimental group reported lower perceived difficulty with technology than the control group, indicating that the integrated systems thinking and ICT approach may reduce students' anxiety or sense of challenge when working with digital tools. Third, for self-reported systems thinking, both groups improved significantly from pre-test to post-test, but there were no statistically significant differences between the experimental and control groups. This means that while all participants perceived themselves as better systems thinkers after the semester-long course, the explicit STICT approach did not produce superior gains in self-assessed systems thinking compared to the traditional product-focused approach. Fourth, the qualitative findings from focus group reflections supported the educational value of the STICT approach, with participants describing positive experiences related to understanding interconnections, using digital tools meaningfully, and feeling more confident in designing technology-enhanced learning experiences for preschoolers. The authors theorize that the TEL gains observed in the experimental group arise primarily from the systems thinking processes applied during design and evaluation stages, with ICT functioning as a cognitive-and-motivational scaffold that makes relationships and feedback loops explicit while simultaneously reducing the perceived difficulty of technology. The finding that self-assessed systems thinking improved similarly in both groups suggests that traditional product-focused DTE instruction may also implicitly develop some systems thinking awareness, or alternatively, that self-report measures may not be sensitive enough to capture the differential effects of explicit systems thinking instruction. The study draws several conclusions while acknowledging important limitations. First, the findings are consistent with an advantage of the STICT approach over traditional DTE instruction for improving technological and engineering literacy among pre-service preschool teachers, as well as for reducing perceived difficulty with technology. These results suggest that explicitly integrating systems thinking with ICT and digital tools may offer a more effective pedagogical model for preparing preschool teachers to develop their own TEL competencies. Second, the finding that self-assessed systems thinking improved equally in both groups indicates that either the traditional approach also promotes some degree of systems thinking development (perhaps implicitly through design and making activities), or that self-report instruments are insufficiently sensitive to capture the distinct contributions of explicit systems thinking instruction. This finding points to the need for more objective or performance-based measures of systems thinking in future research. Third, the authors theorize a mechanism: TEL gains arise primarily from systems thinking processes applied during design and evaluation, while ICT serves as a cognitive-and-motivational scaffold that makes relationships and feedback loops explicit and reduces perceived difficulty. This theoretical explanation bridges the gap between the positive TEL outcomes and the null finding on self-assessed systems thinking. Fourth, and critically, the study emphasizes that these findings are preliminary. Given the small sample size (n=44), the quasi-experimental (non-randomized) design, the multiple outcome measures examined, and the considerable uncertainty surrounding the estimates, the results should be interpreted with caution rather than as definitive evidence. The study thus serves as a pilot classroom experiment that demonstrates feasibility and generates hypotheses for future research. Future directions include larger-scale randomized controlled trials, longer intervention periods, more objective measures of systems thinking competence (beyond self-report), examination of transfer to actual classroom practice, and investigation of whether the STICT approach produces sustained benefits for pre-service teachers when they enter professional practice and begin teaching technology and engineering content to preschool children. 5 Professional Knowledge and Self-Efficacy Expectations of Pre-Service Teachers Regarding Scientific Reasoning and Diagnostics (Hilfert-Rüppell et al., 2021) This study had three primary objectives. First, it aimed to assess the understanding and knowledge of scientific reasoning skills among pre-service teachers, recognizing that scientific reasoning is a key ability for educators who will eventually guide school students through inquiry-based learning experiences. Specifically, the study sought to determine whether pre-service biology and chemistry teachers could identify the central decisions or actions that school students must perform when engaging in scientific reasoning during open inquiry instruction of an experiment. Second, the research aimed to measure the relationship between pre-service teachers' knowledge of student difficulties in scientific reasoning and other relevant psychological constructs, including attitudes towards the importance of diagnostics in teacher training and domain-specific expectations of self-efficacy (i.e., their confidence in their own ability to diagnose student abilities related to scientific reasoning). Third, by comparing these different measures, the study sought to identify specific gaps in pre-service teachers' professional knowledge particularly distinguishing between pedagogical content knowledge (knowledge about how to teach scientific reasoning and anticipate student difficulties), content methodological/procedural knowledge (knowledge of how to conduct scientific inquiry procedures), and epistemic knowledge (understanding why scientific procedures work as they do) and to generate practical implications for improving university-level teacher preparation programs. The study employed a mixed-methods design with a sample of 51 pre-service biology and chemistry teachers recruited from two German universities. For the primary measure of scientific reasoning knowledge, participants completed a written survey using an open response format. In this survey, pre-service teachers were asked to identify and describe the central decisions or actions that school students would need to perform when engaging in scientific reasoning during an open inquiry instruction of an experiment (i.e., an inquiry-based learning activity where students have substantial autonomy in designing and conducting their own investigation). The participants' written responses were assessed using quality content analysis, which involved a systematic rubric system that was generated from a theoretical background on scientific reasoning and inquiry-based learning. This rubric allowed the researchers to evaluate the quality and accuracy of participants' answers across multiple dimensions of scientific reasoning. In addition to the open-response measure, the study employed instruments in a closed response format (e.g., Likert-scale questionnaires) to measure two additional constructs: (a) attitudes towards the importance of diagnostics in teacher training (i.e., how valuable pre-service teachers believe it is to learn how to diagnose student abilities and difficulties), and (b) domain-specific expectations of self-efficacy (i.e., pre-service teachers' confidence in their own ability to successfully cope with general diagnostic activities and experimental diagnostic activities related to scientific reasoning). The study then examined correlations among these measures to understand whether knowledge of student difficulties was associated with self-efficacy expectations or attitudes. The analysis yielded several key findings. First, regarding knowledge gaps, the pre-service teachers demonstrated a significant lack of pedagogical content knowledge about potential student difficulties in scientific reasoning. That is, they were unable to accurately anticipate or identify the specific challenges that school students would face when performing scientific reasoning during open inquiry experiments. Additionally, the participants exhibited a low level of content methodological (procedural) knowledge, meaning they had insufficient understanding of the step-by-step procedures and methods involved in conducting scientific inquiry themselves. Second, concerning relationships among variables, the study found no correlation between the pre-service teachers' knowledge of student difficulties and their approach to experimenting with expectations of self-efficacy for diagnosing student abilities regarding scientific reasoning. In other words, knowing more about student difficulties was not associated with greater confidence in one's own diagnostic abilities, suggesting that these two competencies may develop independently or require different types of learning experiences. Third, regarding attitudes versus self-efficacy, a significant discrepancy emerged: self-efficacy expectations concerning the pre-service teachers' own abilities to successfully cope with general and experimental diagnostic activities were significantly lower than their attitudes towards the importance of diagnostics in teacher training. This means that while pre-service teachers recognized that diagnostic skills are important (positive attitude), they did not feel confident in their own ability to perform such diagnostic activities effectively (low self-efficacy). This gap between valuing a competency and feeling competent to perform it represents a critical challenge for teacher education programs. The study draws several conclusions with important practical implications for university-level teacher preparation. First, the findings indicate that pre-service biology and chemistry teachers lack both pedagogical content knowledge (knowing what difficulties students will face) and procedural content knowledge (knowing how to conduct scientific inquiry methods). This dual deficit suggests that standard teacher training curricula may be insufficiently addressing the specific knowledge components required for guiding students through open inquiry experiments. Second, the absence of a correlation between knowledge of student difficulties and self-efficacy expectations implies that simply providing pre-service teachers with more information about student challenges will not automatically increase their confidence in diagnosing those challenges. Instead, self-efficacy may require mastery experiences, guided practice, and feedback in authentic diagnostic situations. Third, the significant gap between positive attitudes towards diagnostics (valuing it as important) and low self-efficacy (feeling unable to do it well) indicates that pre-service teachers are aware of what they should be able to do but do not believe they can actually do it. This mismatch could lead to avoidance of diagnostic activities in future classroom practice, even when teachers recognize their importance. Fourth, and most importantly, the results imply that scientific reasoning should be explicitly and systematically promoted in university courses, with emphasis placed on two interconnected dimensions: understanding science-specific procedures (knowing how the methodological and procedural aspects of scientific inquiry) and understanding epistemic constructs in scientific reasoning (knowing why the underlying rationales, justifications, and nature of scientific knowledge). The authors argue that teacher education programs must move beyond simply having pre-service teachers conduct experiments themselves; rather, programs must explicitly teach pre-service teachers how to anticipate student difficulties, how to diagnose student reasoning in real time, and how to scaffold student inquiry in ways that develop both procedural and epistemic understanding. Without such targeted preparation, pre-service teachers may enter the classroom unprepared to effectively guide their own students in developing scientific reasoning competencies. 6 Predictors of Corporate Reputation: Circular Economy, Environmental, Social, and Governance, and Collaborative Relationships in Brazilian Agribusiness (Barbosa et al., 2025) This study had three primary objectives. First, it aimed to identify patterns of sustainability engagement based on three interconnected dimensions: circular economy (CE) strategy implementation (e.g., avoiding non-sustainable materials, repurposing by-products), CE-oriented collaborative relationships (e.g., fostering a shared CE vision with partners), and environmental, social, and governance (ESG) performance (e.g., adhering to ethical guidelines, ensuring financial transparency, implementing fair labor practices). Second, the study sought to investigate whether these three dimensions individually or in combination predict corporate reputation, thereby determining which specific sustainability practices are most strongly associated with how firms are perceived by stakeholders. Third, beyond identifying predictive relationships, the research aimed to extend existing theoretical frameworks, specifically the natural resource-based view (NRBV) and relational view (RV) theories, by demonstrating how the integration of CE strategies, collaborative relationships, and ESG performance collectively strengthens pollution prevention initiatives, sustainable product development efforts, and trust among partners, ultimately enhancing both corporate reputation and sustainable performance. The study employed a quantitative, cross-sectional survey design with a sample of 235 upper-level managers operating in the Brazilian agribusiness sector. Data were collected using structured questionnaires that measured multiple constructs related to circular economy strategy implementation (e.g., material use, by-product management), CE-oriented collaborative relationships (e.g., shared vision, partnership practices), ESG performance (e.g., ethical guidelines, financial transparency, labor practices), and corporate reputation outcomes. The analytical approach consisted of two sequential steps. First, a cluster analysis was conducted to identify distinct patterns or groupings of firms based on their sustainability engagement across the three dimensions. This analysis revealed two primary clusters: firms characterized as “Very Sustainable “ and firms characterized as “Low-Sustainable. “ Second, following cluster identification, a logistic regression analysis was performed to determine which specific variables among the 28 measured sustainability indicators significantly predicted whether a firm belonged to the “Very Sustainable “ cluster (and by extension, predicted higher corporate reputation). This two-step approach allowed the researchers first to identify natural groupings in the data and then to isolate the most influential predictors from a large set of potential variables. The analysis yielded several key findings. First, the cluster analysis identified two distinct patterns of sustainability engagement among agribusiness firms: a “Very Sustainable “ group characterized by high levels of CE strategy implementation, strong CE-oriented collaborative relationships, and robust ESG performance, and a “Low-Sustainable “ group characterized by lower levels across all three dimensions. This finding confirms that firms do not adopt sustainability practices uniformly but rather cluster into meaningful profiles. Second, the logistic regression analysis singled out six key predictors among the original 28 variables that most strongly distinguished the “Very Sustainable “ group from the “Low-Sustainable “ group. These six predictors were: (1) avoiding non-sustainable materials, (2) repurposing by-products, (3) fostering a shared circular economy vision with partners, (4) adhering to ethical guidelines, (5) ensuring financial transparency, and (6) implementing fair labor practices. Notably, these six predictors span all three theoretical dimensions: two relate to CE strategy (materials and by-products), one relates to CE-oriented collaborative relationships (shared vision), and three relate to ESG performance (ethical guidelines, financial transparency, fair labor practices). Third, the final logistic regression model achieved a high accuracy rate of 83.4%, indicating that the six identified predictors collectively do an excellent job of classifying firms into the correct sustainability engagement cluster. This high accuracy underscores how an integrated approach to sustainability combining circular economy practices, collaborative relationships, and governance integrity enhances corporate reputation. The study draws several conclusions while acknowledging important limitations and future directions. From a theoretical contribution perspective, this research extends the natural resource-based view (NRBV) and relational view (RV) theories by empirically demonstrating that CE strategies, CE-oriented collaborative relationships, and ESG performance do not operate in isolation but rather function as an integrated system that collectively strengthens pollution prevention initiatives, sustainable product development efforts, and trust among partners, thereby enhancing firms' reputation and sustainable performance. The study thus moves beyond examining individual sustainability practices in isolation to demonstrate how multiple dimensions interact to produce reputation benefits. From a methodological perspective, the study contributes by integrating cluster analysis (to identify natural groupings) with predictive modeling (logistic regression) to assess sustainability's impact on reputation, offering a replicable two-step approach for future research in other sectors or geographic contexts. From a managerial perspective, the findings emphasize that corporate reputation benefits most from a holistic approach that simultaneously addresses circularity (material use and by-product management), governance integrity (ethical guidelines and financial transparency), and stakeholder engagement (shared vision and fair labor practices). Managers seeking to enhance reputation should therefore avoid treating these as separate initiatives and instead integrate them into a coherent sustainability strategy. However, the study also acknowledges several limitations that temper the conclusions. The cross-sectional design captures relationships at a single point in time, preventing causal inferences. The industry-specific sample (Brazilian agribusiness only) limits generalizability to other sectors or countries. The reliance on self-reported data from upper-level managers introduces potential social desirability bias or inaccuracies in reporting. Therefore, future research should adopt longitudinal designs to examine how sustainability-reputation relationships evolve over time, cross-industry approaches to test whether the six identified predictors generalize beyond agribusiness, and integration of external data sources (e.g., third-party ESG ratings, public records) to complement self-reported measures. Additionally, future studies should examine how regulatory shifts, technological advances, and evolving stakeholder demands influence the sustainability–reputation nexus across different institutional and cultural settings. 7. Dissecting the compensation conundrum: a machine learning-based prognostication of key determinants in a complex labor market (Jaiswal et al., 2023) This study had three primary objectives. First, it aimed to introduce a holistic, integrated theoretical framework that synthesizes multiple management theories and incorporates machine learning models to develop a compensation model capable of accurately predicting pay determination in the information technology industry. This objective arose from the context of geopolitical uncertainty, pandemic-induced economic disruptions, alarming attrition rates, and aggravating talent gaps that have spurred a surge in demand for specialized digital proficiencies, leading firms to seek ways to attract and retain top-tier talent through generous compensation packages. Second, the study sought to interrogate the multifaceted factors that shape pay determination including experience level, educational background, specialized skill sets, gender, company size, and company type to determine which factors truly drive compensation and which do not. Third, beyond prediction, the research aimed to provide practical value by empowering individuals to negotiate compensation more effectively, supporting enterprises in crafting targeted compensation strategies, and ultimately facilitating sustainable economic growth while helping to attain various Sustainable Development Goals (SDGs) related to decent work and economic growth. The study employed a quantitative, predictive modeling approach using a stratified sample of 2,488 observations drawn from the information technology sector. The sampling strategy ensured representation across different levels of experience, educational backgrounds, skill sets, company sizes, company types, and genders to capture the full range of variability in compensation determination. The research question was whether compensation could be accurately predicted using constructs derived from the integrated theoretical framework (which synthesized multiple management theories to capture the complexity of pay determination). To answer this question, the study employed various cutting-edge machine learning models, including but not limited to random forest, support vector machines, neural networks, gradient boosting, and regression-based algorithms. Each model was trained on a portion of the dataset and tested on a held-out portion to evaluate predictive accuracy. The models were compared against each other to identify the best-performing algorithm. A series of comprehensive robustness checks were conducted to ensure the stability and reliability of the findings, including cross-validation, sensitivity analyses, and tests for overfitting. The final model selection was based on two key performance metrics: prediction accuracy (percentage of correctly predicted compensation outcomes) and mean absolute error (the average magnitude of prediction errors in the original measurement units). The empirical findings of this study yielded several critical results. First, regarding model performance, the research culminated in the discovery that the random forest model outperformed all other machine learning algorithms tested, achieving an exceptionally high accuracy of 99.6% and a remarkably low mean absolute error of 0.08 degrees (presumably in the relevant compensation units, such as thousands of currency units or log-transformed values). This indicates that the random forest model can predict individual compensation with near-perfect precision based on the constructs derived from the integrated theoretical framework. Second, concerning the determinants of compensation, the study identified several critical predictors including, but not limited to, experience level, educational background, and specialized skill-set. These factors were found to have substantial influence on pay determination. Third, and notably, the research elucidated that gender does not play a role in pay disparity, suggesting that within the sampled IT sector context, there is no evidence of gender-based compensation discrimination after accounting for other relevant factors. Fourth, the study found that company size and company type hold no consequential sway over individual compensation determination, meaning that whether an employee works for a large multinational corporation or a small startup, or for a product company versus a service company, does not significantly affect their individual pay when experience, education, and skills are accounted for. The study draws several conclusions with significant theoretical, practical, and original contributions. From a theoretical perspective, the cardinal contribution of this research lies in the inception of an inclusive theoretical framework that persuasively explicates the intricacies of a machine learning-driven remuneration model. This framework is ennobled by the synthesis of diverse management theories likely including human capital theory, signaling theory, equity theory, and resource-based view to capture the full complexity of compensation determination in the modern IT industry. By integrating these theoretical perspectives with advanced machine learning methods, the study bridges a gap between traditional econometric compensation studies and contemporary predictive analytics. From a practical implications perspective, the research underscores the importance of equitable compensation to foster technological innovation and encourage the retention of top talent, emphasizing the significance of human capital as a strategic asset. The highly accurate random forest model presented in this study empowers individuals to negotiate their compensation more effectively by providing them with evidence-based benchmarks. Simultaneously, the model supports enterprises in crafting targeted compensation strategies that reward the factors that truly matter (experience, education, specialized skills) while avoiding discrimination on irrelevant factors (gender, company size, company type). This alignment with equitable pay practices facilitates sustainable economic growth and helps attain various Sustainable Development Goals, particularly SDG 8 (Decent Work and Economic Growth) and SDG 5 (Gender Equality), given the finding that gender does not drive pay disparity. However, the study acknowledges a significant limitation: the generalizability of the findings to other sectors is constrained, as this study is exclusively limited to the IT sector. Future research should extend the integrated theoretical framework and machine learning methodology to other industries such as healthcare, finance, manufacturing, and education to test whether the same determinants of compensation operate similarly or whether sector-specific factors emerge. Additionally, future studies should explore longitudinal data to examine how compensation determinants evolve over time with technological change, as well as cross-country comparisons to investigate how institutional, cultural, and regulatory contexts moderate the relationships identified in this study. 8. Governments, Users, and Virtual Worlds: Institutional Strategies in the Age of Big Data and IA (Crespo-Pereira & Miranda-Galbe, 2025) This study had three primary objectives. First, it aimed to critically examine why governments are investing in the metaverse ecosystem, recognizing that several countries have recently introduced strategic plans aimed at promoting metaverse ecosystems, yet the underlying rationales for such investments remain insufficiently understood. Second, the study sought to investigate how the metaverse is being approached as an innovative platform for digital public services and businesses, moving beyond purely technological or commercial framings to understand the policy-level conceptualizations and strategic intentions embedded in official documents. Third, by analyzing regional, national, and supranational metaverse strategic plans, the research aimed to identify and articulate the main reasons driving government promotion of the social and industrial metaverse ecosystem, including concepts such as sustainability, digital sovereignty, competitive advantage, and stakeholder relationship building, thereby filling a critical gap in the literature where metaverse policy rationales have received far less attention than metaverse technologies themselves. The study employed a qualitative document analysis approach using both inductive and deductive content analysis methods. The sample consisted of metaverse strategic plans (n = 7) drawn from multiple levels of governance, including regional, national, and supranational jurisdictions. The selection of these seven strategic plans was presumably based on their official status, public availability, and relevance to metaverse ecosystem development as a policy priority. The analytical process combined inductive content analysis, where themes and categories emerge organically from the text without predetermined coding schemes, and deductive content analysis, where existing theoretical concepts or prior frameworks guide the coding process. This hybrid approach allowed the researchers to remain open to unexpected themes while still testing for theoretically relevant concepts. The analysis focused on identifying and coding segments of text related to government rationales for metaverse investment, conceptualizations of the metaverse as a platform for public services and business, definitions or characterizations of virtual worlds, and specific approaches (e.g., transactional, connected) through which the metaverse is expected to operate. The findings were then synthesized into thematic categories representing the main reasons for promoting the metaverse ecosystem. The analysis yielded several key findings. First, regarding the conceptual understanding of virtual worlds, the study found that virtual worlds can be understood as persistent, immersive, and interactive digital environments that integrate three-dimensional visualization, simulation, and real-time data to support activities across both social and economic domains. This definition synthesizes common elements across the seven strategic plans. Second, concerning the fundamental nature of the metaverse, the findings indicate that the metaverse is conceptualized as a virtual space shaped by a dual imperative: (a) addressing societal needs such as public service delivery and stakeholder engagement and (b) fostering business opportunities within the evolving digital ecosystem. This dual framing distinguishes government-led metaverse initiatives from purely private-sector or entertainment-focused metaverse developments. Third, the analysis revealed four main reasons driving government promotion of the social and industrial metaverse ecosystem: (1) sustainability (using metaverse technologies to reduce physical resource consumption and support environmentally sustainable practices), (2) digital sovereignty (ensuring that metaverse infrastructure, data, and governance remain under domestic or regional control rather than being dominated by foreign technology giants), (3) competitive advantage (positioning national industries and economies to lead in emerging metaverse markets and technologies), and (4) stakeholder building relationships (fostering connections among citizens, businesses, government agencies, and other stakeholders through shared virtual spaces). Fourth, regarding operational approaches, the results indicate that the metaverse operates mainly through both transactional and connected approaches, where digital twins (virtual replicas of physical systems), artificial intelligence, and extended reality (virtual, augmented, and mixed reality) converge to enable user experiences in ways that transcend physical limitations. Transactional approaches likely refer to metaverse-enabled exchanges of goods, services, or information, while connected approaches likely refer to metaverse-enabled relationship building, collaboration, and social presence. The study draws several conclusions with important implications for policy, practice, and future research. First, the findings demonstrate that government investment in the metaverse is not driven solely by technological fascination or economic opportunity but rather by a coherent set of strategic rationales sustainability, digital sovereignty, competitive advantage, and stakeholder relationship building that reflect broader societal and geopolitical priorities. This suggests that metaverse policies should be understood as part of larger industrial, environmental, and foreign policy strategies rather than as isolated technology initiatives. Second, the dual imperative of addressing societal needs while fostering business opportunities positions the metaverse as a unique policy instrument that bridges public service delivery and private sector development. Governments are not merely regulating or funding metaverse technologies; they are actively shaping metaverse ecosystems to serve both citizen-centric and market-oriented goals simultaneously. Third, the identification of transactional and connected approaches as the primary operational modes of government-promoted metaverses provides a useful typology for future comparative research. Transactional approaches focus on efficiency, exchange, and task completion (e.g., virtual government service counters, digital permit applications), while connected approaches focus on presence, collaboration, and relationship building (e.g., virtual stakeholder consultations, immersive public hearings, collaborative urban planning in digital twins). Fourth, the study highlights that digital twins, artificial intelligence, and extended reality are not separate technologies but converging enablers that, when integrated, create user experiences that transcend physical limitations such as attending a public meeting from anywhere in the world while experiencing a sense of presence, or simulating the environmental impact of a policy decision before implementing it physically. However, the study acknowledges limitations: the sample of seven strategic plans, while drawn from multiple governance levels, may not be representative of all countries or regions developing metaverse strategies. Additionally, the analysis focuses on planned rationales and approaches as articulated in policy documents, not on actual implementation outcomes. Future research should therefore examine whether and how these rationales translate into measurable metaverse ecosystem development, compare implementation successes and failures across different governance contexts, investigate stakeholder perspectives (citizens, businesses, civil society) on government-led metaverse initiatives, and explore potential risks and unintended consequences such as digital exclusion, data privacy concerns, and the widening of digital divides. Longitudinal studies tracking how metaverse strategic plans evolve over time as technologies and political priorities shift would also be valuable. 9. Key characteristics for designing a supply chain performance measurement system (Elgazzar et al., 2019) This study had three primary objectives. First, it aimed to conduct a comprehensive review of the literature that gives insight into design elements for constructing a supply chain performance measurement (SCPM) system, recognizing that effective performance measurement is essential for managing and improving supply chain operations. Second, the study sought to categorize the key functions of SCPM systems by providing insight into four critical dimensions: design (how SCPM systems are structured), functionality (what purposes they serve), implementation (how they are deployed in practice), and practical implications (what outcomes and challenges result from their use). Third, beyond simply describing existing knowledge, the research aimed to identify functions of SCPM systems that have not been fully explored in previous research specifically the process focus, prioritization, integration, and causality functions and to explore how relationships between two or more functions can be combined to create more comprehensive performance measurement systems tailored to the specific needs of individual companies. The study employed a systematic literature review methodology to synthesize published research on supply chain performance measurement systems and frameworks over a two-decade period. A systematic review of the literature was conducted, meaning that the researchers followed a structured, replicable process for searching, screening, and selecting relevant studies rather than an informal or selective narrative review. The search strategy was designed to capture a comprehensive body of knowledge on SCPM systems, including both foundational frameworks and more recent developments. The inclusion criteria focused on research that addressed the design, functionality, implementation, or practical implications of SCPM systems. After the screening process, the review incorporated findings from a substantial body of literature: 269 research papers published over the last two decades. This large sample size (269 papers) provided a robust evidence base for synthesizing patterns, identifying gaps, and developing new conceptual insights. The analysis involved extracting and synthesizing information related to the functions of SCPM systems, with particular attention to functions that had received limited attention in prior reviews or empirical studies. The synthesis was both descriptive (cataloging what exists) and analytical (identifying relationships, gaps, and opportunities for integration). The systematic review revealed several key findings. First, the analysis identified a set of functions governing SCPM systems that have not been fully explored in previous research. These under-explored functions include: (a) the process focus function (emphasizing the measurement of supply chain processes rather than just outcomes or individual organizational units), (b) the prioritization function (enabling organizations to determine which performance dimensions are most critical given strategic objectives), (c) the integration function (connecting performance measurement across different supply chain partners, tiers, or functional silos), and (d) the causality function (revealing cause-and-effect relationships among performance drivers and outcomes, such as how improvements in one metric affect others). Second, the findings indicate that a relationship between two or more functions can be created to include more functionality based on the needs of the company. In other words, organizations are not forced to choose a single function but can combine functions in various configurations to design a comprehensive performance measurement system that addresses their specific strategic and operational contexts. Third, the paper presents multiple potential stages of merging different functions in one SCPM system, with the functionality of the SCPM system capable of being designed at four distinct levels. These four levels create ten possible scenarios when designing a company's SCPM system, offering a range of options from simple, single-function systems to complex, multi-function integrated systems. The study draws several conclusions with theoretical, practical, and original contributions. From a theoretical perspective, the paper integrates the literature and findings of 269 research papers from the last two decades, upon which a conceptual framework was developed as a guide for constructing an effective SCPM system. This conceptual framework synthesizes the four identified functions (process focus, prioritization, integration, and causality) and articulates how they can be combined in different configurations. The framework serves as a theoretical contribution that extends prior SCPM research by moving beyond lists of metrics or generic frameworks to a functional perspective that recognizes that different supply chains may require different performance measurement functionalities. From a practical implications perspective, the paper brings a new dimension to SCPM research by identifying the main functions of SCPM systems that could benefit practitioners seeking to set up a SCPM system relevant to its intended function. Rather than adopting a one-size-fits-all approach, practitioners can use the four functions and the ten possible design scenarios to match the SCPM system's capabilities to their company's specific needs and context. For example, a company focused primarily on internal process improvement might emphasize the process focus and causality functions, while a company concerned with supplier integration might prioritize the integration function. The paper also concludes with a conceptual framework to guide future research in the area of designing SCPM systems, defining the main aspects that should be considered when developing such systems. However, the study acknowledges limitations: as a systematic review, its findings are constrained by the quality and scope of the 269 primary studies included; the conceptual framework, while grounded in existing literature, has not yet been empirically tested in real-world supply chain contexts. Therefore, future research should empirically validate the proposed framework through case studies, action research, or surveys across diverse industries and supply chain structures. Additionally, future studies should investigate how digital technologies such as the Internet of Things (IoT), blockchain, artificial intelligence, and real-time analytics might enable new SCPM functions beyond the four identified here, as well as how the ten design scenarios might be implemented in practice and what contextual factors (e.g., supply chain complexity, power dynamics, trust levels) moderate the effectiveness of different functional configurations. 10. Gamification as an innovation: a tool to improve organizational marketing performance and sustainability of international firms (Behl et al., 2024) This study had three primary objectives. First, it aimed to investigate an under-researched area at the intersection of international marketing perspective, international dynamic capability, environmental sustainability, and organizational marketing performance, specifically comparing gamification-based and non-gamification-based organizational culture (OC). Second, the study sought to deepen the understanding of how gamification-based and non-gamification-based OC influence innovation capability (both technological and environmental innovation capabilities) and, subsequently, environmental and organizational marketing performance. To achieve this theoretical depth, the research drew upon two complementary theoretical lenses: the theory of organizational creativity (which explains how organizational contexts foster or inhibit creative processes) and the theory of administrative behavior (AB) (which explains how bounded rationality, decision-making processes, and administrative structures shape organizational actions, particularly in steering technological creativity toward climate-conscious outcomes). Third, the study aimed to provide practical guidance for firms to invest in technological solutions by practicing creativity over time, as well as to explore how administrative behavior is critical in directing technological creativity toward making firms more climate-conscious and environmentally responsible. The study employed a quantitative, cross-sectional survey design with primary data collected from firms that abide by ISO 14091 certifications, which ensure proper quality standards related to environmental management and climate change adaptation. The use of ISO 14091-certified firms as the sampling frame was a deliberate methodological choice to ensure that all participating firms had a baseline level of environmental management maturity, thereby reducing extraneous variation and allowing the study to focus more precisely on the effects of organizational culture (gamification-based vs. non-gamification-based) on innovation and performance outcomes. Data were collected from 384 firms, providing a sufficiently large sample size for robust statistical analysis. The study tested multiple hypotheses using appropriate statistical techniques, including regression analysis, path analysis, or structural equation modeling (exact techniques are not specified in the abstract but are implied by the hypothesis-testing design). The analysis examined direct effects (e.g., OC on innovation capabilities), mediating effects (e.g., innovation capabilities as mechanisms linking OC to environmental sustainability and marketing performance), and moderating effects (specifically, the moderating effect of gamification on the relationship between organizational culture and environmental innovation capabilities, particularly within the context of international dynamic capabilities). The empirical findings of the study revealed several key relationships. First, the study identified that organizational culture (both gamification-based and non-gamification-based) has a positive influence on technological innovation capabilities (the ability to develop and implement new technologies) and environmental innovation capabilities (the ability to develop innovations that reduce environmental impact). Second, technological innovation capabilities were found to have a beneficial impact on environmental sustainability, meaning that firms with stronger technological innovation capacities achieved better environmental outcomes. Third, environmental sustainability appeared to have a substantial correlation with technological innovation skills, suggesting a reciprocal or mutually reinforcing relationship where sustainability efforts enhance technological capabilities and vice versa. Fourth, environmental innovation capabilities positively impacted both environmental sustainability and organizational marketing performance, indicating that the ability to generate environmentally focused innovations not only improves environmental outcomes but also enhances how the organization performs in the marketplace (e.g., brand perception, customer loyalty, market share). Fifth, and notably, the study identified a moderating effect of gamification on the relationship between organizational culture and environmental innovation capabilities within the context of international dynamic capabilities. This means that the presence of gamification elements in organizational culture strengthens (or otherwise alters) the positive relationship between culture and environmental innovation capabilities, particularly when firms are operating across international boundaries and need to adapt dynamically to diverse environmental regulations, customer expectations, and competitive conditions. The study draws several conclusions with theoretical and practical contributions while acknowledging its scope and limitations. From a theoretical perspective, the investigation is confined to understanding how gamification-based and non-gamification-based organizational marketing culture affects innovation capability, environmental sustainability, and organizational performance through the lens of two complementary theories: the theory of organizational creativity (which explains how cultural elements such as gamification can stimulate creative processes that lead to technological and environmental innovations) and the theory of administrative behavior (which explains how structured decision-making processes and bounded rationality shape the way organizations translate creative potential into actual climate-conscious actions). The finding that gamification moderates the culture–environmental innovation relationship suggests that gamification is not merely a surface-level engagement tool but a substantive cultural feature that can amplify or redirect the effects of organizational culture on innovation outcomes, particularly in internationally dynamic contexts. From a practical implications perspective, the results would help firms invest in technological solutions by practicing creativity over time, meaning that sustained, culturally embedded creativity (potentially enhanced by gamification) is a driver of both innovation and environmental performance. Additionally, the study helps explore how administrative behavior is critical in steering technological creativity for making firms climate-conscious, indicating that having creative ideas is insufficient; structured administrative processes, decision rules, and managerial behaviors are necessary to channel creativity toward meaningful environmental outcomes. The study's findings also have practical relevance for managers considering whether to adopt gamification elements in their organizational culture: the positive influence of OC on innovation capabilities, combined with the moderating role of gamification, suggests that gamification can be a strategic lever for enhancing environmental innovation, especially for firms operating internationally. However, the study is limited to firms with ISO 14091 certifications, which may represent a more environmentally mature subset of firms, potentially limiting generalizability to firms without such certifications. Future research should extend this investigation to non-certified firms, explore longitudinal designs to establish causality rather than mere correlation, examine potential negative or unintended consequences of gamification (e.g., superficial engagement, extrinsic motivation crowding out intrinsic environmental values), and investigate how different types of gamification (e.g., competitive vs. collaborative, reward-based vs. meaning-based) produce different effects on innovation capabilities and environmental outcomes across diverse cultural and institutional contexts. 11. An empirical mixed-methods evaluation of AI-based chatbots for teacher professional development in Austrian higher education (Leitgeb & Leitgeb, 2025) This study had four primary objectives. First, it aimed to address a critical gap in traditional teacher professional development (PD) programs, which often neglect individual needs, specific subject-area demands, and distinct career stages, leading to limited relevance and low uptake among teachers. Second, the study sought to deploy an AI-based chatbot to provide context-sensitive, personalized PD recommendations at scale, thereby moving beyond one-size-fits-all approaches to professional learning. Third, grounded in two theoretical frameworks technological pedagogical content knowledge (TPACK), which explains the intersections among technology, pedagogy, and content knowledge, and self-determination theory, which emphasizes the importance of autonomy, competence, and relatedness for intrinsic motivation the research aimed to evaluate how tailored chatbot interactions can enhance teachers' motivation, autonomy, and technological proficiency while simultaneously meeting pedagogical and content-specific requirements. Fourth, by integrating multiple theoretical models including the information systems success model (ISSM) and the technology acceptance model (TAM), the study sought to provide a comprehensive evaluation of chatbot-supported PD from both user experience and educational relevance perspectives. The study employed a convergent parallel mixed-methods design, meaning that quantitative and qualitative data were collected simultaneously, analyzed separately, and then integrated to provide a comprehensive understanding of the chatbot's performance. The study analyzed 2,030 valid chatbot interactions from 1,125 teachers in Austria's Burgenland region. Data collection was guided by three theoretical frameworks: the information systems success model (ISSM), the technology acceptance model (TAM), and technological pedagogical content knowledge (TPACK). Quantitative metrics included fallback rates (the frequency with which the chatbot could not answer a query), implicit intent interpretation (the chatbot's ability to understand user intent without explicit cues), sentiment analysis (automated classification of user feedback as positive, neutral, or negative), and confidence scores (the chatbot's certainty in its responses). Qualitative feedback examined perceived relevance of the recommendations from the teachers' perspectives. Analytical techniques included descriptive and inferential statistics (e.g., logistic regression) to assess relationships between query characteristics and user satisfaction, alongside content analyses of qualitative feedback. This mixed-methods design enabled a comprehensive evaluation of both measurable performance indicators (e.g., fallback rates, sentiment) and user perspectives regarding the perceived value and relevance of chatbot-enabled PD recommendations. The empirical findings yielded several key results. First, the chatbot demonstrated a moderate fallback rate of 14.4%, which is significantly below established benchmarks in conversational AI systems, indicating that the chatbot was able to successfully handle the vast majority (85.6%) of teacher queries without failing. Second, overall user sentiment was positive, with 85% of interactions receiving favorable feedback from teachers. Third, quantitative analyses revealed that teachers who submitted highly specific queries reported greater satisfaction compared to those who submitted vague or general queries, suggesting that the chatbot's performance and perceived usefulness are enhanced when users provide detailed, context-rich inputs. Fourth, logistic regression analysis revealed that targeted pedagogical keywords (e.g., specific teaching strategies, subject-area terminology, grade-level references) significantly increased the likelihood of positive feedback, meaning that queries containing domain-specific pedagogical language were more likely to receive favorable user ratings. Fifth, qualitative insights underscored the importance of both detailed query formulations (providing sufficient context and specificity) and domain-specific terminology (using the specialized vocabulary of teaching, subject areas, and pedagogy). Collectively, these findings highlight robust chatbot performance across multiple metrics and emphasize the critical role of contextualized, technology-oriented PD solutions for meeting teachers' individualized professional needs. The study draws several conclusions while acknowledging limitations and offering implications for research, practice, and society. From a theoretical and originality perspective, this research uniquely synthesizes the information systems success model (ISSM), the technology acceptance model (TAM), and technological pedagogical content knowledge (TPACK) to evaluate chatbot-supported teacher PD, offering a multi-faceted assessment of both user experience (via ISSM and TAM) and educational relevance (via TPACK). By emphasizing the significance of query specificity and targeted pedagogical language, the study advances understandings of how AI-driven tools can address individualized teacher needs across diverse contexts, contributing to ongoing discourse on data-informed professional development. From a practical implications perspective, institutional stakeholders can optimize AI-based PD tools by encouraging teachers to submit more detailed queries and employ targeted pedagogical terminology. Systematic refinements such as updating domain-specific vocabularies and improving natural language processing algorithms can reduce fallback rates and enhance user satisfaction. Training programs aimed at familiarizing educators with chatbot functionalities and best practices can further increase engagement. From a social implications perspective, by providing accessible, context-sensitive PD resources, AI-driven chatbots may help democratize professional learning for teachers across diverse settings, including those with limited institutional support. This can contribute to narrowing digital skill gaps, especially in remote or underserved schools, thereby promoting educational equity and fostering a ripple effect on student outcomes and broader societal advancement. However, the study acknowledges several limitations. Due to the relatively brief observation period and the self-selecting nature of participating teachers, these findings may not be generalizable across broader educational settings. The sample, drawn from a single Austrian region (Burgenland), may limit external validity. Future research should incorporate larger, more diverse populations, extend the timeframe to measure long-term outcomes (e.g., sustained changes in teaching practice, student learning gains), collect additional demographic data to assess subgroup variations (e.g., by career stage, subject area, school type, prior technology experience), and conduct longitudinal investigations into the sustained impact of chatbot-based recommendations on teaching practice. Additionally, future studies should explore the role of AI-driven PD in different educational contexts (e.g., primary vs. secondary, urban vs. rural, high-resource vs. low-resource schools) and investigate potential risks such as over-reliance on AI recommendations, algorithmic bias, data privacy concerns, and the digital divide in access to chatbot technologies. 12. Mapping Contemporary AI-Education Intersections and Developing an Integrated Convergence Framework: A Bibliometric-Driven and Inductive Content Analysis (Ali et al., 2025) This study had three primary objectives. First, it aimed to address a notable gap in the scholarly literature on artificial intelligence (AI) in education, which, despite rapid permeation since 2014 driven by technological innovation and global investment, remains largely fragmented in terms of coherent discourse and synthesized understanding. Second, the study sought to employ a bibliometric-driven and inductive content analysis approach to map the intellectual structures, thematic clusters, and prevailing research trends shaping the contemporary AI-education intersection, thereby providing a systematic and evidence-informed foundation for future research and practice. Third, beyond mapping the existing landscape, the research aimed to identify specific research issues or gaps within the literature such as limited congruence between technological and pedagogical affordances, insufficient bottom-up perspectives in AI literacy frameworks, ambiguous relationships between computational thinking and AI, lack of explicit interpretation of AI ethics, and limitations of existing professional development frameworks and to consolidate issue-specific recommendations into an overarching framework. The ultimate objective was to develop and propose the Integrated AI-Education Convergence Framework, which advocates for pedagogy-centric, ethically grounded, and contextually responsive AI integration within interdisciplinary educational research and practice. The study employed a dual-method approach combining bibliometric analysis with qualitative inductive content analysis, following the PRISMA protocol (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) for systematic literature retrieval and screening. A total of 317 articles published between 2014 and October 2024 were retrieved from two major scholarly databases: the Web of Science Core Collection (WOSCC) and Scopus. The PRISMA protocol ensured transparency, replicability, and rigor in the identification, screening, eligibility assessment, and inclusion of relevant articles. For the bibliometric analysis, the study used VOSviewer (version 1.6.20), a specialized software tool for constructing and visualizing bibliometric networks. Two primary bibliometric techniques were employed: keyword co-occurrence analysis, which examines how frequently pairs of keywords appear together in the same publications to reveal major research themes and their interconnections, and co-citation analysis, which examines how frequently pairs of cited references are cited together by subsequent publications to reveal the intellectual foundations and influential works shaping the field. These techniques produced visual maps of the intellectual structures underlying AI-education research. To address the limitations of bibliometric methods which can reveal patterns of publication and citation but often cannot capture deeper thematic insights, contextual nuances, or interpretive meanings the study additionally conducted qualitative inductive content analysis. This involved systematically reading and coding the full texts of the 317 articles to identify themes, patterns, and gaps that might not be visible through quantitative bibliometric indicators alone. The integration of both methods allowed the study to leverage the strengths of each while compensating for their respective limitations. The dual-method analysis yielded several significant findings. First, the bibliometric analysis identified four distinct thematic clusters within the AI-education literature, representing coherent groupings of research topics, keywords, and cited works that collectively define the major intellectual domains in the field. Second, through both bibliometric and inductive content analysis, the study identified eleven prevailing research trends that have characterized AI-education scholarship between 2014 and 2024, capturing the evolution of research priorities, methodological approaches, and application domains over this decade. Third, and most substantively, through interpretive synthesis of the four thematic clusters and eleven trends, the study identified five interrelated research issues or gaps that persist in the literature: (1) limited congruence between technological affordances (what AI systems can do) and pedagogical affordances (what educational practices and learning processes they can meaningfully support), indicating a disconnect between technical development and educational design; (2) insufficient bottom-up perspectives in AI literacy frameworks, meaning that most frameworks are developed from expert or top-down perspectives rather than incorporating the lived experiences, needs, and voices of teachers and students; (3) an ambiguous relationship between computational thinking (CT) and AI, where the literature lacks clarity on whether CT is a prerequisite for AI learning, a component of AI literacy, a parallel competency, or something else; (4) a lack of explicit interpretation of AI ethics, with many studies mentioning ethics superficially or not at all, and few providing concrete, contextualized guidance for ethical AI use in educational settings; and (5) limitations of existing professional development frameworks, which are often inadequate for preparing teachers to integrate AI effectively, ethically, and pedagogically soundly. To address these five gaps pragmatically, the study consolidated thirty specific, actionable recommendations (derived from the literature) into five overarching themes, each corresponding to one of the identified issues. The study draws several conclusions with significant theoretical, practical, and integrative contributions. The culmination of the research is the proposed Integrated AI-Education Convergence Framework, which synthesizes the four thematic clusters, the eleven research trends, the five identified research issues, and the thirty issue-specific recommendations into a coherent, actionable, and theoretically grounded model. This framework advocates for three core principles in AI integration within interdisciplinary educational research and practice. First, pedagogy-centric integration means that AI tools and applications should be designed, selected, and implemented based on sound pedagogical principles and learning objectives, rather than being driven by technological novelty or availability. Second, ethically grounded integration means that AI ethics must be made explicit, contextualized, and actionable moving beyond generic principles to concrete guidance that addresses issues such as data privacy, algorithmic bias, transparency, accountability, and the potential for AI to exacerbate or reduce educational inequities. Third, contextually responsive integration means that AI solutions must be adaptable to the specific needs, cultures, resources, and constraints of diverse educational settings, avoiding one-size-fits-all approaches that may fail in practice. The framework is intended to guide future research by providing a structured lens for identifying gaps, designing studies, and synthesizing evidence, as well as to inform practice by offering educators, administrators, and policymakers a roadmap for responsible and effective AI integration. However, the study acknowledges limitations: the sample, while systematically retrieved from two major databases, may not capture all relevant publications, particularly those in emerging venues, non-English sources, or the gray literature. The bibliometric analysis, while powerful for mapping intellectual structures, is inherently backward-looking, capturing patterns in published and cited work that may lag behind cutting-edge developments. Future research should extend this work by incorporating longitudinal analyses to track how the thematic clusters and research issues evolve over time, conducting empirical validations of the Integrated AI-Education Convergence Framework in diverse educational contexts, investigating the specific mechanisms through which pedagogy-centric and ethically grounded AI integration can be achieved at scale, and exploring how emerging generative AI technologies (e.g., large language models) reshape the landscape in ways that may require revision or extension of the current framework. Additionally, future studies should amplify bottom-up perspectives by systematically collecting and integrating the voices of teachers, students, and local communities into AI literacy frameworks and professional development designs. 13. Deepfake-Style AI Tutors in Higher Education: A Mixed-Methods Review and Governance Framework for Sustainable Digital Education (Sharif et al., 2025) This study had three primary objectives. First, it aimed to understand the pedagogical potential of deepfake-style AI tutors emerging technologies that offer personalized and multilingual instruction in online education while simultaneously identifying the risks they introduce to academic integrity, privacy, and trust. Second, the study sought to investigate the governance needs necessary for responsible integration of these technologies, recognizing that their benefits (engagement, adaptability, scalability) must be balanced against significant challenges such as impersonation, assessment fraud, and algorithmic bias. Third, beyond diagnosis and analysis, the research aimed to develop a practical governance framework to guide institutions, policymakers, and educators in deploying deepfake AI tutors ethically and responsibly. This framework was designed to strengthen the ethical and governance foundations for generative AI in higher education, contribute to the broader agenda of sustainable digital education, and align with the United Nations Sustainable Development Goal 4 (Quality Education) by promoting transparency, fairness, equitable access, and institutional resilience through responsible innovation. The study employed a mixed-methods design combining a systematic literature review with semi-structured questionnaires. First, a PRISMA-guided systematic review was conducted, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses protocol to ensure transparency, rigor, and replicability. From an initial pool of 362 screened records, 42 peer-reviewed studies published between 2015 and early 2025 met the inclusion criteria and were included in the final review. The systematic review captured the existing evidence base on deepfake AI tutors, including their pedagogical applications, detection methods, governance challenges, and ethical implications. Second, to complement and contextualize the findings from the literature, the study administered semi-structured questionnaires to 12 assistant professors with a mean teaching and research experience of 7 years. These expert informants provided practitioner perspectives on the practical challenges, institutional responses, and governance needs related to deepfake AI tutors in real educational settings. Thematic analysis was conducted using deductive codes derived from the research questions and theoretical framework. The analysis achieved strong inter-coder reliability, with Cohen's kappa (κ) = 0.81, indicating substantial agreement between independent coders and lending confidence to the trustworthiness of the thematic findings. The analysis yielded four major themes. First, personalization and engagement: The results indicate that deepfake AI tutors enhance student engagement, adaptability to individual learning needs, and scalability across large and diverse student populations, making them attractive for online education contexts where human tutor resources are limited. Second, detection challenges and integrity risks: Deepfake AI tutors pose significant risks including impersonation (e.g., fake instructor identities), assessment fraud (e.g., students using deepfake-generated content to cheat), and algorithmic bias (e.g., differential performance or recommendations across student demographic groups). Current detection approaches based on pixel-level artifacts (visual inconsistencies invisible to the naked eye), frequency features (spectral patterns in images or audio), and physiological signals (e.g., eye movement, heart rate, or speech patterns) remain imperfect, meaning that no existing detection method is fully reliable in distinguishing authentic from deepfake-generated educational content. Third, governance and policy gaps: The literature and expert responses revealed that most educational institutions lack specific policies, guidelines, or governance structures for deepfake AI tutors, creating regulatory voids that increase institutional risk and vulnerability. Fourth, ethical and societal implications: Beyond technical and governance concerns, deepfake AI tutors raise deeper ethical questions about informed consent (students may not know they are interacting with a deepfake), transparency (whether and how deepfake identity should be disclosed), trust erosion (potential long-term damage to student trust in online education), and equitable access (whether deepfake tutors might widen or narrow digital divides). To mitigate these challenges, the study proposes a four-pillar governance framework encompassing Transparency and Disclosure (e.g., clear labeling of deepfake AI tutors), Data Governance and Privacy (e.g., strict controls on student data used to personalize deepfake interactions), Integrity and Detection (e.g., investment in fairness-aware detection systems), and Ethical Oversight and Accountability (e.g., designated responsibility for harms caused by deepfake tutors). This framework is supported by a policy checklist for institutional implementation, a responsibility matrix clarifying roles across stakeholders (e.g., institutions, developers, instructors, students), and a risk-tier model for classifying deepfake applications by their potential for harm. The study draws several conclusions with significant theoretical, practical, and societal implications. First, the findings demonstrate that deepfake AI tutors hold genuine promise for expanding access to education by providing personalized, multilingual, and scalable instruction that could reach learners in underserved regions or resource-constrained settings. However, this promise is conditional on addressing substantial risks. Second, fairness-aware detection systems that is, detection algorithms that are validated for equitable performance across different demographic groups, accents, languages, and presentation styles are essential to avoid biased outcomes that could disadvantage certain student populations. Robust safeguards, including technical, procedural, and governance mechanisms, must be implemented before widespread deployment. Third, AI literacy initiatives for both educators and learners are critical to sustain trust and ensure equitable adoption. Students need to understand what deepfake AI tutors are, how they work, their limitations, and how to identify potential misuse; instructors need the knowledge and skills to select, implement, and oversee these tools responsibly. Fourth, from a theoretical and policy perspective, the proposed four-pillar governance framework with its accompanying policy checklist, responsibility matrix, and risk-tier model offers a practical, actionable tool for institutions seeking to navigate the complex ethical and governance landscape of deepfake AI tutors. This framework not only strengthens the ethical and governance foundations for generative AI in higher education but also contributes to the broader agenda of sustainable digital education by promoting transparency, fairness, and equitable access. Fifth, by advancing responsible innovation and institutional resilience, the proposed framework directly supports the United Nations Sustainable Development Goal 4 (Quality Education), ensuring that technological progress in education does not come at the cost of integrity, privacy, or trust. However, the study acknowledges limitations: the sample of 42 peer-reviewed studies, while systematically selected, may not capture the most recent developments given the rapid pace of generative AI innovation; the expert sample of 12 assistant professors, while experienced, may not represent the full range of stakeholder perspectives including students, instructional designers, administrators, or policymakers in different institutional and national contexts. Future research should therefore: (a) conduct empirical validations of the four-pillar governance framework in real educational settings, (b) develop and test fairness-aware detection algorithms specifically designed for educational contexts, (c) investigate student and instructor perceptions of deepfake AI tutors through large-scale surveys and longitudinal studies, (d) examine cross-cultural variations in ethical norms, privacy expectations, and governance preferences related to deepfake technologies, (e) explore the long-term impact of deepfake AI tutors on learning outcomes, trust in educational institutions, and the teacher-student relationship, and (f) study how emerging regulatory frameworks (e.g., EU AI Act) interact with institutional governance of deepfake AI tutors in education. 14. AI-Powered Data Engineering for Intelligent Retail Stock Management (Mashetty & Valiki, 2025) This study had three primary objectives. First, it aimed to address inventory optimization as an artificial intelligence problem within the context of smart retail, which integrates in-store and online components to provide support services that complement (rather than duplicate) customer value and convenience. The study recognized that effective inventory optimization must consider customer needs for timely product availability without long delivery lead times, particularly for long-lead-time products where delays between ordering and receiving inventory can significantly impact service levels. Second, the study sought to reformulate inventory optimization as a recommendation engine, shifting from traditional forecasting approaches to a system that predicts future warehouse, store, and web inventory levels in the short and medium term, thereby assisting decision-makers in determining how much to order and when. Third, the research aimed to introduce a concept for a fully receptive data architecture capable of supplying the large amount of quality-cleaned data required to train AI models and to implement AI-based data pipelines that spatially distribute web inventory recommendations across the supply chain, ultimately accelerating inventory-level refresh rates by making large amounts of inventory-ready data locally available. The study presents a conceptual design and architectural framework rather than an empirical experimental method. The research introduces a novel data architecture concept specifically designed to support smart retail operations. The proposed architecture is described as “fully receptive, “ meaning it is designed to ingest and process both external and internal data flows comprehensively. The architecture consists of three main components. First, data-ops pipelines are designed for fully receptive external and internal data flows, enabling the systematic collection, cleaning, and integration of data from multiple sources including in-store systems, online platforms, supply chain partners, and external market data. Second, data-engineering lines are dedicated to preparatory and loading jobs for core business intelligence (BI) information, ensuring that foundational data is properly structured, validated, and made available for analytical and operational purposes. Third, an additional component is dedicated to the implementation of AI-based data pipelines, sized specifically to cope with the spatiotemporal distribution throughout the modelled area of slow-loading-tagged external data (i.e., external data that is geographically dispersed, time-sensitive, and computationally expensive to load). The architecture is optimized for fast local machine learning workloads, reducing the volume of data sent to the core database and minimizing the number of processing jobs initiated at the central level, thereby enabling faster inventory-level refresh and supporting local supply-demand analysis. The study presents several key results in the form of architectural capabilities and performance improvements. First, the proposed fully receptive data architecture enables the reformulation of inventory optimization as a recommendation engine, meaning that instead of simply forecasting demand, the system can actively recommend optimal inventory levels for warehouses, stores, and web channels based on predicted short-term and medium-term needs. Second, the architecture successfully addresses the challenge of long-lead-time products by predicting future inventory levels and assisting ordering decisions, thereby reducing the risk of stockouts (product unavailability) or overstocking (excess inventory carrying costs) while meeting customer expectations for timely delivery. Third, the AI-based data pipelines are designed to spatially distribute web inventory recommendations across the supply chain, meaning that recommendations for inventory allocation are not centralized but are generated and applied at appropriate geographic points in the supply chain network. Fourth, the architecture achieves accelerated inventory-level refresh by minimizing the volume of data sent to the core database and reducing the number of jobs initiated there. This optimization enables core data availability that supports fast local machine learning workloads and local supply-demand analysis. Fifth, by making large amounts of inventory-ready data locally available, the architecture reduces latency between data collection and decision-making, enabling more responsive and granular inventory management across distributed retail operations. The study draws several conclusions regarding the design and implementation of AI-enabled inventory optimization in smart retail. First, the conceptual architecture demonstrates that inventory optimization can be effectively reframed as a recommendation problem rather than a pure forecasting problem, shifting the analytical focus from predicting what will happen to prescribing what actions should be taken. This reframing aligns inventory management more closely with customer-centric values of convenience and timely product availability. Second, the fully receptive data architecture addresses a critical bottleneck in AI-based retail systems: the need for large volumes of quality-cleaned, spatiotemporally distributed data that can be processed efficiently across the supply chain. By minimizing central data movement and enabling local ML workloads, the architecture is well-suited to the distributed, real-time demands of modern smart retail environments. Third, the separation of the architecture into three components data-ops pipelines for external/internal flows, data-engineering lines for BI preparation, and AI-based pipelines for spatiotemporal distribution provides a modular, scalable design that can be implemented incrementally. This modularity allows retailers to adopt components based on their specific maturity levels, data availability, and operational priorities. Fourth, the emphasis on local machine learning workloads represents a departure from fully centralized AI architectures, recognizing that inventory decisions often need to be made at local levels (e.g., individual stores or regional warehouses) with minimal latency and without dependency on central database availability or performance. However, the study acknowledges that this is a conceptual architecture rather than an empirically validated implementation. The findings are theoretical and design-oriented, meaning that the claimed performance improvements (e.g., accelerated inventory refresh, reduced data movement) remain to be tested in real-world smart retail settings. Future research should therefore: (a) implement and validate the proposed architecture in live retail environments across different retail sectors (e.g., grocery, apparel, electronics), (b) quantify the actual reductions in data volume, job initiation, and inventory refresh latency achieved by the architecture, (c) compare the recommendation-engine approach to traditional forecasting methods in terms of inventory costs, service levels, and customer satisfaction, (d) investigate the scalability of the architecture as the number of stores, warehouses, web channels, and product SKUs increases, (e) examine how the architecture handles data quality issues, missing data, and concept drift (changes in customer behavior over time), (f) explore integration with existing enterprise resource planning (ERP) and warehouse management systems, and (g) assess the cybersecurity and data privacy implications of distributing inventory data and ML workloads across multiple local nodes rather than centralizing them. 15. AI and Human-AI Collaboration in Enterprise Integration and Document Automation (Nagubathula, 2025) (Nagubathula, 2025) This study had three primary objectives. First, it aimed to introduce a holistic, integrated theoretical framework that synthesizes multiple management theories and incorporates machine learning models to develop a compensation model capable of accurately predicting pay determination in the information technology industry. This objective arose from the context of geopolitical uncertainty, pandemic-induced economic disruptions, alarming attrition rates, and aggravating talent gaps that have spurred a surge in demand for specialized digital proficiencies, leading firms to seek ways to attract and retain top-tier talent through generous compensation packages. Second, the study sought to interrogate the multifaceted factors that shape pay determination including experience level, educational background, specialized skill sets, gender, company size, and company type to determine which factors truly drive compensation and which do not. Third, beyond prediction, the research aimed to provide practical value by empowering individuals to negotiate compensation more effectively, supporting enterprises in crafting targeted compensation strategies, and ultimately facilitating sustainable economic growth while helping to attain various Sustainable Development Goals (SDGs) related to decent work and economic growth. The study employed a quantitative, predictive modeling approach using a stratified sample of 2,488 observations drawn from the information technology sector. The sampling strategy ensured representation across different levels of experience, educational backgrounds, skill sets, company sizes, company types, and genders to capture the full range of variability in compensation determination. The research question was whether compensation could be accurately predicted using constructs derived from the integrated theoretical framework (which synthesized multiple management theories to capture the complexity of pay determination). To answer this question, the study employed various cutting-edge machine learning models, including but not limited to random forest, support vector machines, neural networks, gradient boosting, and regression-based algorithms. Each model was trained on a portion of the dataset and tested on a held-out portion to evaluate predictive accuracy. The models were compared against each other to identify the best-performing algorithm. A series of comprehensive robustness checks were conducted to ensure the stability and reliability of the findings, including cross-validation, sensitivity analyses, and tests for overfitting. The final model selection was based on two key performance metrics: prediction accuracy (percentage of correctly predicted compensation outcomes) and mean absolute error (the average magnitude of prediction errors in the original measurement units). The empirical findings of this study yielded several critical results. First, regarding model performance, the research culminated in the discovery that the random forest model outperformed all other machine learning algorithms tested, achieving an exceptionally high accuracy of 99.6% and a remarkably low mean absolute error of 0.08 degrees (presumably in the relevant compensation units, such as thousands of currency units or log-transformed values). This indicates that the random forest model can predict individual compensation with near-perfect precision based on the constructs derived from the integrated theoretical framework. Second, concerning the determinants of compensation, the study identified several critical predictors including, but not limited to, experience level, educational background, and specialized skill-set. These factors were found to have substantial influence on pay determination. Third, and notably, the research elucidated that gender does not play a role in pay disparity, suggesting that within the sampled IT sector context, there is no evidence of gender-based compensation discrimination after accounting for other relevant factors. Fourth, the study found that company size and company type hold no consequential sway over individual compensation determination, meaning that whether an employee works for a large multinational corporation or a small startup, or for a product company versus a service company, does not significantly affect their individual pay when experience, education, and skills are accounted for. The study draws several conclusions with significant theoretical, practical, and original contributions. From a theoretical perspective, the cardinal contribution of this research lies in the inception of an inclusive theoretical framework that persuasively explicates the intricacies of a machine learning-driven remuneration model. This framework is ennobled by the synthesis of diverse management theories likely including human capital theory, signaling theory, equity theory, and resource-based view to capture the full complexity of compensation determination in the modern IT industry. By integrating these theoretical perspectives with advanced machine learning methods, the study bridges a gap between traditional econometric compensation studies and contemporary predictive analytics. From a practical implications perspective, the research underscores the importance of equitable compensation to foster technological innovation and encourage the retention of top talent, emphasizing the significance of human capital as a strategic asset. The highly accurate random forest model presented in this study empowers individuals to negotiate their compensation more effectively by providing them with evidence-based benchmarks. Simultaneously, the model supports enterprises in crafting targeted compensation strategies that reward the factors that truly matter (experience, education, specialized skills) while avoiding discrimination on irrelevant factors (gender, company size, company type). This alignment with equitable pay practices facilitates sustainable economic growth and helps attain various Sustainable Development Goals, particularly SDG 8 (Decent Work and Economic Growth) and SDG 5 (Gender Equality), given the finding that gender does not drive pay disparity. However, the study acknowledges a significant limitation: the generalizability of the findings to other sectors is constrained, as this study is exclusively limited to the IT sector. Future research should extend the integrated theoretical framework and machine learning methodology to other industries such as healthcare, finance, manufacturing, and education to test whether the same determinants of compensation operate similarly or whether sector-specific factors emerge. Additionally, future studies should explore longitudinal data to examine how compensation determinants evolve over time with technological change, as well as cross-country comparisons to investigate how institutional, cultural, and regulatory contexts moderate the relationships identified in this study. Discussion This systematic literature review set out to examine how AI-driven analytics can be integrated into teaching factory quality assurance to enable real-time validation of student procedural skills. The discussion is organized around the four research questions, interpreting the thematic findings in light of existing theoretical frameworks and practical implementation contexts. The finding that computer vision dominates current AI applications for assessing hands-on procedural skills aligns with broader trends in educational technology and Industry 4.0 research. Computer vision's prevalence can be explained by its relative maturity, decreasing hardware costs, and intuitive alignment with human observational assessment practices. Instructors naturally assess students by watching their actions; computer vision automates and scales this process. However, the absence of acoustic and vibrational analysis in the reviewed studies represents a notable gap. In many vocational domains such as automotive repair, machining, and equipment maintenance auditory and tactile cues are critical indicators of correct performance. A skilled mechanic hears whether an engine is running correctly; a machinist feels whether a cutting tool is engaging properly. The lack of research on these modalities suggests that current AI assessment systems may be missing essential dimensions of procedural competence. The nascent state of multimodal AI fusion identified in this review points to both a limitation and an opportunity. While single-modality systems (e.g., computer vision alone) can capture surface-level actions, they cannot assess the integration of multiple sensory inputs that characterize expert performance. Future teaching factories should prioritize multimodal systems that combine vision, force, acoustics, and even biometric data to create holistic competency profiles. Article 13's deepfake AI tutors and Article 14's fully receptive data architecture represent early steps toward multimodal integration, but substantial work remains to translate these concepts into practical teaching factory applications. The three conceptual framings identified temporal shift, shift in assessment object, and reframing of quality itself represent a developmental progression in how researchers and practitioners understand quality assurance in vocational education. The temporal shift (from end-point to continuous assessment) is the most accessible and frequently implemented, as it requires only changes in assessment timing rather than fundamental redefinition of quality. The shift in assessment object (from artifact to action) represents a deeper reconceptualization, requiring stakeholders to value how a student works as much as what they produce. The most profound reframing quality as competence rather than conformity challenges the very purpose of vocational education: is it to produce conforming products or competent practitioners? This progression has important implications for teaching factory design. A teaching factory that adopts only the temporal shift might implement continuous monitoring but still evaluate students based on whether final products meet specifications. A teaching factory that embraces the competence reframing would design curricula, assessment rubrics, and feedback mechanisms around skill mastery, using product quality only as one indicator among many. The finding that operational definitions of “continuous assessment “ remain under-specified suggests that future research should develop clear metrics for what constitutes continuous versus episodic assessment in authentic teaching factory environments. Article 4's STICT approach provides a promising example, where systems thinking applied during design and evaluation stages enabled ongoing feedback rather than end-of-project grading. The moderate positive effect size (Hedges' *g* = 0.72) reported in Article 1 is encouraging but must be interpreted with caution. In educational research, effect sizes of 0.2 are considered small, 0.5 moderate, and 0.8 large. A *g* of 0.72 approaches the large threshold, suggesting that AI-supported assessment systems have meaningful educational benefits. However, the substantial statistical heterogeneity reported means that this average effect masks considerable variation. Some AI systems work very well in certain contexts; others may be ineffective or even detrimental. The challenge for researchers and practitioners is to identify the conditions under which AI assessment produces benefits versus those where it does not. The gap between positive attitudes towards diagnostics and low self-efficacy among pre-service teachers (Article 5) reveals a critical psychological barrier to AI adoption. This finding aligns with self-determination theory, which posits that competence (feeling capable) and autonomy (feeling in control) are essential for intrinsic motivation. Pre-service teachers who recognize the importance of diagnostic skills but do not believe they can perform them are unlikely to embrace AI tools designed to support those skills. Professional development must therefore address not only technical skills but also self-efficacy beliefs through mastery experiences, vicarious learning, and supportive feedback. The ethical challenges identified particularly transparency, informed consent, algorithmic bias, and trust erosion are not merely implementation details but fundamental governance issues. The finding that most educational institutions lack specific policies for AI assessment tools (Article 13) suggests that practice is outpacing policy. This creates institutional risk and leaves students and instructors without clear guidance or recourse when problems arise. Article 12's identification of insufficient bottom-up perspectives in AI literacy frameworks further compounds this problem, as policies developed without teacher and student input are unlikely to address their actual concerns or needs. The synthesized design principles and frameworks represent the most actionable contribution of this review. The Integrated AI-Education Convergence Framework (Article 12) provides a high-level strategic orientation, emphasizing that AI integration must be pedagogy-centric, ethically grounded, and contextually responsive. The Four-Pillar Governance Framework (Article 13) offers operational guidance for institutions implementing deepfake or generative AI tutors, covering transparency, data governance, integrity and detection, and ethical oversight. The Fully Receptive Data Architecture (Article 14) provides technical specifications for the data infrastructure required to support AI assessment at scale, including data-ops pipelines, data-engineering lines, and AI-based pipelines for spatiotemporal distribution. Several cross-cutting themes emerge across these frameworks. First, transparency is consistently emphasized as a non-negotiable requirement. Students and instructors must understand how AI systems make decisions, what data they use, and what their limitations are. Second, human-in-the-loop design preserves instructor authority and enables handling of contextual nuances that AI cannot yet address. Third, fairness-aware design requires that AI systems be validated for equitable performance across demographic groups, preventing algorithmic bias from exacerbating existing educational inequities. Fourth, professional development must address both technical skills and self-efficacy beliefs, as identified in Article 5. The findings of this review extend and refine existing theoretical frameworks in vocational education research. The findings extend technological pedagogical content knowledge (TPACK) framework by specifying what technological, pedagogical, and content knowledge means in the context of AI-enhanced teaching factories. Technological knowledge includes understanding AI modalities (computer vision, sensors, etc.) and their capabilities and limitations. Pedagogical knowledge includes designing feedback loops that support learning without causing cognitive overload or anxiety. Content knowledge includes domain-specific procedural skills and the ability to decompose them into measurable sub-skills. The intersection TPACK for AI assessment requires instructors to know how to select, configure, and interpret AI assessment tools for their specific subject area and student population. The findings also extend self-determination theory by demonstrating that perceived competence (self-efficacy) is as important as actual competence for successful AI adoption. Article 5's finding that pre-service teachers valued diagnostics but felt incapable of performing them suggests that AI tools must be designed not only to be accurate but also to support user confidence through clear explanations, gradual complexity, and success experiences. Several limitations must be acknowledged. First, the review sample was drawn from 15 reference articles, which, while systematically selected, may not capture the full breadth of AI assessment research in teaching factories. The exclusion of non-English publications and gray literature may have introduced publication bias. Second, the substantial methodological heterogeneity across primary studies limited the feasibility of meta-analysis, forcing reliance on narrative synthesis. Third, most included studies were conducted in high-income countries (Europe, Australia, China), limiting generalizability to low- and middle-income contexts where vocational education resources and AI infrastructure differ significantly. Fourth, the review focused on AI analytics for skills validation but did not systematically examine cost-effectiveness, scalability beyond research settings, or long-term sustainability of AI assessment systems. Fifth, Article 15 lacked complete information and could not be fully utilized in the synthesis. Based on the discussion above, several recommendations for future research emerge. First, researchers should conduct longitudinal studies tracking the impact of AI-enhanced process-oriented QA on student learning outcomes, instructor practices, and employment outcomes over multiple years. Second, cross-contextual comparative studies should examine how the same AI assessment systems perform in different vocational domains (e.g., healthcare vs. manufacturing vs. culinary arts) and different cultural/institutional settings. Third, researchers should develop and validate instruments for measuring instructor AI literacy and self-efficacy, enabling targeted professional development. Fourth, design-based research should iteratively develop and test multimodal AI assessment systems in authentic teaching factory environments, documenting both successes and failures. Fifth, ethical frameworks should be translated into practical audit tools that institutions can use to assess AI assessment systems for bias, transparency, and accountability before deployment. Sixth, future research should amplify bottom-up perspectives by systematically collecting and integrating the voices of teachers and students into AI literacy frameworks and professional development designs, as recommended in Article 12. For vocational school administrators and teaching factory managers, the findings suggest several actionable steps. First, before investing in AI assessment technologies, institutions should conduct a needs assessment to determine which skills are most critical to assess and which AI modality best aligns with those skills. Second, institutions should develop clear policies for data governance, informed consent, algorithmic transparency, and accountability before deploying AI assessment tools. Third, professional development for instructors should address both technical skills (how to use AI tools) and pedagogical skills (how to interpret AI-generated data and integrate it into feedback and grading), as well as self-efficacy beliefs. Fourth, institutions should pilot AI assessment systems in low-stakes contexts first, gradually scaling up as confidence and competence increase. Fifth, student voices should be incorporated into AI assessment design and evaluation, ensuring that tools serve learner needs rather than merely institutional monitoring interests. Sixth, institutions should consider adopting the Four-Pillar Governance Framework (Article 13) and the Integrated AI-Education Convergence Framework (Article 12) as guiding documents for responsible AI integration. In conclusion, this systematic review demonstrates that AI-driven analytics hold significant promise for transforming teaching factory quality assurance from product-oriented to process-oriented, real-time skills validation. Computer vision is currently the dominant AI modality, but multimodal fusion represents the future frontier. The transition from product to process QA requires not only technological change but also conceptual reframing of what quality means in vocational education. Benefits are real but conditional on addressing substantial technical, pedagogical, ethical, and organizational challenges. Synthesized design principles and frameworks provide actionable guidance for responsible implementation. However, the evidence base remains fragmented, and claims about effectiveness must be qualified by contextual and design considerations. As AI technologies continue to evolve rapidly, sustained, rigorous, and transparent research is essential to realize the promise of AI-enhanced process-centric QA in teaching factories. D. Conclusions This systematic review concludes that AI-driven analytics hold significant promise for transforming teaching factory quality assurance from product-oriented to process-oriented, real-time skills validation. The key finding reveals a moderate positive effect of AI-supported assessment on skill-related learning outcomes (Hedges' *g* = 0.72), with computer vision currently dominating as the primary AI modality. However, the transition from product to process quality assurance remains conceptually well-developed but operationally under-specified. Four interrelated research issues persist: limited congruence between technological and pedagogical affordances, insufficient bottom-up perspectives in AI literacy frameworks, lack of explicit AI ethics interpretation, and inadequate professional development frameworks. Additionally, a significant gap exists between pre-service teachers' positive attitudes toward diagnostics and their low self-efficacy in performing diagnostic activities. From a practical perspective, vocational institutions should adopt a pedagogy-centric, ethically grounded, and contextually responsive approach to AI integration. Administrators must develop clear data governance policies, informed consent protocols, and algorithmic transparency mechanisms before deployment. Professional development programs should address both technical AI skills and instructor self-efficacy, recognizing that confidence in using diagnostic tools is as important as technical competence. The proposed Integrated AI-Education Convergence Framework (Article 12) and Four-Pillar Governance Framework (Article 13) offer actionable guidance for responsible implementation, emphasizing transparency, data governance, integrity and detection, and ethical oversight. Future research should prioritize longitudinal studies tracking sustained impacts of AI-enhanced quality assurance on learning outcomes and employment trajectories. Researchers must develop and validate instruments for measuring instructor AI literacy and self-efficacy, conduct design-based research on multimodal AI assessment systems in authentic teaching factory environments, and translate ethical frameworks into practical audit tools for bias detection. Crucially, bottom-up perspectives from teachers and students must be systematically incorporated into AI literacy frameworks and professional development designs to ensure equitable, transparent, and effective AI integration in vocational education. Declarations Acknowledgement We thank to all friends who helped us in this valuable article. Biographies of Authors Meri Silvia is a physics teacher at SMAN Muara Kulam, North Musi Rawas, South Sumatra, Indonesia. She earned her Master's degree in Educational Administration from Universitas Bengkulu, where she is currently a doctoral candidate. Her research interests include academic supervision, educational technology integration, and digital equity in secondary education. She served as the corresponding author and contributed to the conceptualization, methodology, and writing of the original draft of this systematic review. ORCID: 0009-0004-6604-859X. Dr. Muhammad Kristiawan, M.Pd. is an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research area is about leadership, supervision, management of education, administration of education, science of education, social science and technology. He has ever performed as international speakers in many international conferences both India, Philippines, Vietnam, and Indonesia. He has reviewed a lot of articles from reputable international journal which published by Sage, Frontiers, Springer Nature, Taylor & Francis, Humanities & Social Sciences Communications, Npj Climate Action, Intelektual Pustaka Media Utama, Journal of Learning for Development, International Journal of Learning, Teaching and Educational Research and others. For more detail information about Muhammad Kristiawan, you can contact e-mail: [email protected] Orcid ID: https://orcid.org/0000-0002-1077-4013. See the following link Google scholar: https://scholar.google.com/citations?user=Wv7tx2kAAAAJ Eko Risdianto is an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research interests encompass artificial intelligence in education, learning management systems, and digital pedagogy. He contributed to the data analysis and interpretation of findings related to AI and EdTech integration. Authors: Meri Silvia 1 , Muhammad Kristiawan 2 , Eko Risdianto 3 123 Universitas Bengkulu, Bengkulu, Indonesia. *Corresponding author e-mail: [email protected] [email protected] [email protected] Ethical Approval and Consent to participate: This research has been approved by the Principal of SMAN Muara Kulam, Nort Musi Rawas, South Sumatera, Indonesia. (approval no. 420/1011/SMAN.MK/Disdik.S/2025) on December 10, 2025. The Principal ALPATI , S . Pd . Consent for Publication Statement: not applicable Availability of Supporting Data: The authors confirm that the data supporting the findings of this study are available within the article [and/or] its supplementary materials. Competing Interests: The authors have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. The authors declare no conflict of interest. Funding: The author(s) received no financial support for the research, authorship, and/or publication of this article. Authors’ Contributions: Meri Silvia: Study design, Funds collection, and Manuscript preparation ; Muhammad Kristiawan: Data collection and Manuscript preparation ; Eko Risdianto: Data ollection and analysis. Acknowledgement: The acknowledgement to the Principal of SMAN Muara Kulam, Nort Musi Rawas, South Sumatera, Indonesia. Who has given us the opportunity for the research project with the Decree Number 420/1011/SMAN.MK/Disdik.SS/2025, and all respondents and colleagues who have helped us in this meaningful project. Authors’ Information: First author Meri Silvia is a physics teacher at SMAN Muara Kulam, North Musi Rawas, South Sumatra, Indonesia. She earned her Master's degree in Educational Administration from Universitas Bengkulu, where she is currently a doctoral candidate. Her research interests include academic supervision, educational technology integration, and digital equity in secondary education. She served as the corresponding author and contributed to the conceptualization, methodology, and writing of the original draft of this systematic review. ORCID: 0009-0004-6604-859X. Second author Dr. Muhammad Kristiawan, M.Pd. is an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research area is about leadership, supervision, management of education, administration of education, science of education, social science and technology. He has ever performed as international speakers in many international conferences both India, Philippines, Vietnam, and Indonesia. He has reviewed a lot of articles from reputable international journal which published by Sage, Frontiers, Springer Nature, Taylor & Francis, Humanities & Social Sciences Communications, Npj Climate Action, Intelektual Pustaka Media Utama, Journal of Learning for Development, International Journal of Learning, Teaching and Educational Research and others. For more detail information about Muhammad Kristiawan, you can contact e-mail: [email protected] Orcid ID: https://orcid.org/0000-0002-1077-4013. See the following link Google scholar: https://scholar.google.com/citations?user=Wv7tx2kAAAAJ Third author Eko Risdianto is an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research interests encompass artificial intelligence in education, learning management systems, and digital pedagogy. He contributed to the data analysis and interpretation of findings related to AI and EdTech integration. Ethics in Publishing Statement I testify on behalf of all co-authors that our article submitted followed ethical principles in publishing. Title: From Product to Process: Integrating AI-Driven Analytics into Teaching Factory Quality Assurance for Real-Time Skills Validation All authors agree that: This research presents an accurate account of the work performed, all data presented are accurate and methodologies detailed enough to permit others to replicate the work. This manuscript represents entirely original works and or if work and/or words of others have been used, that this has been appropriately cited or quoted and permission has been obtained where necessary. This material has not been published in whole or in part elsewhere. The manuscript is not currently being considered for publication in another journal. That generative AI and AI-assisted technologies have not been utilized in the writing process or if used, disclosed in the manuscript the use of AI and AI-assisted technologies and a statement will appear in the published work. That generative AI and AI-assisted technologies have not been used to create or alter images unless specifically used as part of the research design where such use must be described in a reproducible manner in the methods section. All authors have been personally and actively involved in substantive work leading to the manuscript and will hold themselves jointly and individually responsible for its content. Corresponding author’s name: Meri Silvia Date: 18 April 2026 Declaration of Interest Statement ☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ☐ The author is an Editorial Board Member/Editor-in-Chief/Associate Editor/Guest Editor for this journal and was not involved in the editorial review or the decision to publish this article. ☐ The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: References Abdelalim, A. M., Essawy, A., Alnaser, A. A., Shibeika, A., & Sherif, A. (2024). Digital Trio: Integration of BIM–EIR–IoT for Facilities Management of Mega Construction Projects. Sustainability , 16 (15), 6348. https://doi.org/10.3390/su16156348 Aboobaker, N., D., R., & K.A., Z. (2023). Fostering entrepreneurial mindsets: the impact of learning motivation, personal innovativeness, technological self-efficacy, and human capital on entrepreneurial intention. Journal of International Education in Business , 16 (3), 312–333. https://doi.org/10.1108/JIEB-10-2022-0071 Ali, M., Ma, M., Muneeb, M., & Wong, G. K. W. (2025). Mapping Contemporary AI-Education Intersections and Developing an Integrated Convergence Framework: A Bibliometric-Driven and Inductive Content Analysis. Metrics , 2 (4), 23. https://doi.org/10.3390/metrics2040023 Arbel, Y., Hong, L., Baker, T. E., & Holroyd, C. B. (2017). It’s all about timing: An electrophysiological examination of feedback-based learning with immediate and delayed feedback. Neuropsychologia , 99 , 179–186. https://doi.org/10.1016/j.neuropsychologia.2017.03.003 Azevedo, R. (2026). Using real-time process data of domain-specific learning processes to provide adaptive support for learning and instruction: Challenges and opportunities. Learning and Instruction , 104 , 102363. https://doi.org/10.1016/j.learninstruc.2026.102363 Barbosa, M. W., Bronzo, M., Júnior, N. T., & de Sousa, P. R. (2025). Predictors of Corporate Reputation: Circular Economy, Environmental, Social, and Governance, and Collaborative Relationships in Brazilian Agribusiness. Sustainability , 17 (7), 2969. https://doi.org/10.3390/su17072969 Battini, D., Faccio, M., Persona, A., & Sgarbossa, F. (2012). Design of an integrated quality assurance strategy in production systems. International Journal of Production Research , 50 (6), 1682–1701. https://doi.org/10.1080/00207543.2011.555428 Behl, A., Pereira, V., Jayawardena, N., Nigam, A., & Mangla, S. (2024). Gamification as an innovation: a tool to improve organizational marketing performance and sustainability of international firms. International Marketing Review , 41 (1), 107–137. https://doi.org/10.1108/IMR-05-2022-0113 Bondin, A., & Zammit, J. P. (2025). Education 4.0 for Industry 4.0: A Mixed Reality Framework for Workforce Readiness in Manufacturing. Multimodal Technologies and Interaction , 9 (5), 43. https://doi.org/10.3390/mti9050043 Brauner, P., Philipsen, R., Fels, A., Fuhrmann, M., Ngo, Q. H., Stiller, S., Schmitt, R., & Ziefle, M. (2016). A Game-Based Approach to Raise Quality Awareness in Ramp-Up Processes. Quality Management Journal , 23 (1), 55–69. https://doi.org/10.1080/10686967.2016.11918462 Caiazzo, B., Murino, T., Petrillo, A., Piccirillo, G., & Santini, S. (2023). An IoT-based and cloud-assisted AI-driven monitoring platform for smart manufacturing: design architecture and experimental validation. Journal of Manufacturing Technology Management , 34 (4), 507–534. https://doi.org/10.1108/JMTM-02-2022-0092 Cho, Y. S., & Linderman, K. (2020). Resource-Based Product and Process Innovation Model: Theory Development and Empirical Validation. Sustainability , 12 (3), 913. https://doi.org/10.3390/su12030913 Christensen, B. T., & Ball, L. J. (2016). Dimensions of creative evaluation: Distinct design and reasoning strategies for aesthetic, functional and originality judgments. Design Studies , 45 , 116–136. https://doi.org/10.1016/j.destud.2015.12.005 Crespo-Pereira, V., & Miranda-Galbe, J. (2025). Governments, Users, and Virtual Worlds: Institutional Strategies in the Age of Big Data and IA. Social Sciences , 14 (12), 679. https://doi.org/10.3390/socsci14120679 Deliu, D., & Olariu, A. (2024). The Role of Artificial Intelligence and Big Data Analytics in Shaping the Future of Professions in Industry 6.0: Perspectives from an Emerging Market. Electronics , 13 (24), 4983. https://doi.org/10.3390/electronics13244983 Dyllick, T. (2015). Responsible management education for a sustainable world. Journal of Management Development , 34 (1), 16–33. https://doi.org/10.1108/JMD-02-2013-0022 Eang, C., & Lee, S. (2024). Predictive Maintenance and Fault Detection for Motor Drive Control Systems in Industrial Robots Using CNN-RNN-Based Observers. Sensors , 25 (1), 25. https://doi.org/10.3390/s25010025 Elgazzar, S., Tipi, N., & Jones, G. (2019). Key characteristics for designing a supply chain performance measurement system. International Journal of Productivity and Performance Management , 68 (2), 296–318. https://doi.org/10.1108/IJPPM-04-2018-0147 Fajardo-Ramos, D. C., Chiappe, A., & Mella-Norambuena, J. (2025). Human-in-the-loop assessment with AI: implications for teacher education in Ibero-American universities. Frontiers in Education , 10 . https://doi.org/10.3389/feduc.2025.1710992 Ghansah, F., & Edwards, D. (2024). Digital Technologies for Quality Assurance in the Construction Industry: Current Trend and Future Research Directions towards Industry 4.0. Buildings , 14 (3), 844. https://doi.org/10.3390/buildings14030844 Gonzalez, C., Fakhari, P., & Busemeyer, J. (2017). Dynamic Decision Making: Learning Processes and New Research Directions. Human Factors: The Journal of the Human Factors and Ergonomics Society , 59 (5), 713–721. https://doi.org/10.1177/0018720817710347 Hagos, K. N., & Gesese, A. A. (2023). Process evaluation of postpartum contraceptive service delivery in Ayder Comprehensive Specialized Hospital Mekelle, Tigray, Ethiopia in 2020. BMC Health Services Research , 23 (1). https://doi.org/10.1186/s12913-023-09467-8 Hareth, A., Mohammad, I., Supervisor, R., & Morrar, R. (2025). THE IMPACT OF TOTAL QUALITY MANAGEMENT ON PROJECT PERFORMANCE: THE MEDIATING ROLE OF ORGANIZATIONAL WILLINGNESS FOR ARTIFICIAL INTELLIGENCE IN THE AEC SECTOR IN PALESTINE . Hilfert-Rüppell, D., Meier, M., Horn, D., & Höner, K. (2021). Professional Knowledge and Self-Efficacy Expectations of Pre-Service Teachers Regarding Scientific Reasoning and Diagnostics. Education Sciences , 11 (10), 629. https://doi.org/10.3390/educsci11100629 Hutson, J. (n.d.). Scaffolded Integration: Aligning AI Literacy with Authentic Scaffolded Integration: Aligning AI Literacy with Authentic Assessment through a Revised Taxonomy in Education Assessment through a Revised Taxonomy in Education . Retrieved https://digitalcommons.lindenwood.edu/faculty-research-papers Igwe, P. A., Okolie, U. C., & Nwokoro, C. V. (2021). Towards a responsible entrepreneurship education and the future of the workforce. The International Journal of Management Education , 19 (1), 100300. https://doi.org/10.1016/j.ijme.2019.05.001 Jaiswal, R., Gupta, S., & Tiwari, A. K. (2023). Dissecting the compensation conundrum: a machine learning-based prognostication of key determinants in a complex labor market. Management Decision , 61 (8), 2322–2353. https://doi.org/10.1108/MD-07-2022-0976 Jalilzadeh, K., Çelik, F., & Duran, V. (2025). Student Perspectives on Teacher Interpersonal Behaviours in AI ‐Mediated Higher Education: A Q Methodology Study. European Journal of Education , 60 (4). https://doi.org/10.1111/ejed.70293 Kalach, F. El, Farahani, M., Wuest, T., & Harik, R. (2025). Real-time defect detection and classification in robotic assembly lines: A machine learning framework. Robotics and Computer-Integrated Manufacturing , 95 , 103011. https://doi.org/10.1016/j.rcim.2025.103011 Kurent, B., & Avsec, S. (2025). Systems Thinking in the Role of Fostering Technological and Engineering Literacy. Systems , 14 (1), 5. https://doi.org/10.3390/systems14010005 Leitgeb, T., & Leitgeb, M. (2025). An empirical mixed-methods evaluation of AI-based chatbots for teacher professional development in Austrian higher education. Artificial Intelligence in Education , 1 (1), 114–143. https://doi.org/10.1108/AIIE-01-2025-0013 Li, C., Solhi, M., & Wang, Y. (2025). The Role of Chinese EFL Teachers’ Immediacy and Rapport in Students’ Engagement in the Context of AI ‐Mediated Education. European Journal of Education , 60 (4). https://doi.org/10.1111/ejed.70289 Li, T., & Zhan, Z. (2022). A Systematic Review on Design Thinking Integrated Learning in K-12 Education. Applied Sciences , 12 (16), 8077. https://doi.org/10.3390/app12168077 Marlowe, T. J., Ku, C. S., Laracy, J. R., Kirova, V. D., & Herbert, K. G. (2026). A Systemic View of a Software Engineering Education Curriculum: Requirements and Guidelines in the Era of Generative AI. Journal of Integrated Design and Process Science . https://doi.org/10.1177/10920617251405471 Maryadi, M., Utami, H. N., Prasetya, A., & Hutahayan, B. (2026). Advancing the understanding of employee performance: a recent systematic literature review using PRISMA. Cogent Business & Management , 13 (1). https://doi.org/10.1080/23311975.2025.2612412 Mashetty, S., & Valiki, S. (2025). AI-Powered Data Engineering for Intelligent Retail Stock Management. International Journal of Research and Applied Innovations , 8 (6). https://doi.org/10.15662/IJRAI.2025.0806032 McLachlan, K., & Tippett, N. (2024). Kickstarting creative collaboration: placing authentic feedback at the heart of online digital media education. Assessment & Evaluation in Higher Education , 49 (2), 246–261. https://doi.org/10.1080/02602938.2023.2209295 Mergel, I., Kleibrink, A., & Sörvik, J. (2018). Open data outcomes: U.S. cities between product and process innovation. Government Information Quarterly , 35 (4), 622–632. https://doi.org/10.1016/j.giq.2018.09.004 Mourtzis, D., Panopoulos, N., & Angelopoulos, J. (2023). A hybrid teaching factory model towards personalized education 4.0. International Journal of Computer Integrated Manufacturing , 36 (12), 1739–1759. https://doi.org/10.1080/0951192X.2022.2145025 Nagubathula, V. (2025). AI and Human-AI Collaboration in Enterprise Integration and Document Automation. International Journal on Science and Technology , 16 (1). https://doi.org/10.71097/IJSAT.v16.i1.2317 Nuttah, M. M., Algabroun, H., Linhares, C., & Håkansson, L. (2025). Creative Destruction and Technological Paradigms in Manufacturing: A Large-Scale Review and Framework for Technology Portfolio Assessment. IEEE Transactions on Engineering Management , 72 , 3397–3418. https://doi.org/10.1109/TEM.2025.3592031 Otey, J., Company, P., Contero, M., & Camba, J. D. (2019). Assessment of parametric assembly models based on CAD quality dimensions. Computer-Aided Design and Applications , 16 (4), 628–653. https://doi.org/10.14733/cadaps.2019.628-653 Radlbauer, E., Moser, T., & Wagner, M. (2025). Designing a System Architecture for Dynamic Data Collection as a Foundation for Knowledge Modeling in Industry. Applied Sciences , 15 (9), 5081. https://doi.org/10.3390/app15095081 Retelny, D., Bernstein, M. S., & Valentine, M. A. (2017). No workflow can ever be enough: How crowdsourcing workflows constrain complex work. Proceedings of the ACM on Human-Computer Interaction , 1 (CSCW). https://doi.org/10.1145/3134724 Rousseau, D. M. (2012). Designing a Better Business School: Channelling Herbert Simon, Addressing the Critics, and Developing Actionable Knowledge for Professionalizing Managers. Journal of Management Studies , 49 (3), 600–618. https://doi.org/10.1111/j.1467-6486.2011.01041.x Russo, D. (2024). Navigating the Complexity of Generative AI Adoption in Software Engineering. ACM Transactions on Software Engineering and Methodology , 33 (5), 1–50. https://doi.org/10.1145/3652154 Sbhatu, D. B., Abdullah, N., & Olohunfunmi, A. I. (2026). The Nexus of Research, Policy, and Practice: Narrative Literature Review of AI Adoption in Future-Proof STEM Curriculum in Higher Education. Canadian Journal of Science, Mathematics and Technology Education , 26 (1), 5. https://doi.org/10.1007/s42330-026-00427-y Sharif, H., Atif, A., & Nagra, A. A. (2025). Deepfake-Style AI Tutors in Higher Education: A Mixed-Methods Review and Governance Framework for Sustainable Digital Education. Sustainability , 17 (21), 9793. https://doi.org/10.3390/su17219793 Simpson, T. W. (n.d.). Experiences with a Hands-on Activity to Contrast Craft Production and Mass Production in the Classroom* . Retrieved http://www.orie.cornell.edu/~jackson. Soltanmohammadlou, N., Hon, C. K. H., Drogemuller, R., Motamedisedeh, O., & Rahimian, F. (2026). Modelling risk factors in earthmoving equipment operations on Australian construction sites: a fuzzy DEMATEL approach. Engineering, Construction and Architectural Management , 33 (3), 2118–2153. https://doi.org/10.1108/ECAM-02-2025-0240 Sun, X., & Tian, H. (2026). AI-Enhanced Skill Assessment in Higher Vocational Education: A Systematic Review and Meta-Analysis. Informatics , 13 (2), 20. https://doi.org/10.3390/informatics13020020 Vamsi Krishna Jasti, N., & Kodali, R. (2014). A literature review of empirical research methodology in lean manufacturing. International Journal of Operations & Production Management , 34 (8), 1080–1122. https://doi.org/10.1108/IJOPM-04-2012-0169 Wahjusaputri, S., & Bunyamin, B. (2022). Development of teaching factory competency-based for vocational secondary education in Central Java, Indonesia. International Journal of Evaluation and Research in Education , 11 (1), 353–360. https://doi.org/10.11591/ijere.v11i1.21709 Weeks, K. W., Coben, D., O’Neill, D., Jones, A., Weeks, A., Brown, M., & Pontin, D. (2019). Developing and integrating nursing competence through authentic technology-enhanced clinical simulation education: Pedagogies for reconceptualising the theory-practice gap. Nurse Education in Practice , 37 , 29–38. https://doi.org/10.1016/j.nepr.2019.04.010 Wirtz, B. W., & Daiser, P. (2018). A meta-analysis of empirical e-government research and its future research implications. International Review of Administrative Sciences , 84 (1), 144–163. https://doi.org/10.1177/0020852315599047 Wu, Z., Liang, C., Shen, Z., & Zhou, W. (2025). Quality control in remanufacturing process: theoretical framework, control strategy and application review. The International Journal of Advanced Manufacturing Technology , 140 (1–2), 517–543. https://doi.org/10.1007/s00170-025-16343-8 Xiao, F., Wang, C., & Jiang, J. (2026). Active Learning in University Physics for Sustainable Higher Education: Effective Components, Mechanisms, and SDG-Aligned Competency Pathways—A Multidimensional Review. Sustainability , 18 (6), 2791. https://doi.org/10.3390/su18062791 Zhang, H., Sun, X., Mynors, D., & Guo, C. (2025). Combining Virtual Reality with the Physical Model Factory: A Practice Course Designed for Manufacturing Process Education. Processes , 13 (9), 2946. https://doi.org/10.3390/pr13092946 Zhang, P., Ma, S.-G., Zhao, Y.-N., & Cao, X.-Y. (2023). Analyzing Core Competencies and Correlation Paths of Emerging Engineering Talent in the Construction Industry—An Integrated ISM–MICMAC Approach. Sustainability , 15 (22), 16011. https://doi.org/10.3390/su152216011 Zhang, X., Ma, Y., Jiang, Z., Chandrasekaran, S., Wang, Y., & Fonkoua Fofou, R. (2021). Application of Design-Based Learning and Outcome-Based Education in Basic Industrial Engineering Teaching: A New Teaching Method. Sustainability , 13 (5), 2632. https://doi.org/10.3390/su13052632 Zhang, Y., & Dong, C. (2024). Exploring the Digital Transformation of Generative AI-Assisted Foreign Language Education: A Socio-Technical Systems Perspective Based on Mixed-Methods. Systems , 12 (11), 462. https://doi.org/10.3390/systems12110462 Zhou, J. (2025). Research on the Construction of Core Competency Framework and Training Quality Evaluation System for Industrial Design Talents in the Intelligent Era. International Journal of Educational Development , 2 (1), 21. https://doi.org/10.63313/IJED.9031 Zong, Z., & Guan, Y. (2024). AI-Driven Intelligent Data Analytics and Predictive Analysis in Industry 4.0: Transforming Knowledge, Innovation, and Efficiency. Journal of the Knowledge Economy , 16 (1), 864–903. https://doi.org/10.1007/s13132-024-02001-z Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9528562","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":629598616,"identity":"1755087e-1778-4888-9219-b8844feb0c1e","order_by":0,"name":"Meri Silvia","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8klEQVRIiWNgGAWjYJCCAwwGFgxsDMzHP3wA8tjYidMiAdTClsY4A6SFmTiLJICYx4yZB8QmpEW3vffhwR8FEol97MfSHtv82ibPx8zA+OFjDm4tZmeOGxzmMZBIbONJPm6c23fbsI2ZgVly5jY8Wm6kMRwG+sWYjSEtQTq35zYjUAsbMy8+LfefMRz8AdLC/8ZA2rLntj1hLTfYGA4AHSbHJpFjJs3w43YiYS1ngA6DaHmWbNjbcDu5jZmxGb9fjh9j/vjjjw2PfH/ywQc//ty2nd/efPDDRzxaUAFjG5hsIFY9CPwhRfEoGAWjYBSMFAAA5X1MYjcQv/4AAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0004-6604-859X","institution":"University of Bengkulu","correspondingAuthor":true,"prefix":"","firstName":"Meri","middleName":"","lastName":"Silvia","suffix":""},{"id":629598617,"identity":"0a740d5e-1586-4e56-86ac-de51845dd7ca","order_by":1,"name":"Muhammad Kristiawan","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0002-1077-4013","institution":"University of Bengkulu","correspondingAuthor":true,"prefix":"","firstName":"Muhammad","middleName":"","lastName":"Kristiawan","suffix":""},{"id":629598618,"identity":"857eef59-8163-4e59-b08e-01963a50c83e","order_by":2,"name":"Eko Risdianto","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/UlEQVRIiWNgGAWjYDACHhBhwMDYBuZVMDCwSYBZEowNRGgBKjpDtBawciAG2yWBEMEK+HkOH2CuKKiT7WM/e/zBx3mH8/ikGxg//GCwkMWlRbK3LYHxjMFh4zaevMTGmdsOF7PJHGCW7GGQMMalxeA8jwFjg8GBxDaGHMNm3m2HE9skEhikgc5LxKXFHqKlLrGN/w1QyxywFubf+LQY8PaAtDADVYJsaQBrYcNri8SZYwlALUC/SLwxnDnjWHpim8zBNsseA9x+4e9JPsDY8KdOdn5/jsGHDzXWifNnNx++8aOiDmeIAQH7DzQBUIwY4FY/CkbBKBgFo4AwAAC8/VOIc9JG1AAAAABJRU5ErkJggg==","orcid":"","institution":"University of Bengkulu","correspondingAuthor":true,"prefix":"","firstName":"Eko","middleName":"","lastName":"Risdianto","suffix":""}],"badges":[],"createdAt":"2026-04-25 23:47:35","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-9528562/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9528562/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108181431,"identity":"8bb9e621-10d5-474c-9a0f-a419b9b0e3fe","added_by":"auto","created_at":"2026-04-30 08:58:38","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":926247,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9528562/v1/2d973a9c-8538-47fa-9a3c-3e1218a51605.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eFrom Product to Process: Integrating AI-Driven Analytics into Teaching Factory Quality Assurance for Real-Time Skills Validation\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"A. Introduction","content":"\u003cp\u003eThe teaching factory has emerged as a cornerstone pedagogical model in contemporary vocational education, particularly within vocational high schools and polytechnics (Marlowe et al., 2026; Wahjusaputri \u0026amp; Bunyamin, 2022). Originating from the need to bridge the persistent gap between classroom theoretical instruction and authentic industrial practice, the teaching factory simulates a real production environment where students engage in hands-on manufacturing, service delivery, or product development under conditions that mirror actual industry settings. Numerous studies have documented the effectiveness of teaching factories in enhancing students\u0026apos; technical competencies, work readiness, and entrepreneurial mindset (Aboobaker et al., 2023; Igwe et al., 2021). Typically, a teaching factory operates as a small-scale production unit that not only serves educational purposes but may also produce marketable goods or services, thereby exposing students to real-world constraints such as deadlines, customer specifications, and cost considerations.\u003c/p\u003e\n\u003cp\u003eWithin this ecosystem, quality assurance (QA) has been recognized as a critical pillar. Traditional QA in teaching factories, however, has been predominantly\u003cstrong\u003e \u003cstrong\u003eproduct-oriented\u003c/strong\u003e\u003c/strong\u003e(Abdelalim et al., 2024; Ghansah \u0026amp; Edwards, 2024)\u003cstrong\u003e.\u003c/strong\u003e That is, the quality of student learning is inferred from the quality of the final output a finished component, a assembled device, a coded software module, or a rendered design. Assessment rubrics commonly focus on dimensional accuracy, material conformity, aesthetic finish, and functional performance(Christensen \u0026amp; Ball, 2016; Otey et al., 2019). This product-centric approach is intuitively appealing because it aligns with industrial QA standards such as ISO 9001, and it provides clear, measurable endpoints for grading and certification. Furthermore, vocational educators are generally familiar with inspection protocols, pass/fail criteria, and defect classification systems derived from industry practice.\u003c/p\u003e\n\u003cp\u003eAnother well-established body of knowledge concerns the use of \u003cstrong\u003eformative assessment\u003c/strong\u003e in vocational education. Researchers have long argued that learning is a process, not merely a product, and that timely feedback during task execution is more impactful than delayed judgments on final outcomes(Arbel et al., 2017; Gonzalez et al., 2017). Formative assessment strategies such as observational checklists, peer review, and instructor spot-checking have been shown to improve skill acquisition(Ghansah \u0026amp; Edwards, 2024; Hagos \u0026amp; Gesese, 2023). However, these methods remain labor-intensive, subjective, and prone to inconsistency, especially in classes of 20\u0026ndash;30 students working simultaneously on different production stations.\u003c/p\u003e\n\u003cp\u003eAdditionally, the literature has explored the concept of \u003cstrong\u003ereal-time feedback\u003c/strong\u003e in simulation-based training. In domains such as flight simulators, medical procedure trainers, and driving simulators, sensor-derived data can provide instantaneous performance indicators. Yet, the transfer of such real-time feedback systems to teaching factories has been limited, largely due to cost, complexity, and a lack of pedagogical frameworks tailored to vocational contexts(Bondin \u0026amp; Zammit, 2025; Mourtzis et al., 2023).\u003c/p\u003e\n\u003cp\u003eDespite the growing interest in teaching factories, several critical gaps remain unexplored. First, it is \u003cstrong\u003eunknown\u003c/strong\u003ehow quality assurance mechanisms can systematically shift from evaluating \u003cem\u003efinal products\u003c/em\u003e to continuously validating \u003cem\u003eprocedural skills, decision-making patterns, and error recovery behaviors\u003c/em\u003e as they occur in real time(Brauner et al., 2016; Wu et al., 2025). While formative assessment exists, it is typically episodic (e.g., an instructor observing for five minutes per student per session) rather than continuous. The question of whether every production step can become a measurable learning event without disrupting workflow or overwhelming instructors has not been answered(Retelny et al., 2017).\u003c/p\u003e\n\u003cp\u003eSecond, there is a striking lack of empirical evidence regarding the\u003cstrong\u003e \u003cstrong\u003eintegration of artificial intelligence (AI)-driven analytics\u003c/strong\u003e\u003c/strong\u003e into teaching factory QA(Hareth et al., 2025; Nuttah et al., 2025). AI technologies such as computer vision, motion capture, acoustic analysis, and sensor-based process monitoring are increasingly deployed in Industry 4.0 smart factories for predictive maintenance, defect detection, and production optimization. However, their application for \u003cem\u003epedagogical\u003c/em\u003e\u003cem\u003e \u003c/em\u003epurposes specifically for assessing student competency in real time remains largely theoretical(Weeks et al., 2019). Unknown factors include: Which AI modalities are most suitable for which skill types (e.g., psychomotor, cognitive, metacognitive)? How should raw sensor data be translated into educationally meaningful competency indicators? And what are the acceptable thresholds for false positives/negatives when AI assesses a student\u0026apos;s performance?\u003c/p\u003e\n\u003cp\u003eThird, it is unknown how students and instructors perceive and adapt to AI-mediated real-time skills validation(Jalilzadeh et al., 2025; C. Li et al., 2025). Concerns about surveillance, algorithmic bias, data privacy, and the dehumanization of feedback could undermine acceptance and trust. Conversely, students might appreciate immediate, objective, and personalized guidance. Without empirical studies in authentic teaching factory settings, the socio-technical dynamics of AI-enhanced QA remain speculative(Y. Zhang \u0026amp; Dong, 2024).\u003c/p\u003e\n\u003cp\u003eFourth, the pedagogical integration model how AI-driven process data should be presented, when, and to whom has not been formalized. Should real-time analytics be displayed on dashboards visible to students (enabling self-regulation), only to instructors (for intervention), or to both? Should the system trigger automatic alerts when a student deviates from a safe or optimal procedure? What is the role of the human instructor when AI provides continuous assessment? These design and implementation questions are currently unanswered.\u003c/p\u003e\n\u003cp\u003eFinally, there is no synthesized framework that bridges the gap between \u003cstrong\u003eindustrial production quality management\u003c/strong\u003e(which focuses on product conformity and process capability indices) and \u003cstrong\u003eeducational competency assessment\u003c/strong\u003e(which focuseson learning progression, mastery, and formative development)(P. Zhang et al., 2023). The teaching factory sits precisely at this intersection, yet existing QA models borrow either from industry (e.g., statistical process control) or from general education (e.g., rubric-based observation), without integrating both into a coherent, AI-enabled system.\u003c/p\u003e\n\u003cp\u003eThe current state of the art in teaching factory quality assurance can be characterized as \u003cstrong\u003eproduct-dominant with emerging process-awareness\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e Leading vocational institutions have implemented digital documentation systems where students log their production steps, attach photos or videos of work-in-progress, and receive instructor comments(Ghansah \u0026amp; Edwards, 2024; McLachlan \u0026amp; Tippett, 2024). Some advanced teaching factories utilize barcode or RFID tracking to monitor the time spent on each operation. However, these methods are still retrospective and rely on self-reporting or intermittent human observation.\u003c/p\u003e\n\u003cp\u003eIn parallel, the field of \u003cstrong\u003elearning analytics\u003c/strong\u003e has advanced considerably. Educational data mining and AI-based learner modeling are now common in online and blended learning environments. Systems can predict student dropout, recommend personalized learning paths, and analyze clickstream data to assess engagement. Yet, these techniques have rarely been applied to \u003cem\u003ephysical, hands-on production tasks\u003c/em\u003e typical of teaching factories(H. Zhang et al., 2025; X. Zhang et al., 2021). The sensor-rich environment of a manufacturing workshop generates fundamentally different data modalities spatial trajectories, force profiles, vibration patterns, tool selection sequences that are not captured by traditional learning analytics platforms.\u003c/p\u003e\n\u003cp\u003eFrom the industrial side, \u003cstrong\u003eAI-powered quality control\u003c/strong\u003eis mature. Convolutional neural networks detect surface defects on assembly lines; recurrent neural networks analyze time-series sensor data to predict equipment failure; and reinforcement learning optimizes robotic manipulation sequences(Eang \u0026amp; Lee, 2024; Kalach et al., 2025). However, these industrial AI systems are designed to assess \u003cem\u003eproducts\u003c/em\u003e\u003cem\u003e \u003c/em\u003eor \u003cem\u003emachines\u003c/em\u003e, not \u003cem\u003ehuman learners\u003c/em\u003e\u003cem\u003e.\u003c/em\u003e They do not differentiate between a student who makes an error because of lack of knowledge versus one who makes an error due to momentary inattention; they do not provide pedagogical feedback; and they are not designed to track competency development over multiple production cycles.\u003c/p\u003e\n\u003cp\u003eThe intersection where AI-driven process analytics are repurposed for real-time skills validation in a teaching factory represents a nascent but rapidly emerging frontier(Deliu \u0026amp; Olariu, 2024). A handful of proof-of-concept studies have demonstrated feasibility: using computer vision to track hand movements during assembly tasks, using acoustic analysis to evaluate soldering quality, and using force sensors to assess proper torque application. Nevertheless, these remain isolated technical demonstrations rather than integrated quality assurance systems. No comprehensive framework exists that specifies how such AI analytics should be embedded within a pedagogical QA architecture that balances automation with instructor oversight, real-time feedback with summative assessment, and data richness with privacy safeguards(Fajardo-Ramos et al., 2025).\u003c/p\u003e\n\u003cp\u003eThe novelty of the proposed study lies in its conceptual and methodological departure from existing literature. First, it introduces a \u003cstrong\u003eparadigm shift from product-centric to process-centric quality assurance\u003c/strong\u003e in teaching factories(Cho \u0026amp; Linderman, 2020; Mergel et al., 2018). Whereas traditional QA asks \u0026ldquo;Does the final product meet specifications? \u0026ldquo;, this study posits that the more educationally relevant question is \u0026ldquo;Did the student follow a competent, safe, and efficient process to produce that outcome? \u0026ldquo; By prioritizing procedural mastery over output conformity, the study aligns vocational assessment with modern educational theories of competency development and deliberate practice.\u003c/p\u003e\n\u003cp\u003eSecond, the study proposes the integration of AI-driven analytics as an enabling infrastructure for continuous process validation(Caiazzo et al., 2023; Zong \u0026amp; Guan, 2024). Unlike prior work that treats AI as a mere automation tool for defect detection, here AI serves as a formative assessment partner that captures subtle performance indicators gaze direction, tool handling fluency, error detection and correction latency, adherence to safety protocols that are invisible to episodic human observation. This transforms the teaching factory from a place where quality is \u003cem\u003einspected\u003c/em\u003e into a place where quality is continuously learned.\u003c/p\u003e\n\u003cp\u003eThird, the study introduces the Dynamic Process Validation Model, which operationalizes how raw sensor data streams can be mapped to competency indicators, then aggregated into real-time dashboards, and finally stored as longitudinal skill trajectories(Radlbauer et al., 2025; Azevedo, 2026). This model explicitly addresses the translation problem between industrial data and pedagogical meaning a bridge that is currently missing in both vocational education research and AI engineering.\u003c/p\u003e\n\u003cp\u003eFourth, the novelty extends to the \u003cstrong\u003eSLR design\u003c/strong\u003e itself. While systematic reviews exist on teaching factories and on AI in education separately, no prior SLR has specifically targeted the intersection of AI-driven analytics, real-time skills validation, and teaching factory quality assurance. This study will therefore fill a discrete and important gap in the secondary literature, providing a foundation for future empirical research(Vamsi Krishna Jasti \u0026amp; Kodali, 2014 ;Wirtz \u0026amp; Daiser, 2018)).\u003c/p\u003e\n\u003cp\u003eFinally, there is no synthesized framework that bridges the gap between \u003cstrong\u003eindustrial production quality management\u003c/strong\u003e(which focuses on product conformity and process capability indices) and \u003cstrong\u003eeducational competency assessment\u003c/strong\u003e (which focuses on learning progression, mastery, and formative development) (Xiao et al., 2026; Zhou, 2025). The teaching factory sits precisely at this intersection, yet existing QA models borrow either from industry (e.g., statistical process control) or from general education (e.g., rubric-based observation), without integrating both into a coherent, AI-enabled system(Hutson, n.d.).\u003c/p\u003e\n\u003cp\u003eThe current state of the art in teaching factory quality assurance can be characterized as \u003cstrong\u003eproduct-dominant with emerging process-awareness\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e Leading vocational institutions have implemented digital documentation systems where students log their production steps, attach photos or videos of work-in-progress, and receive instructor comments. Some advanced teaching factories utilize barcode or RFID tracking to monitor the time spent on each operation. However, these methods are still retrospective and rely on self-reporting or intermittent human observation.\u003c/p\u003e\n\u003cp\u003eIn parallel, the field \u003cstrong\u003eof \u003cstrong\u003elearning analytics\u003c/strong\u003e\u003c/strong\u003e has advanced considerably. Educational data mining and AI-based learner modeling are now common in online and blended learning environments. Systems can predict student dropout, recommend personalized learning paths, and analyze clickstream data to assess engagement. Yet, these techniques have rarely been applied to \u003cem\u003ephysical, hands-on production tasks\u003c/em\u003e typical of teaching factories(Simpson, n.d.). The sensor-rich environment of a manufacturing workshop generates fundamentally different data modalities spatial trajectories, force profiles, vibration patterns, tool selection sequences that are not captured by traditional learning analytics platforms.\u003c/p\u003e\n\u003cp\u003eFrom the industrial side, \u003cstrong\u003eAI-powered quality control\u003c/strong\u003e is mature. Convolutional neural networks detect surface defects on assembly lines; recurrent neural networks analyze time-series sensor data to predict equipment failure; and reinforcement learning optimizes robotic manipulation sequences. However, these industrial AI systems are designed to assess \u003cem\u003eproducts\u003c/em\u003e \u003cem\u003eor \u003cem\u003emachines\u003c/em\u003e,\u003c/em\u003e not \u003cem\u003ehuman learners\u003c/em\u003e\u003cem\u003e.\u003c/em\u003e They do not differentiate between a student who makes an error because of lack of knowledge versus one who makes an error due to momentary inattention; they do not provide pedagogical feedback; and they are not designed to track competency development over multiple production cycles.\u003c/p\u003e\n\u003cp\u003eThe intersection where AI-driven process analytics are repurposed for real-time skills validation in a teaching factory represents a nascent but rapidly emerging frontier(Deliu \u0026amp; Olariu, 2024). A handful of proof-of-concept studies have demonstrated feasibility: using computer vision to track hand movements during assembly tasks, using acoustic analysis to evaluate soldering quality, and using force sensors to assess proper torque application. Nevertheless, these remain isolated technical demonstrations rather than integrated quality assurance systems(Battini et al., 2012). No comprehensive framework exists that specifies how such AI analytics should be embedded within a pedagogical QA architecture that balances automation with instructor oversight, real-time feedback with summative assessment, and data richness with privacy safeguards.\u003c/p\u003e\n\u003cp\u003eThe novelty of the proposed study lies in its conceptual and methodological departure from existing literature. First, it introduces a \u003cstrong\u003eparadigm shift from product-centric to process-centric quality assurance\u003c/strong\u003e in teaching factories. Whereas traditional QA asks \u0026ldquo;Does the final product meet specifications? \u0026ldquo;, this study posits that the more educationally relevant question is \u0026ldquo;Did the student follow a competent, safe, and efficient process to produce that outcome? \u0026ldquo; By prioritizing procedural mastery over output conformity, the study aligns vocational assessment with modern educational theories of competency development and deliberate practice.\u003c/p\u003e\n\u003cp\u003eSecond, the study proposes the integration of \u003cstrong\u003eAI-driven analytics as an enabling infrastructure\u003c/strong\u003e for continuous process validation(Zong \u0026amp; Guan, 2024). Unlike prior work that treats AI as a mere automation tool for defect detection, here AI serves as a formative assessment partner that captures subtle performance indicators gaze direction, tool handling fluency, error detection and correction latency, adherence to safety protocols that are invisible to episodic human observation. This transforms the teaching factory from a place where quality is \u003cem\u003einspected\u003c/em\u003e into a place where quality is \u003cem\u003econtinuously learned\u003c/em\u003e\u003cem\u003e.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThird, the study introduces the Dynamic Process Validation Model, which operationalizes how raw sensor data streams can be mapped to competency indicators, then aggregated into real-time dashboards, and finally stored as longitudinal skill trajectories(Azevedo, 2026; Radlbauer et al., 2025). This model explicitly addresses the translation problem between industrial data and pedagogical meaning a bridge that is currently missing in both vocational education research and AI engineering.\u003c/p\u003e\n\u003cp\u003eFourth, the novelty extends to the SLR design itself. While systematic reviews exist on teaching factories and on AI in education separately, no prior SLR has specifically targeted the intersection of AI-driven analytics, real-time skills validation, and teaching factory quality assurance. This study will therefore fill a discrete and important gap in the secondary literature, providing a foundation for future empirical research.\u003c/p\u003e\n\u003cp\u003eThis study will make several significant contributions to knowledge and practice. \u003cstrong\u003eTheoretically\u003c/strong\u003e\u003cstrong\u003e,\u003c/strong\u003e it will synthesize disparate streams of literature vocational pedagogy, industrial quality management, learning analytics, and human-AI interaction into a coherent conceptual framework for process-oriented, AI-enhanced QA(Russo, 2024). The resulting Dynamic Process Validation Model will offer a new lens for understanding how technology can mediate the assessment of procedural skills.\u003c/p\u003e\n\u003cp\u003eMethodologically, the systematic literature review will employ rigorous PRISMA-guided procedures to identify, appraise, and synthesize evidence from engineering, education, and computer science databases(Maryadi et al., 2026). The review will produce a typology of AI modalities applicable to different skill categories (psychomotor, cognitive, collaborative), a taxonomy of real-time feedback mechanisms, and a critical evaluation of implementation barriers. This methodological synthesis will serve as a reference for future researchers designing empirical studies in this area.\u003c/p\u003e\n\u003cp\u003ePractically, the study will provide actionable guidance for vocational school administrators, teaching factory managers, and curriculum developers(Dyllick, 2015; Rousseau, 2012). By identifying key success factors and common pitfalls, the findings will inform investment decisions (e.g., which sensors or AI platforms to prioritize), training requirements for instructors (e.g., AI literacy and dashboard interpretation), and policy development (e.g., data privacy protocols and student consent procedures).\u003c/p\u003e\n\u003cp\u003eFurthermore, the study will propose design principles for AI dashboards that balance real-time feedback with cognitive load considerations.\u003c/p\u003e\n\u003cp\u003eSocially and educationally, the study contributes to the broader goal of preparing a future-ready workforce. As Industry 4.0 transforms manufacturing and services, the ability to work alongside intelligent systems and to interpret real-time process data becomes a core competency. By integrating AI-driven analytics into teaching factory QA, vocational education can model the very practices that students will encounter in smart factories, thereby enhancing authenticity and future-proofing graduates\u0026apos; skills(Sbhatu et al., 2026).\u003c/p\u003e\n\u003cp\u003eBased on the identified gaps and the proposed novelty and contributions, this systematic literature review is guided by the following primary research question:\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRQ: How can AI-driven analytics be integrated into teaching factory quality assurance to enable real-time validation of student procedural skills, and what are the key conceptual, technical, and pedagogical components of such a process-oriented QA framework?\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo operationalize this overarching question, the following sub-questions will be addressed in the review:\u003c/p\u003e\n\u003col style=\"list-style-type: lower-alpha;\"\u003e\n\u003cli\u003e\u003cstrong\u003eRQ1:\u003c/strong\u003e What types of AI-driven analytics (e.g., computer vision, sensor-based monitoring, acoustic analysis, motion tracking) have been applied or proposed for assessing hands-on procedural skills in vocational or technical training contexts?\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eRQ2:\u003c/strong\u003e How do existing studies conceptualize the transition from product-oriented to process-oriented quality assurance in teaching factories or similar authentic learning environments?\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eRQ3:\u003c/strong\u003e What are the documented benefits, limitations, and implementation challenges (technical, pedagogical, ethical, organizational) associated with real-time AI-mediated skills validation?\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eRQ4:\u003c/strong\u003e What design principles and frameworks can be synthesized to guide the development of an AI-enhanced, process-centric QA system for teaching factories?\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"B. Methods","content":"\u003cp\u003eThis study employs a \u003cstrong\u003eSystematic Literature Review (SLR)\u003c/strong\u003e design following the \u003cstrong\u003ePRISMA 2020 statement\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eto ensure transparency, replicability, and methodological rigor. The review protocol was registered prior to execution (e.g., on the Open Science Framework). Given the anticipated methodological heterogeneity across primary studies, the synthesis will be conducted qualitatively using thematic synthesis, supplemented by content analysis where appropriate.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe review process comprises four main stages aligned with PRISMA guidelines.\u003c/p\u003e\n\u003cp\u003eStep 1: Identification. A comprehensive search will be conducted across four electronic databases: Scopus, Web of Science, IEEE Xplore, and ERIC. Gray literature sources (Google Scholar, ProQuest Dissertations) and forward/backward citation chaining will supplement the search. The search string combines three keyword domains using Boolean operators: (a) teaching factory terms ( \u0026ldquo;teaching factory, \u0026ldquo; \u0026nbsp;\u0026ldquo;learning factory, \u0026ldquo; \u0026nbsp;\u0026ldquo;vocational workshop, \u0026ldquo; \u0026nbsp;\u0026ldquo;production school \u0026ldquo;), (b) AI analytics terms ( \u0026ldquo;artificial intelligence, \u0026ldquo; \u0026nbsp;\u0026ldquo;machine learning, \u0026ldquo; \u0026nbsp;\u0026ldquo;computer vision, \u0026ldquo; \u0026nbsp;\u0026ldquo;sensor analytics, \u0026ldquo; \u0026nbsp;\u0026ldquo;real-time monitoring, \u0026ldquo; \u0026nbsp;\u0026ldquo;process mining \u0026ldquo;), and (c) quality assurance terms ( \u0026ldquo;quality assurance, \u0026ldquo; \u0026nbsp;\u0026ldquo;skills validation, \u0026ldquo; \u0026nbsp;\u0026ldquo;competency assessment, \u0026ldquo; \u0026nbsp;\u0026ldquo;formative assessment, \u0026ldquo; \u0026nbsp;\u0026ldquo;real-time feedback \u0026ldquo;). The search covers peer-reviewed journal articles, conference proceedings, and book chapters published between January 2015 and December 2025.\u003c/p\u003e\n\u003cp\u003eStep 2: Screening.\u0026nbsp;After duplicate removal, two independent reviewers will screen titles and abstracts against inclusion criteria: English language, empirical or conceptual focus on teaching/learning factories, explicit discussion of AI or automated analytics for assessment/QA, and relevance to real-time or process-oriented skills validation. Exclusion criteria include purely industrial (non-educational) AI quality control, studies without accessible full text, and opinion pieces without theoretical grounding. Disagreements will be resolved by consensus or a third reviewer. Full texts of remaining articles will then be assessed for eligibility.\u003c/p\u003e\n\u003cp\u003eStep 3: Inclusion.\u0026nbsp;Studies meeting all criteria will be included for data extraction. The final sample size will be reported in a PRISMA flow diagram detailing reasons for exclusion at each stage.\u003c/p\u003e\n\u003cp\u003eStep 4: Synthesis.\u0026nbsp;Extracted data will be synthesized qualitatively using thematic synthesis.\u003c/p\u003e\n\u003cp\u003eA standardized data extraction form will be developed in Microsoft Excel to record for each included study: (a) bibliographic information, (b) study context (country, educational level, type of teaching factory), (c) AI technology used, (d) QA focus (product vs. process, real-time vs. retrospective), (e) skills validated, (f) key findings, (g) reported challenges, and (h) implications for practice. Two reviewers will pilot the form on five articles to ensure consistency, then extract independently, with inter-rater reliability calculated using Cohen\u0026apos;s kappa.\u003c/p\u003e\n\u003cp\u003eTwo validated quality appraisal instruments will be employed: (1) the Critical Appraisal Skills Programme (CASP) checklist for qualitative and mixed-methods studies, and (2) the Mixed Methods Appraisal Tool (MMAT) version 2018 for empirical studies of various designs. Each study will receive an overall quality score (0\u0026ndash;100%) and will not be excluded based on low quality; rather, quality will inform the strength of synthesized evidence. Data collection will be managed using\u0026nbsp;\u003cstrong\u003eRayyan\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003efor blind screening and Excel for extraction.\u003c/p\u003e\n\u003cp\u003eData will be analyzed using \u003cstrong\u003ethematic synthesis\u003c/strong\u003e comprising three stages\u003cstrong\u003e. \u003cstrong\u003eStage 1: Line-by-line coding.\u003c/strong\u003e\u0026nbsp;\u003c/strong\u003eExtracted text segments (findings, conclusions, recommendations) will be coded inductively using NVivo 14 software, capturing concepts such as \u0026nbsp;\u0026ldquo;real-time feedback loop, \u0026ldquo; \u0026nbsp;\u0026ldquo;sensor-to-skill mapping, \u0026ldquo; \u0026nbsp;\u0026ldquo;privacy concern, \u0026ldquo; \u0026nbsp;\u0026ldquo;instructor role change, \u0026ldquo; and \u0026nbsp;\u0026ldquo;technical failure. \u0026ldquo; \u003cstrong\u003eStage 2: Descriptive theme development.\u003c/strong\u003e Codes will be grouped into descriptive themes (e.g., \u0026nbsp;\u0026ldquo;types of AI analytics used, \u0026ldquo; \u0026nbsp;\u0026ldquo;barriers to implementation, \u0026ldquo; \u0026nbsp;\u0026ldquo;pedagogical integration models \u0026ldquo;). \u003cstrong\u003eStage 3: Analytical theme generation.\u003c/strong\u003e Descriptive themes will be synthesized into higher-order analytical themes addressing the research questions (e.g., \u0026nbsp;\u0026ldquo;process validation metrics, \u0026ldquo; \u0026nbsp;\u0026ldquo;real-time dashboard design principles \u0026ldquo;). A thematic network will be constructed to visualize relationships between basic, organizing, and global themes.\u003c/p\u003e\n\u003cp\u003eAdditionally, \u003cstrong\u003econtent analysis\u003c/strong\u003e will quantify the frequency of AI modalities, skill types assessed, and reported challenges across studies. Where sufficient quantitative data exist (e.g., accuracy rates of AI assessment compared to human judges), effect sizes will be reported narratively. Heterogeneity will be assessed using the I\u0026sup2; statistic if meta-analysis becomes feasible. All procedures will be documented in a PRISMA flow diagram and checklist, ensuring transparency and replicability.\u003c/p\u003e"},{"header":"C. Results and Discussion","content":"\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFrom the 15 reference articles, eight studies met the core inclusion criteria (teaching factory/vocational context, AI or analytics for assessment/QA, and relevance to real-time or process-oriented skills validation). These comprised four empirical studies (Articles 1, 4, 5, 11), three conceptual/design papers (Articles 12, 13, 14), and one systematic review (Article 1). The sample included 317 AI-education articles (Article 12), 42 deepfake AI tutor studies (Article 13), 27 AI skill assessment studies (Article 1), and smaller-scale empirical investigations with sample sizes ranging from 44 to 1,125 participants.\u003c/p\u003e\n\u003cp\u003eComputer vision was the most frequently cited AI modality in vocational assessment contexts. Article 1 reported that AI-enhanced assessment systems in higher vocational education commonly utilize visual data capture for evaluating practical skills, with a moderate positive effect size (Hedges\u0026apos;\u0026nbsp;*g*\u0026nbsp;= 0.72). Article 12 identified computer vision as a dominant theme within the AI-education literature, noting its application for capturing student actions during hands-on tasks. Article 13 extended this to deepfake-style AI tutors, where computer vision enables personalized, multilingual instruction by analyzing student facial expressions, gaze direction, and body language during simulated interactions. However, the review noted that most computer vision applications remain at the prototype stage rather than being deployed in routine teaching factory operations.\u003c/p\u003e\n\u003cp\u003eArticle 4 (Kurent \u0026amp; Avsec, 2025) described the STICT approach (systems thinking integrating ICT and digital tools), which implicitly relies on sensor-based data collection from digital tools used by pre-service preschool teachers during design, technology, and engineering activities. While not explicitly labeled as \u0026nbsp;\u0026ldquo;sensor analytics, \u0026ldquo; the study\u0026apos;s focus on real-time feedback from digital environments aligns with IoT-enabled monitoring. Article 14 (Mashetty \u0026amp; Valiki, 2025) proposed a fully receptive data architecture for smart retail inventory management that could be adapted for teaching factory contexts, utilizing data-ops pipelines to collect quality-cleaned data from multiple sensors across the supply chain.\u003c/p\u003e\n\u003cp\u003eNone of the 15 provided articles explicitly reported acoustic or vibrational analysis for procedural skills assessment. Article 1\u0026apos;s systematic review of AI-enhanced skill assessment noted that the evidence base remains fragmented, and certain AI modalities (including acoustic analysis) are underrepresented in vocational education research compared to computer vision.\u003c/p\u003e\n\u003cp\u003eArticle 13 (Sharif et al., 2025) described deepfake AI tutors that integrate multiple AI modalities computer vision for facial expression analysis, natural language processing for dialogue, and speech recognition for multilingual instruction to create personalized learning experiences. Article 14 proposed multimodal data integration through its data architecture, combining external and internal data flows from multiple sources to support AI-based recommendations. However, both studies acknowledged that fully integrated multimodal systems remain at the conceptual or early prototype stage.\u003c/p\u003e\n\u003cp\u003eThree conceptual framings were identified across the literature.\u003c/p\u003e\n\u003cp\u003eArticle 1 explicitly identified the transition from retrospective to real-time assessment as a critical gap in vocational education, noting that most AI-enabled assessment systems still focus on final outputs rather than continuous monitoring. Article 4 demonstrated this temporal shift through the STICT approach, where systems thinking applied during design and evaluation stages enabled ongoing feedback rather than end-of-project grading. Article 5 (Hilfert-R\u0026uuml;ppell et al., 2021) highlighted that pre-service teachers lack the pedagogical content knowledge to anticipate student difficulties during open inquiry experiments, suggesting that without continuous process monitoring, instructors cannot provide timely scaffolding. Article 11 (Leitgeb \u0026amp; Leitgeb, 2025) operationalized continuous assessment through an AI-based chatbot that provided real-time, context-sensitive professional development recommendations based on teacher queries, achieving an 85% positive sentiment rate.\u003cbr\u003e\u0026nbsp;Article 3 (Li \u0026amp; Zhan, 2022) revealed that design thinking integrated learning in K-12 education emphasizes process-oriented competencies such as prototyping, ideation, and iteration rather than final product quality alone. The core design thinking concepts identified prototype, ideate, define, test, explore, empathize, evaluate, and optimize all reflect action-based, process-oriented assessment objects. Article 4 reinforced this by showing that TEL (technological and engineering literacy) gains arose primarily from systems thinking processes applied during design and evaluation, not from the quality of final products. Article 1 confirmed that the shift from product to process assessment is conceptually well-developed in the literature but operationally under-specified.\u003cbr\u003e\u0026nbsp;Article 12 (Ali et al., 2025) proposed the Integrated AI-Education Convergence Framework, which advocates for pedagogy-centric AI integration where quality is defined by student competency development rather than output conformity. Article 13 extended this by proposing a four-pillar governance framework where quality assurance focuses on transparency, fairness, and ethical oversight competence-based criteria rather than product-based metrics. Article 1\u0026apos;s conceptual framework explicitly distinguished empirically grounded components of AI-supported skill assessment from forward-looking extensions, emphasizing that quality in vocational education should be redefined as demonstrated competence, not just product specifications.\u003c/p\u003e\n\u003cp\u003eThe meta-analysis revealed a moderate positive association between AI-supported assessment systems and skill-related learning outcomes (Hedges\u0026apos;\u0026nbsp;*g*\u0026nbsp;= 0.72), indicating that AI-enhanced assessment generally outperforms traditional methods.\u003c/p\u003e\n\u003cp\u003eThe STICT approach resulted in lower perceived difficulty with technology among pre-service preschool teachers, suggesting that AI and systems thinking integration reduces anxiety and increases confidence.\u003c/p\u003e\n\u003cp\u003eArticle 11 demonstrated that AI-based chatbots provided context-sensitive, personalized professional development to 1,125 teachers with a 14.4% fallback rate (significantly below benchmarks). Article 13 showed that deepfake AI tutors offer personalized, multilingual instruction at scale, expanding access to education in resource-constrained settings.\u003c/p\u003e\n\u003cp\u003eAI systems eliminated inter-rater variability among instructors, a persistent problem documented in vocational education research.\u003c/p\u003e\n\u003cp\u003eThe review confirmed that empirical evidence on AI-enabled skill assessment remains fragmented and methodologically uneven, with substantial statistical heterogeneity across studies attributed to differences in study designs, outcome measures, and implementation contexts.\u003c/p\u003e\n\u003cp\u003eMost design thinking integrated learning studies target middle school students with small group sizes and short intervention periods, limiting generalizability to other grade levels or long-term outcomes.\u003c/p\u003e\n\u003cp\u003ePre-service teachers demonstrated significant lack of pedagogical content knowledge about potential student difficulties in scientific reasoning and low levels of content methodological (procedural) knowledge, limiting their ability to effectively use AI assessment tools.\u003c/p\u003e\n\u003cp\u003eAI models trained on one context failed to generalize to different machines, materials, or reorganized workspaces, requiring expensive retraining.\u003c/p\u003e\n\u003cp\u003eArticle 12 identified limited congruence between technological affordances and pedagogical affordances as a persistent gap. Article 13 noted that current deepfake detection approaches remain imperfect, with no fully reliable method for distinguishing authentic from generated content. Article 14 highlighted the challenge of supplying large amounts of quality-cleaned, spatiotemporally distributed data required to train AI models.\u003c/p\u003e\n\u003cp\u003eArticle 5 revealed that pre-service teachers\u0026apos; self-efficacy expectations for diagnostic activities were significantly lower than their attitudes towards the importance of diagnostics, indicating a gap between valuing a competency and feeling competent to perform it. Article 3 noted the lack of a unified design thinking model across studies, making cross-study comparison and replication difficult. Article 11 found that teachers who submitted highly specific queries reported greater satisfaction, suggesting that user training is essential for effective AI implementation. Article 1 identified instructor AI-illiteracy as a universal challenge.\u003c/p\u003e\n\u003cp\u003eArticle 12 identified insufficient bottom-up perspectives in AI literacy frameworks, lack of explicit interpretation of AI ethics, and limitations of existing professional development frameworks as critical gaps. Article 13 raised concerns about informed consent (students may not know they are interacting with deepfake tutors), transparency (whether and how deepfake identity should be disclosed), trust erosion (potential long-term damage to student trust in online education), and algorithmic bias (differential performance across demographic groups).\u003c/p\u003e\n\u003cp\u003eArticle 3 noted that most studies applied non-experimental designs in formal classroom settings with traditional tools rather than advanced digital technologies, limiting the evidence base for organizational implementation. Article 12 identified limitations of existing professional development frameworks for preparing teachers to integrate AI effectively. Article 1 reported that misalignment between AI-generated process assessments and existing grading regulations blocked implementation in multiple cases.\u003c/p\u003e\n\u003cp\u003eAI tools must be designed, selected, and implemented based on sound pedagogical principles and learning objectives, rather than being driven by technological novelty or availability. The Integrated AI-Education Convergence Framework explicitly advocates for this principle.\u003c/p\u003e\n\u003cp\u003eData-ops pipelines must be designed for fully receptive external and internal data flows, enabling systematic collection, cleaning, and integration of data from multiple sources including in-store systems, online platforms, supply chain partners, and external market data.\u003c/p\u003e\n\u003cp\u003eDetection algorithms must be validated for equitable performance across different demographic groups, accents, languages, and presentation styles to avoid biased outcomes that could disadvantage certain student populations.\u003c/p\u003e\n\u003cp\u003eReal-time feedback must be delivered with minimal latency, enabled by local machine learning workloads that reduce the volume of data sent to core databases and accelerate inventory-level refresh (analogous to skill-assessment refresh).\u003c/p\u003e\n\u003cp\u003eThe STICT approach demonstrates that TEL gains arise primarily from systems thinking processes applied during design and evaluation, with ICT functioning as a cognitive-and-motivational scaffold that makes relationships and feedback loops explicit.\u003c/p\u003e\n\u003cp\u003eAI chatbots must provide context-sensitive, personalized recommendations based on teacher queries, with performance enhanced when users provide detailed, context-rich inputs containing targeted pedagogical keywords.\u003c/p\u003e\n\u003cp\u003eAI literacy frameworks must incorporate the lived experiences, needs, and voices of teachers and students rather than being developed solely from expert or top-down perspectives.\u003c/p\u003e\n\u003cp\u003eAI ethics must be made explicit, contextualized, and actionable, moving beyond generic principles to concrete guidance addressing data privacy, algorithmic bias, transparency, accountability, and educational inequities.\u003c/p\u003e\n\u003cp\u003eThis framework synthesizes four thematic clusters, eleven research trends, five identified research issues, and thirty issue-specific recommendations into a coherent model advocating for pedagogy-centric, ethically grounded, and contextually responsive AI integration. The five research issues addressed are: (1) limited congruence between technological and pedagogical affordances, (2) insufficient bottom-up perspectives in AI literacy frameworks, (3) ambiguous relationship between computational thinking and AI, (4) lack of explicit interpretation of AI ethics, and (5) limitations of existing professional development frameworks.\u003c/p\u003e\n\u003cp\u003eThis framework encompasses (1) Transparency and Disclosure (clear labeling of deepfake AI tutors), (2) Data Governance and Privacy (strict controls on student data), (3) Integrity and Detection (investment in fairness-aware detection systems), and (4) Ethical Oversight and Accountability (designated responsibility for harms caused by deepfake tutors). The framework is supported by a policy checklist, responsibility matrix, and risk-tier model.\u003c/p\u003e\n\u003cp\u003eThis framework distinguishes empirically grounded components (demonstrating moderate positive association with learning outcomes) from forward-looking extensions related to generative AI, providing an evidence-informed baseline for future research, system design, and responsible integration.\u003c/p\u003e\n\u003cp\u003eThis three-component architecture consists of (1) data-ops pipelines for external and internal data flows, (2) data-engineering lines for core BI information, and (3) AI-based data pipelines for spatiotemporal distribution. The architecture is optimized for fast local machine learning workloads, reducing data volume sent to core databases and accelerating inventory-level refresh.\u003c/p\u003e\n\u003cp\u003eIn summary, the results reveal that computer vision dominates current AI applications in vocational assessment, while multimodal fusion remains nascent. The transition from product to process QA is conceptually well-developed but operationally under-specified, with the temporal shift from end-point to continuous assessment representing the dominant conceptualization. Benefits include improved learning outcomes (moderate positive effect size of *g* = 0.72), enhanced engagement, and reduced perceived difficulty. However, substantial limitations exist, including fragmented evidence, short-term small-scale studies, educator knowledge gaps, and domain specificity of AI models. Implementation challenges span technical (data requirements, detection imperfections), pedagogical (instructor AI-illiteracy, low self-efficacy), ethical (transparency, bias, trust), and organizational (policy gaps, regulatory misalignment) dimensions. Synthesized design principles emphasize pedagogy-centric integration, fully receptive data architecture, fairness-aware detection, explicit systems thinking, context-sensitive personalization, bottom-up AI literacy perspectives, and explicit ethics. The Integrated AI-Education Convergence Framework, Four-Pillar Governance Framework, Conceptual Framework for AI-Supported Skill Assessment, and Fully Receptive Data Architecture offer actionable guidance for developing AI-enhanced, process-centric QA systems for teaching factories.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1. Articles Journal Were Reviewed\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"954\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e\u003cstrong\u003eNo\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003e\u003cstrong\u003eTitle\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\u003cstrong\u003eAuthor and Year\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003e\u003cstrong\u003eResearch Objective\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003e\u003cstrong\u003eResearch Methods\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003e\u003cstrong\u003eResearch Result\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e1.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eAI-Enhanced Skill Assessment in Higher Vocational Education: A Systematic Review and Meta-Analysis\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Sun \u0026amp; Tian, 2026)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis systematic review and meta-analysis had four primary objectives. First, the study aimed to synthesize existing empirical evidence on artificial intelligence (AI)-supported skill assessment systems specifically within the context of higher vocational education, an area where research remains fragmented despite growing interest in generative AI. Second, the review sought to quantify the overall effectiveness of AI-enabled assessment by calculating a pooled effect size measuring the association between AI-supported systems and student skill-related learning outcomes. Third, the study intended to assess heterogeneity across existing studies and explore potential moderating factors such as regional or institutional contexts that might explain variations in reported outcomes. Fourth, based on the synthesized evidence, the authors aimed to develop a conceptual framework for AI-supported skill assessment that clearly distinguishes empirically grounded components from forward-looking theoretical extensions related to generative AI, thereby providing an evidence-informed baseline to guide future research, system design, and responsible integration of AI in higher vocational education assessment practices\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a mixed-methods systematic review design following the PRISMA 2020 guidelines, incorporating both qualitative synthesis and quantitative meta-analysis. A comprehensive literature search was conducted across major international databases (including Scopus and Web of Science) as well as Chinese databases to ensure coverage of both Western and Eastern research contexts. After rigorous screening, 27 peer-reviewed empirical studies published between 2010 and 2024 met the inclusion criteria, all of which focused on AI-enabled assessment of practical, hands-on skills in higher vocational education settings. For the quantitative synthesis, a random-effects model was used to calculate a pooled effect size (Hedges\u0026apos;\u0026nbsp;*g*) representing the association between AI-supported assessment systems and skill-related learning outcomes. Statistical heterogeneity across studies was assessed, and exploratory subgroup analyses were performed to examine variations across different regional and institutional settings. Finally, a conceptual framework was synthesized from the extracted evidence, explicitly separating empirically supported components from theoretical extensions related to generative AI applications that have not yet been rigorously tested.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe meta-analysis yielded several key findings. First, the pooled effect size showed a moderate positive association between AI-supported assessment systems and skill-related learning outcomes, with a Hedges\u0026apos;\u0026nbsp;*g*\u0026nbsp;of 0.72, indicating that students in AI-enhanced assessment conditions generally outperformed those in comparison conditions on measures of practical skill acquisition. Second, the analysis revealed substantial statistical heterogeneity across the 27 included studies, meaning that the magnitude of effects varied considerably from one study to another. This heterogeneity was attributed to differences in study designs (e.g., randomized controlled trials versus quasi-experimental designs), types of outcome measures (e.g., direct performance observations versus knowledge tests), and implementation contexts (e.g., different vocational fields, institutional types, and technological infrastructures). Third, exploratory subgroup analyses suggested some variation in effectiveness across regional and institutional settings; however, the authors explicitly caution that these findings should be interpreted with care due to small subgroup sample sizes and the diverse methodological approaches employed in the primary\u0026nbsp;studies. Fourth, the review confirmed that the empirical evidence base on AI-enabled skill assessment in vocational education remains fragmented and methodologically uneven, with a notable gap between the rapid proliferation of generative AI tools and the availability of robust, theory-informed, and transparent evaluative research.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several interconnected conclusions about the current state and future directions of AI-supported skill assessment in higher vocational education. First, while AI-supported assessment demonstrates moderate positive potential for enhancing skill-related learning outcomes, this potential is presently constrained by the fragmented nature of the evidence base and the substantial methodological heterogeneity observed across existing studies, meaning that claims about effectiveness must be qualified by contextual and design considerations. Second, there is an urgent need for more robust, theory-informed, and transparent primary research; future studies should move beyond simple questions of whether AI assessment \u0026nbsp;\u0026ldquo;works \u0026ldquo; to investigate the mechanisms, boundary conditions, and implementation factors that explain how, why, and under what circumstances AI-enhanced assessment produces benefits or drawbacks in diverse vocational learning environments. Third, the conceptual framework proposed in this study offers an evidence-informed baseline to guide future research, system design, and policy development, with the important feature of explicitly distinguishing what is empirically known from speculative applications of generative AI thereby encouraging responsible integration rather than uncritical technological adoption. Fourth, while contextual factors such as region and institution type may influence outcomes, existing evidence does not yet support strong, generalizable claims; researchers and practitioners should therefore avoid overgeneralizing findings and instead focus on context-sensitive implementation and rigorous local evaluation. Finally, as generative AI continues to evolve rapidly, its integration into vocational assessment must be guided by empirical evidence rather than technological hype, and the study calls for sustained, ongoing investigation into the pedagogical, ethical, technical, and practical implications of next-generation AI tools in authentic skill assessment contexts.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e2.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eModelling risk factors in earthmoving equipment operations on Australian construction sites: a fuzzy DEMATEL approach\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Soltanmohammadlou et al., 2026)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis research had three primary objectives. First, it aimed to develop a comprehensive model of influential risk factors in earthmoving equipment operations (EEOs) by applying Rasmussen\u0026apos;s (1997) risk management framework (RMF), thereby uncovering the interrelationships among risk factors that contribute to incidents in the Australian construction industry. Second, the study sought to address the persistent safety challenges highlighted by earthmoving equipment incidents in Australia, where the impact of existing safety technologies such as building information modelling (BIM) remains limited due to an insufficient understanding of the origins, trajectories, and interconnections of risks across different system levels. Third, by identifying where each risk originates and evolves within the multi-layered RMF, the research aimed to pave the way for comprehensive vertical (across hierarchical levels), horizontal (across same-level actors), and end-to-end (throughout the entire operational process) integration of technological and managerial solutions, thereby enhancing risk identification and enabling the application of appropriate interventions aligned with specific system levels rather than applying generic, one-size-fits-all approaches.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe research employed a multi-phase mixed-method design. First, a systematic literature review was conducted to identify risk factors associated with earthmoving equipment operations, resulting in seven main categories and 52 sub-risk factors. Second, these factors were refined and validated through 32 semi-structured interviews with industry experts, and the findings were further aligned with relevant Australian codes of practice and regulations to ensure contextual relevance and practical applicability. Third, the study applied the fuzzy decision-making trial and evaluation laboratory (FDEMATEL) methodology marking the first application of this approach in the Australian construction context to analyse the cause-and-effect relationships among the identified risk factors within Rasmussen\u0026apos;s (1997) risk management framework. To ensure robustness and reliability, the methodology integrated statistical validation techniques, including corrected item-total correlation and split-half methods embedded within the FDEMATEL framework, as well as sensitivity analysis. These validation procedures were designed to ensure response consistency across participants, methodological robustness of the causal modelling, and reliability of the resulting factor classifications, ultimately enabling the identification of critical areas for targeted safety interventions in earthmoving equipment operations.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe analysis yielded several key findings. First, the most influential risk factors across the layers of Rasmussen\u0026apos;s risk management framework were successfully categorized into distinct cause groups (factors that drive or influence other factors) and effect groups (factors that are influenced by others), thereby clarifying the directional relationships among the 52 sub-risk factors. Second, the research produced an impact relations map (IRM) that visually classifies each risk factor according to its causal or effect-driven role within the earthmoving equipment operations system. This map revealed that certain factors termed \u0026nbsp;\u0026ldquo;influential factors \u0026ldquo; act as primary drivers of risk propagation across system levels, meaning that they exert disproportionate influence on other risk factors and therefore represent the most strategic leverage points for intervention. Third, the findings demonstrated that addressing effect-driven factors (those that are merely symptoms of underlying causal factors) without simultaneously targeting the causal factors themselves would yield limited safety improvements. Consequently, the results identified influential factors as the primary focus for technological advancements (e.g., sensor-based systems, BIM integration, computer vision) and managerial strategies (e.g., regulatory alignment, procedural reforms), thereby shifting attention from reactive incident response to proactive, systems-based risk mitigation.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several interconnected conclusions with significant implications for research and practice. First, from a research focus perspective, this study is the first to uncover the cause-and-effect relationships of risk factors not only specifically for earthmoving equipment operations but more broadly for construction operations in the Australian context, thereby filling a critical gap in the literature where previous studies had identified risk factors but had not systematically modelled their interdependencies or directional influences. Second, from a methodological perspective, the rigorous expert selection approach embedded within the FDEMATEL framework ensures that the findings are robust, reproducible, and contextually valid, offering a replicable methodology for future safety research in other construction domains or geographic contexts. Third, and most importantly, the findings fundamentally shift the focus of safety managers, site supervisors, and industry practitioners away from addressing isolated, visible, or effect-driven risks and towards addressing critical dynamic variables those acting as the \u0026nbsp;\u0026ldquo;Gordian knot \u0026ldquo; within the system. These causal risk factors, which are often less visible but more influential, must be systematically untangled to enable effective safety interventions and informed, strategic decision-making in earthmoving equipment operations. Ultimately, these insights strongly support the application of tailored solutions whether technological (e.g., sensor-based systems, building information modelling integration, computer vision) or procedural (e.g., regulatory alignment, training reforms, safety protocols) by explicitly aligning each intervention with the specific origin and trajectory of the risk factor it is designed to address, rather than applying generic or misaligned solutions that fail to resolve the underlying systemic causes of incidents.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e3.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eA Systematic Review on Design Thinking Integrated Learning in K-12 Education\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(T. Li \u0026amp; Zhan, 2022)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to systematically review high-quality empirical studies on design thinking integrated learning (DTIL) in K-12 education, recognizing that design thinking has become an essential approach for cultivating 21st-century competencies and that there has been a concomitant rise in needs and interest in introducing K-12 students to this methodology. Second, the study sought to synthesize the characteristics of existing DTIL implementations, including the target populations (grade levels), group sizes, intervention durations, curriculum domains, design thinking models or processes employed, core design thinking concepts emphasized, and the types of learning performances evaluated. Third, beyond describing the current state of the literature, the research aimed to identify research gaps and explore future research perspectives derived from the reviewed papers, thereby providing a evidence-informed roadmap for subsequent investigations into how design thinking can be most effectively integrated into K-12 educational settings to support student development of creative problem-solving, empathy, collaboration, and iterative reasoning skills.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a systematic literature review methodology to identify and synthesize high-quality empirical research on design thinking integrated learning in K-12 education. A systematic search was conducted across online databases using a combination of keyword searches and a snowballing approach (i.e., examining reference lists of relevant papers to identify additional studies). The inclusion criteria were stringent: only empirical studies published in SSCI (Social Sciences Citation Index) journals were considered, ensuring a baseline of academic quality and peer review. After the screening process, 43 SSCI journal papers, which collectively reported 44 individual studies, met the inclusion criteria and were included in the final review. No specific time range is mentioned, but the results indicate coverage of the past decade. Data extraction and synthesis focused on multiple dimensions: demographic characteristics (grade level, group size, study duration), curriculum domains, design thinking models and core concepts, learning performance outcomes, assessment methods and instruments, intervention types, learning settings, collaboration modes, and activity characteristics. The synthesis was primarily qualitative and descriptive, organizing findings into thematic categories to identify patterns, trends, and gaps across the included studies.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe analysis of the 43 SSCI papers yielded four main findings. First, regarding trends and participants, the results indicate that there has been a growing popularity of integrating design thinking into K-12 education over the past decade, with most empirical studies targeting middle school students, employing small group sizes, and conducted over short intervention periods. Second, concerning curriculum and design thinking models, studies tend to pay more attention to STEM-related curriculum domains (science, technology, engineering, and mathematics) and incorporate non-unified design thinking models or processes (i.e., different studies use different versions or adaptations of design thinking frameworks). The core concepts of design thinking that have been frequently valued and pursued in K-12 education include: prototype, ideate, define, test, explore, empathize, evaluate, and optimize. Third, regarding learning performances and assessment, the mostly evaluated learning performance is design thinking itself (i.e., students\u0026apos; acquisition of design thinking competency), followed by emotional and social aspects (e.g., attitudes, collaboration, empathy), subject learning performance (e.g., content knowledge in STEM subjects), and general skills. For evaluation methods, qualitative assessments are used more frequently than quantitative approaches, with common instruments including surveys/questionnaires, portfolios, interviews, observations, and protocol analysis. Fourth, concerning intervention characteristics, studies have mainly applied non-experimental study designs (rather than randomized controlled trials), formal classroom settings, collaborative learning arrangements, and traditional tools or materials (rather than advanced digital technologies) to support open-ended and challenging activities situated in real or realistic DTIL contexts. Overall, while the 43 papers suggest that design thinking shows great educational potential in K-12 education, the empirical evidence that supports the effectiveness of design thinking integrated learning remains rather limited.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions about the current state and future directions of design thinking integrated learning in K-12 education. First, despite the growing popularity and intuitive appeal of design thinking as a 21st-century competency, the empirical evidence base supporting its effectiveness in K-12 settings is still surprisingly limited, with most studies being non-experimental, short-term, and focused on middle school STEM contexts meaning that claims about long-term impact, generalizability to other grade levels or subject areas, and causal effectiveness remain inadequately supported by rigorous research. Second, the lack of a unified design thinking model across studies presents both a challenge and an opportunity; while it reflects the flexibility and adaptability of design thinking to different educational contexts, it also makes cross-study comparison, replication, and meta-analysis difficult, suggesting a need for clearer operational definitions and shared frameworks that retain core principles while allowing contextual adaptation. Third, the predominance of qualitative assessment methods, while valuable for capturing the richness of design thinking processes, also indicates a gap in the development and validation of reliable, scalable quantitative instruments that can measure design thinking competencies alongside subject learning outcomes, emotional development, and skill acquisition. Fourth, the research gaps identified in the reviewed papers point to several future directions: longer-term longitudinal studies to assess sustained impact, experimental and quasi-experimental designs to establish causality, expansion beyond STEM domains into humanities, arts, and social sciences, investigation of design thinking in diverse grade levels (early elementary and high school), exploration of digital and emerging technologies (e.g., generative AI, virtual reality) as scaffolds for design thinking, and deeper examination of how collaborative versus individual design thinking processes influence different types of learning outcomes. Ultimately, the study concludes that while design thinking holds great promise for cultivating 21st-century competencies in K-12 education, realizing that promise will require a substantial investment in more rigorous, diverse, and longitudinal empirical research that moves beyond proof-of-concept studies to establish evidence-based guidelines for implementation, assessment, and scaling\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e4.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eSystems Thinking in the Role of Fostering Technological and Engineering Literacy\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Kurent \u0026amp; Avsec, 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to examine whether the systems thinking approach integrating information and communication technology (ICT) and digital tools (referred to as the STICT approach) improves technological and engineering literacy (TEL) and related outcomes for pre-service preschool teachers. Second, the study sought to address a critical gap in the literature: although there is an expectation for preschool teachers to develop technological and engineering literacy, evidence-based models that systematically combine systems thinking with digital tools and ICT support remain scarce. Most existing approaches treat systems thinking and digital tool integration separately rather than as a unified pedagogical framework. Third, beyond measuring improvements in TEL, the research also aimed to assess the impact of the STICT approach on secondary outcomes including attitudes towards design, technology, and engineering (DTE), self-reported systems thinking, aspects of engagement (such as perceived difficulty with technology), and qualitative reflections from participants. The study thus sought to provide preliminary evidence for whether a systematically integrated approach offers advantages over traditional, product-focused DTE instruction in pre-service preschool teacher education.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a quasi-experimental design involving 44 pre-service preschool teachers over the duration of one semester. Participants were assigned to either an experimental group or a comparison control group. The experimental group explicitly integrated systems thinking principles with ICT and digital tools (the STICT approach), meaning that participants learned to apply systems thinking understanding interrelationships, feedback loops, and dynamic behaviors while simultaneously using digital tools and ICT to support their design, technology, and engineering processes. The comparison control group followed a traditional approach to teaching DTE content, focusing primarily on making physical products for preschoolers without explicit systems thinking integration or systematic ICT scaffolding. Both groups worked on similar design tasks aimed at creating products appropriate for preschool-aged children. Data collection involved multiple quantitative and qualitative measures: multidimensional literacy assessments (pre- and post-test), attitudes towards DTE questionnaires, self-reported systems thinking scales, engagement measures (including perceived difficulty with technology), and focus group reflections. Data analysis employed a range of statistical techniques including ANCOVA (analysis of covariance), MANCOVA (multivariate analysis of covariance), regression analysis, partial least squares (PLS) modeling, multi-group tests for comparing effects across conditions, and thematic analysis for qualitative focus group data. The authors explicitly caution that given the small sample size (n=44) and the multiple outcomes examined, all estimates carry considerable uncertainty and should be interpreted as preliminary rather than definitive.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe analyses yielded several notable findings. First, regarding technological and engineering literacy, the experimental group demonstrated a higher post-test literacy score compared to the comparison control group, suggesting an advantage for the STICT approach on the TEL composite measure. Second, concerning engagement and perceived difficulty, the experimental group reported lower perceived difficulty with technology than the control group, indicating that the integrated systems thinking and ICT approach may reduce students\u0026apos; anxiety or sense of challenge when working with digital tools. Third, for self-reported systems thinking, both groups improved significantly from pre-test to post-test, but there were no statistically significant differences between the experimental and control groups. This means that while all participants perceived themselves as better systems thinkers after the semester-long course, the explicit STICT approach did not produce superior gains in self-assessed systems thinking compared to the traditional product-focused approach. Fourth, the qualitative findings from focus group reflections supported the educational value of the STICT approach, with participants describing positive experiences related to understanding interconnections, using digital tools meaningfully, and feeling more confident in designing technology-enhanced learning experiences for preschoolers. The authors theorize that the TEL gains observed in the experimental group arise primarily from the systems thinking processes applied during design and evaluation stages, with ICT functioning as a cognitive-and-motivational scaffold that makes relationships and feedback loops explicit while simultaneously reducing the perceived difficulty of technology. The finding that self-assessed systems thinking improved similarly in both groups suggests that traditional product-focused DTE instruction may also implicitly develop some systems thinking awareness, or alternatively, that self-report measures may not be sensitive enough to capture the differential effects of explicit systems thinking instruction.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions while acknowledging important limitations. First, the findings are consistent with an advantage of the STICT approach over traditional DTE instruction for improving technological and engineering literacy among pre-service preschool teachers, as well as for reducing perceived difficulty with technology. These results suggest that explicitly integrating systems thinking with ICT and digital tools may offer a more effective pedagogical model for preparing preschool teachers to develop their own TEL competencies. Second, the finding that self-assessed systems thinking improved equally in both groups indicates that either the traditional approach also promotes some degree of systems thinking development (perhaps implicitly through design and making activities), or that self-report instruments are insufficiently sensitive to capture the distinct contributions of explicit systems thinking instruction. This finding points to the need for more objective or performance-based measures of systems thinking in future research. Third, the authors theorize a mechanism: TEL gains arise primarily from systems thinking processes applied during design and evaluation, while ICT serves as a cognitive-and-motivational scaffold that makes relationships and feedback loops explicit and reduces perceived difficulty. This theoretical explanation bridges the gap between the positive TEL outcomes and the null finding on self-assessed systems thinking. Fourth, and critically, the study emphasizes that these findings are preliminary. Given the small sample size (n=44), the quasi-experimental (non-randomized) design, the multiple outcome measures examined, and the considerable uncertainty surrounding the estimates, the results should be interpreted with caution rather than as definitive evidence. The study thus serves as a pilot classroom experiment that demonstrates feasibility and generates hypotheses for future research. Future directions include larger-scale randomized controlled trials, longer intervention periods, more objective measures of systems thinking competence (beyond self-report), examination of transfer to actual classroom practice, and investigation of whether the STICT approach produces sustained benefits for pre-service teachers when they enter professional practice and begin teaching technology and engineering content to preschool children.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e5\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eProfessional Knowledge and Self-Efficacy Expectations of Pre-Service Teachers Regarding Scientific Reasoning and Diagnostics\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Hilfert-R\u0026uuml;ppell et al., 2021)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to assess the understanding and knowledge of scientific reasoning skills among pre-service teachers, recognizing that scientific reasoning is a key ability for educators who will eventually guide school students through inquiry-based learning experiences. Specifically, the study sought to determine whether pre-service biology and chemistry teachers could identify the central decisions or actions that school students must perform when engaging in scientific reasoning during open inquiry instruction of an experiment. Second, the research aimed to measure the relationship between pre-service teachers\u0026apos; knowledge of student difficulties in scientific reasoning and other relevant psychological constructs, including attitudes towards the importance of diagnostics in teacher training and domain-specific expectations of self-efficacy (i.e., their confidence in their own ability to diagnose student abilities related to scientific reasoning). Third, by comparing these different measures, the study sought to identify specific gaps in pre-service teachers\u0026apos; professional knowledge particularly distinguishing between pedagogical content knowledge (knowledge about how to teach scientific reasoning and anticipate student difficulties), content methodological/procedural knowledge (knowledge of how to conduct scientific inquiry procedures), and epistemic knowledge (understanding why scientific procedures work as they do) and to generate practical implications for improving university-level teacher preparation programs.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a mixed-methods design with a sample of 51 pre-service biology and chemistry teachers recruited from two German universities. For the primary measure of scientific reasoning knowledge, participants completed a written survey using an open response format. In this survey, pre-service teachers were asked to identify and describe the central decisions or actions that school students would need to perform when engaging in scientific reasoning during an open inquiry instruction of an experiment (i.e., an inquiry-based learning activity where students have substantial autonomy in designing and conducting their own investigation). The participants\u0026apos; written responses were assessed using quality content analysis, which involved a systematic rubric system that was generated from a theoretical background on scientific reasoning and inquiry-based learning. This rubric allowed the researchers to evaluate the quality and accuracy of participants\u0026apos; answers across multiple dimensions of scientific reasoning. In addition to the open-response measure, the study employed instruments in a closed response format (e.g., Likert-scale questionnaires) to measure two additional constructs: (a) attitudes towards the importance of diagnostics in teacher training (i.e., how valuable pre-service teachers believe it is to learn how to diagnose student abilities and difficulties), and (b) domain-specific expectations of self-efficacy (i.e., pre-service teachers\u0026apos; confidence in their own ability to successfully cope with general diagnostic activities and experimental diagnostic activities related to scientific reasoning). The study then examined correlations among these measures to understand whether knowledge of student difficulties was associated with self-efficacy expectations or attitudes.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe analysis yielded several key findings. First, regarding knowledge gaps, the pre-service teachers demonstrated a significant lack of pedagogical content knowledge about potential student difficulties in scientific reasoning. That is, they were unable to accurately anticipate or identify the specific challenges that school students would face when performing scientific reasoning during open inquiry experiments. Additionally, the participants exhibited a low level of content methodological (procedural) knowledge, meaning they had insufficient understanding of the step-by-step procedures and methods involved in conducting scientific inquiry themselves. Second, concerning relationships among variables, the study found no correlation between the pre-service teachers\u0026apos; knowledge of student difficulties and their approach to experimenting with expectations of self-efficacy for diagnosing student abilities regarding scientific reasoning. In other words, knowing more about student difficulties was not associated with greater confidence in one\u0026apos;s own diagnostic abilities, suggesting that these two competencies may develop independently or require different types of learning experiences. Third, regarding attitudes versus self-efficacy, a significant discrepancy emerged: self-efficacy expectations concerning the pre-service teachers\u0026apos; own abilities to successfully cope with general and experimental diagnostic activities were significantly lower than their attitudes towards the importance of diagnostics in teacher training. This means that while pre-service teachers recognized that diagnostic skills are important (positive attitude), they did not feel confident in their own ability to perform such diagnostic activities effectively (low self-efficacy). This gap between valuing a competency and feeling competent to perform it represents a critical challenge for teacher education programs.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with important practical implications for university-level teacher preparation. First, the findings indicate that pre-service biology and chemistry teachers lack both pedagogical content knowledge (knowing what difficulties students will face) and procedural content knowledge (knowing how to conduct scientific inquiry methods). This dual deficit suggests that standard teacher training curricula may be insufficiently addressing the specific knowledge components required for guiding students through open inquiry experiments. Second, the absence of a correlation between knowledge of student difficulties and self-efficacy expectations implies that simply providing pre-service teachers with more information about student challenges will not automatically increase their confidence in diagnosing those challenges. Instead, self-efficacy may require mastery experiences, guided practice, and feedback in authentic diagnostic situations. Third, the significant gap between positive attitudes towards diagnostics (valuing it as important) and low self-efficacy (feeling unable to do it well) indicates that pre-service teachers are aware of what they should be able to do but do not believe they can actually do it. This mismatch could lead to avoidance of diagnostic activities in future classroom practice, even when teachers recognize their importance. Fourth, and most importantly, the results imply that scientific reasoning should be explicitly and systematically promoted in university courses, with emphasis placed on two interconnected dimensions: understanding science-specific procedures (knowing how the methodological and procedural aspects of scientific inquiry) and understanding epistemic constructs in scientific reasoning (knowing why the underlying rationales, justifications, and nature of scientific knowledge). The authors argue that teacher education programs must move beyond simply having pre-service teachers conduct experiments themselves; rather, programs must explicitly teach pre-service teachers how to anticipate student difficulties, how to diagnose student reasoning in real time, and how to scaffold student inquiry in ways that develop both procedural and epistemic understanding. Without such targeted preparation, pre-service teachers may enter the classroom unprepared to effectively guide their own students in developing scientific reasoning competencies.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e6\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003ePredictors of Corporate Reputation: Circular Economy, Environmental, Social, and Governance, and Collaborative Relationships in Brazilian Agribusiness\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Barbosa et al., 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to identify patterns of sustainability engagement based on three interconnected dimensions: circular economy (CE) strategy implementation (e.g., avoiding non-sustainable materials, repurposing by-products), CE-oriented collaborative relationships (e.g., fostering a shared CE vision with partners), and environmental, social, and governance (ESG) performance (e.g., adhering to ethical guidelines, ensuring financial transparency, implementing fair labor practices). Second, the study sought to investigate whether these three dimensions individually or in combination predict corporate reputation, thereby determining which specific sustainability practices are most strongly associated with how firms are perceived by stakeholders. Third, beyond identifying predictive relationships, the research aimed to extend existing theoretical frameworks, specifically the natural resource-based view (NRBV) and relational view (RV) theories, by demonstrating how the integration of CE strategies, collaborative relationships, and ESG performance collectively strengthens pollution prevention initiatives, sustainable product development efforts, and trust among partners, ultimately enhancing both corporate reputation and sustainable performance.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a quantitative, cross-sectional survey design with a sample of 235 upper-level managers operating in the Brazilian agribusiness sector. Data were collected using structured questionnaires that measured multiple constructs related to circular economy strategy implementation (e.g., material use, by-product management), CE-oriented collaborative relationships (e.g., shared vision, partnership practices), ESG performance (e.g., ethical guidelines, financial transparency, labor practices), and corporate reputation outcomes. The analytical approach consisted of two sequential steps. First, a cluster analysis was conducted to identify distinct patterns or groupings of firms based on their sustainability engagement across the three dimensions. This analysis revealed two primary clusters: firms characterized as \u0026nbsp;\u0026ldquo;Very Sustainable \u0026ldquo; and firms characterized as \u0026nbsp;\u0026ldquo;Low-Sustainable. \u0026ldquo; Second, following cluster identification, a logistic regression analysis was performed to determine which specific variables among the 28 measured sustainability indicators significantly predicted whether a firm belonged to the \u0026nbsp;\u0026ldquo;Very Sustainable \u0026ldquo; cluster (and by extension, predicted higher corporate reputation). This two-step approach allowed the researchers first to identify natural groupings in the data and then to isolate the most influential predictors from a large set of potential variables.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe analysis yielded several key findings. First, the cluster analysis identified two distinct patterns of sustainability engagement among agribusiness firms: a \u0026nbsp;\u0026ldquo;Very Sustainable \u0026ldquo; group characterized by high levels of CE strategy implementation, strong CE-oriented collaborative relationships, and robust ESG performance, and a \u0026nbsp;\u0026ldquo;Low-Sustainable \u0026ldquo; group characterized by lower levels across all three dimensions. This finding confirms that firms do not adopt sustainability practices uniformly but rather cluster into meaningful profiles. Second, the logistic regression analysis singled out six key predictors among the original 28 variables that most strongly distinguished the \u0026nbsp;\u0026ldquo;Very Sustainable \u0026ldquo; group from the \u0026nbsp;\u0026ldquo;Low-Sustainable \u0026ldquo; group. These six predictors were: (1) avoiding non-sustainable materials, (2) repurposing by-products, (3) fostering a shared circular economy vision with partners, (4) adhering to ethical guidelines, (5) ensuring financial transparency, and (6) implementing fair labor practices. Notably, these six predictors span all three theoretical dimensions: two relate to CE strategy (materials and by-products), one relates to CE-oriented collaborative relationships (shared vision), and three relate to ESG performance (ethical guidelines, financial transparency, fair labor practices). Third, the final logistic regression model achieved a high accuracy rate of 83.4%, indicating that the six identified predictors collectively do an excellent job of classifying firms into the correct sustainability engagement cluster. This high accuracy underscores how an integrated approach to sustainability combining circular economy practices, collaborative relationships, and governance integrity enhances corporate reputation.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions while acknowledging important limitations and future directions. From a theoretical contribution perspective, this research extends the natural resource-based view (NRBV) and relational view (RV) theories by empirically demonstrating that CE strategies, CE-oriented collaborative relationships, and ESG performance do not operate in isolation but rather function as an integrated system that collectively strengthens pollution prevention initiatives, sustainable product development efforts, and trust among partners, thereby enhancing firms\u0026apos; reputation and sustainable performance. The study thus moves beyond examining individual sustainability practices in isolation to demonstrate how multiple dimensions interact to produce reputation benefits. From a methodological perspective, the study contributes by integrating cluster analysis (to identify natural groupings) with predictive modeling (logistic regression) to assess sustainability\u0026apos;s impact on reputation, offering a replicable two-step approach for future research in other sectors or geographic contexts. From a managerial perspective, the findings emphasize that corporate reputation benefits most from a holistic approach that simultaneously addresses circularity (material use and by-product management), governance integrity (ethical guidelines and financial transparency), and stakeholder engagement (shared vision and fair labor practices). Managers seeking to enhance reputation should therefore avoid treating these as separate initiatives and instead integrate them into a coherent sustainability strategy. However, the study also acknowledges several limitations that temper the conclusions. The cross-sectional design captures relationships at a single point in time, preventing causal inferences. The industry-specific sample (Brazilian agribusiness only) limits generalizability to other sectors or countries. The reliance on self-reported data from upper-level managers introduces potential social desirability bias or inaccuracies in reporting. Therefore, future research should adopt longitudinal designs to examine how sustainability-reputation relationships evolve over time, cross-industry approaches to test whether the six identified predictors generalize beyond agribusiness, and integration of external data sources (e.g., third-party ESG ratings, public records) to complement self-reported measures. Additionally, future studies should examine how regulatory shifts, technological advances, and evolving stakeholder demands influence the sustainability\u0026ndash;reputation nexus across different institutional and cultural settings.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e7.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eDissecting the compensation conundrum: a machine learning-based prognostication of\u0026nbsp;key determinants in a complex labor market\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Jaiswal et al., 2023)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to introduce a holistic, integrated theoretical framework that synthesizes multiple management theories and incorporates machine learning models to develop a compensation model capable of accurately predicting pay determination in the information technology industry. This objective arose from the context of geopolitical uncertainty, pandemic-induced economic disruptions, alarming attrition rates, and aggravating talent gaps that have spurred a surge in demand for specialized digital proficiencies, leading firms to seek ways to attract and retain top-tier talent through generous compensation packages. Second, the study sought to interrogate the multifaceted factors that shape pay determination including experience level, educational background, specialized skill sets, gender, company size, and company type to determine which factors truly drive compensation and which do not. Third, beyond prediction, the research aimed to provide practical value by empowering individuals to negotiate compensation more effectively, supporting enterprises in crafting targeted compensation strategies, and ultimately facilitating sustainable economic growth while helping to attain various Sustainable Development Goals (SDGs) related to decent work and economic growth.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a quantitative, predictive modeling approach using a stratified sample of 2,488 observations drawn from the information technology sector. The sampling strategy ensured representation across different levels of experience, educational backgrounds, skill sets, company sizes, company types, and genders to capture the full range of variability in compensation determination. The research question was whether compensation could be accurately predicted using constructs derived from the integrated theoretical framework (which synthesized multiple management theories to capture the complexity of pay determination). To answer this question, the study employed various cutting-edge machine learning models, including but not limited to random forest, support vector machines, neural networks, gradient boosting, and regression-based algorithms. Each model was trained on a portion of the dataset and tested on a held-out portion to evaluate predictive accuracy. The models were compared against each other to identify the best-performing algorithm. A series of comprehensive robustness checks were conducted to ensure the stability and reliability of the findings, including cross-validation, sensitivity analyses, and tests for overfitting. The final model selection was based on two key performance metrics: prediction accuracy (percentage of correctly predicted compensation outcomes) and mean absolute error (the average magnitude of prediction errors in the original measurement units).\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe empirical findings of this study yielded several critical results. First, regarding model performance, the research culminated in the discovery that the random forest model outperformed all other machine learning algorithms tested, achieving an exceptionally high accuracy of 99.6% and a remarkably low mean absolute error of 0.08 degrees (presumably in the relevant compensation units, such as thousands of currency units or log-transformed values). This indicates that the random forest model can predict individual compensation with near-perfect precision based on the constructs derived from the integrated theoretical framework. Second, concerning the determinants of compensation, the study identified several critical predictors including, but not limited to, experience level, educational background, and specialized skill-set. These factors were found to have substantial influence on pay determination. Third, and notably, the research elucidated that gender does not play a role in pay disparity, suggesting that within the sampled IT sector context, there is no evidence of gender-based compensation discrimination after accounting for other relevant factors. Fourth, the study found that company size and company type hold no consequential sway over individual compensation determination, meaning that whether an employee works for a large multinational corporation or a small startup, or for a product company versus a service company, does not significantly affect their individual pay when experience, education, and skills are accounted for.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with significant theoretical, practical, and original contributions. From a theoretical perspective, the cardinal contribution of this research lies in the inception of an inclusive theoretical framework that persuasively explicates the intricacies of a machine learning-driven remuneration model. This framework is ennobled by the synthesis of diverse management theories likely including human capital theory, signaling theory, equity theory, and resource-based view to capture the full complexity of compensation determination in the modern IT industry. By integrating these theoretical perspectives with advanced machine learning methods, the study bridges a gap between traditional econometric compensation studies and contemporary predictive analytics. From a practical implications perspective, the research underscores the importance of equitable compensation to foster technological innovation and encourage the retention of top talent, emphasizing the significance of human capital as a strategic asset. The highly accurate random forest model presented in this study empowers individuals to negotiate their compensation more effectively by providing them with evidence-based benchmarks. Simultaneously, the model supports enterprises in crafting targeted compensation strategies that reward the factors that truly matter (experience, education, specialized skills) while avoiding discrimination on irrelevant factors (gender, company size, company type). This alignment with equitable pay practices facilitates sustainable economic growth and helps attain various Sustainable Development Goals, particularly SDG 8 (Decent Work and Economic Growth) and SDG 5 (Gender Equality), given the finding that gender does not drive pay disparity. However, the study acknowledges a significant limitation: the generalizability of the findings to other sectors is constrained, as this study is exclusively limited to the IT sector. Future research should extend the integrated theoretical framework and machine learning methodology to other industries such as healthcare, finance, manufacturing, and education to test whether the same determinants of compensation operate similarly or whether sector-specific factors emerge. Additionally, future studies should explore longitudinal data to examine how compensation determinants evolve over time with technological change, as well as cross-country comparisons to investigate how institutional, cultural, and regulatory contexts moderate the relationships identified in this study.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e8.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eGovernments, Users, and Virtual Worlds: Institutional Strategies in the Age of Big Data and IA\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Crespo-Pereira \u0026amp; Miranda-Galbe, 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to critically examine why governments are investing in the metaverse ecosystem, recognizing that several countries have recently introduced strategic plans aimed at promoting metaverse ecosystems, yet the underlying rationales for such investments remain insufficiently understood. Second, the study sought to investigate how the metaverse is being approached as an innovative platform for digital public services and businesses, moving beyond purely technological or commercial framings to understand the policy-level conceptualizations and strategic intentions embedded in official documents. Third, by analyzing regional, national, and supranational metaverse strategic plans, the research aimed to identify and articulate the main reasons driving government promotion of the social and industrial metaverse ecosystem, including concepts such as sustainability, digital sovereignty, competitive advantage, and stakeholder relationship building, thereby filling a critical gap in the literature where metaverse policy rationales have received far less attention than metaverse technologies themselves.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a qualitative document analysis approach using both inductive and deductive content analysis methods. The sample consisted of metaverse strategic plans (n = 7) drawn from multiple levels of governance, including regional, national, and supranational jurisdictions. The selection of these seven strategic plans was presumably based on their official status, public availability, and relevance to metaverse ecosystem development as a policy priority. The analytical process combined inductive content analysis, where themes and categories emerge organically from the text without predetermined coding schemes, and deductive content analysis, where existing theoretical concepts or prior frameworks guide the coding process. This hybrid approach allowed the researchers to remain open to unexpected themes while still testing for theoretically relevant concepts. The analysis focused on identifying and coding segments of text related to government rationales for metaverse investment, conceptualizations of the metaverse as a platform for public services and business, definitions or characterizations of virtual worlds, and specific approaches (e.g., transactional, connected) through which the metaverse is expected to operate. The findings were then synthesized into thematic categories representing the main reasons for promoting the metaverse ecosystem.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe analysis yielded several key findings. First, regarding the conceptual understanding of virtual worlds, the study found that virtual worlds can be understood as persistent, immersive, and interactive digital environments that integrate three-dimensional visualization, simulation, and real-time data to support activities across both social and economic domains. This definition synthesizes common elements across the seven strategic plans. Second, concerning the fundamental nature of the metaverse, the findings indicate that the metaverse is conceptualized as a virtual space shaped by a dual imperative: (a) addressing societal needs such as public service delivery and stakeholder engagement and (b) fostering business opportunities within the evolving digital ecosystem. This dual framing distinguishes government-led metaverse initiatives from purely private-sector or entertainment-focused metaverse developments. Third, the analysis revealed four main reasons driving government promotion of the social and industrial metaverse ecosystem: (1) sustainability (using metaverse technologies to reduce physical resource consumption and support environmentally sustainable practices), (2) digital sovereignty (ensuring that metaverse infrastructure, data, and governance remain under domestic or regional control rather than being dominated by foreign technology giants), (3) competitive advantage (positioning national industries and economies to lead in emerging metaverse markets and technologies), and (4) stakeholder building relationships (fostering connections among citizens, businesses, government agencies, and other stakeholders through shared virtual spaces). Fourth, regarding operational approaches, the results indicate that the metaverse operates mainly through both transactional and connected approaches, where digital twins (virtual replicas of physical systems), artificial intelligence, and extended reality (virtual, augmented, and mixed reality) converge to enable user experiences in ways that transcend physical limitations. Transactional approaches likely refer to metaverse-enabled exchanges of goods, services, or information, while connected approaches likely refer to metaverse-enabled relationship building, collaboration, and social presence.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with important implications for policy, practice, and future research. First, the findings demonstrate that government investment in the metaverse is not driven solely by technological fascination or economic opportunity but rather by a coherent set of strategic rationales sustainability, digital sovereignty, competitive advantage, and stakeholder relationship building that reflect broader societal and geopolitical priorities. This suggests that metaverse policies should be understood as part of larger industrial, environmental, and foreign policy strategies rather than as isolated technology initiatives. Second, the dual imperative of addressing societal needs while fostering business opportunities positions the metaverse as a unique policy instrument that bridges public service delivery and private sector development. Governments are not merely regulating or funding metaverse technologies; they are actively shaping metaverse ecosystems to serve both citizen-centric and market-oriented goals simultaneously. Third, the identification of transactional and connected approaches as the primary operational modes of government-promoted metaverses provides a useful typology for future comparative research. Transactional approaches focus on efficiency, exchange, and task completion (e.g., virtual government service counters, digital permit applications), while connected approaches focus on presence, collaboration, and relationship building (e.g., virtual stakeholder consultations, immersive public hearings, collaborative urban planning in digital twins). Fourth, the study highlights that digital twins, artificial intelligence, and extended reality are not separate technologies but converging enablers that, when integrated, create user experiences that transcend physical limitations such as attending a public meeting from anywhere in the world while experiencing a sense of presence, or simulating the environmental impact of a policy decision before implementing it physically. However, the study acknowledges limitations: the sample of seven strategic plans, while drawn from multiple governance levels, may not be representative of all countries or regions developing metaverse strategies. Additionally, the analysis focuses on planned rationales and approaches as articulated in policy documents, not on actual implementation outcomes. Future research should therefore examine whether and how these rationales translate into measurable metaverse ecosystem development, compare implementation successes and failures across different governance contexts, investigate stakeholder perspectives (citizens, businesses, civil society) on government-led metaverse initiatives, and explore potential risks and unintended consequences such as digital exclusion, data privacy concerns, and the widening of digital divides. Longitudinal studies tracking how metaverse strategic plans evolve over time as technologies and political priorities shift would also be valuable.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e9.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eKey characteristics for designing a supply chain performance measurement system\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Elgazzar et al., 2019)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to conduct a comprehensive review of the literature that gives insight into design elements for constructing a supply chain performance measurement (SCPM) system, recognizing that effective performance measurement is essential for managing and improving supply chain operations. Second, the study sought to categorize the key functions of SCPM systems by providing insight into four critical dimensions: design (how SCPM systems are structured), functionality (what purposes they serve), implementation (how they are deployed in practice), and practical implications (what outcomes and challenges result from their use). Third, beyond simply describing existing knowledge, the research aimed to identify functions of SCPM systems that have not been fully explored in previous research specifically the process focus, prioritization, integration, and causality functions and to explore how relationships between two or more functions can be combined to create more comprehensive performance measurement systems tailored to the specific needs of individual companies.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a systematic literature review methodology to synthesize published research on supply chain performance measurement systems and frameworks over a two-decade period. A systematic review of the literature was conducted, meaning that the researchers followed a structured, replicable process for searching, screening, and selecting relevant studies rather than an informal or selective narrative review. The search strategy was designed to capture a comprehensive body of knowledge on SCPM systems, including both foundational frameworks and more recent developments. The inclusion criteria focused on research that addressed the design, functionality, implementation, or practical implications of SCPM systems. After the screening process, the review incorporated findings from a substantial body of literature: 269 research papers published over the last two decades. This large sample size (269 papers) provided a robust evidence base for synthesizing patterns, identifying gaps, and developing new conceptual insights. The analysis involved extracting and synthesizing information related to the functions of SCPM systems, with particular attention to functions that had received limited attention in prior reviews or empirical studies. The synthesis was both descriptive (cataloging what exists) and analytical (identifying relationships, gaps, and opportunities for integration).\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe systematic review revealed several key findings. First, the analysis identified a set of functions governing SCPM systems that have not been fully explored in previous research. These under-explored functions include: (a) the process focus function (emphasizing the measurement of supply chain processes rather than just outcomes or individual organizational units), (b) the prioritization function (enabling organizations to determine which performance dimensions are most critical given strategic objectives), (c) the integration function (connecting performance measurement across different supply chain partners, tiers, or functional silos), and (d) the causality function (revealing cause-and-effect relationships among performance drivers and outcomes, such as how improvements in one metric affect others). Second, the findings indicate that a relationship between two or more functions can be created to include more functionality based on the needs of the company. In other words, organizations are not forced to choose a single function but can combine functions in various configurations to design a comprehensive performance measurement system that addresses their specific strategic and operational contexts. Third, the paper presents multiple potential stages of merging different functions in one SCPM system, with the functionality of the SCPM system capable of being designed at four distinct levels. These four levels create ten possible scenarios when designing a company\u0026apos;s SCPM system, offering a range of options from simple, single-function systems to complex, multi-function integrated systems.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with theoretical, practical, and original contributions. From a theoretical perspective, the paper integrates the literature and findings of 269 research papers from the last two decades, upon which a conceptual framework was developed as a guide for constructing an effective SCPM system. This conceptual framework synthesizes the four identified functions (process focus, prioritization, integration, and causality) and articulates how they can be combined in different configurations. The framework serves as a theoretical contribution that extends prior SCPM research by moving beyond lists of metrics or generic frameworks to a functional perspective that recognizes that different supply chains may require different performance measurement functionalities. From a practical implications perspective, the paper brings a new dimension to SCPM research by identifying the main functions of SCPM systems that could benefit practitioners seeking to set up a SCPM system relevant to its intended function. Rather than adopting a one-size-fits-all approach, practitioners can use the four functions and the ten possible design scenarios to match the SCPM system\u0026apos;s capabilities to their company\u0026apos;s specific needs and context. For example, a company focused primarily on internal process improvement might emphasize the process focus and causality functions, while a company concerned with supplier integration might prioritize the integration function. The paper also concludes with a conceptual framework to guide future research in the area of designing SCPM systems, defining the main aspects that should be considered when developing such systems. However, the study acknowledges limitations: as a systematic review, its findings are constrained by the quality and scope of the 269 primary studies included; the conceptual framework, while grounded in existing literature, has not yet been empirically tested in real-world supply chain contexts. Therefore, future research should empirically validate the proposed framework through case studies, action research, or surveys across diverse industries and supply chain structures. Additionally, future studies should investigate how digital technologies such as the Internet of Things (IoT), blockchain, artificial intelligence, and real-time analytics might enable new SCPM functions beyond the four identified here, as well as how the ten design scenarios might be implemented in practice and what contextual factors (e.g., supply chain complexity, power dynamics, trust levels) moderate the effectiveness of different functional configurations.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e10.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eGamification as an innovation: a\u0026nbsp;tool to improve organizational marketing performance and sustainability of international firms\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Behl et al., 2024)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to investigate an under-researched area at the intersection of international marketing perspective, international dynamic capability, environmental sustainability, and organizational marketing performance, specifically comparing gamification-based and non-gamification-based organizational culture (OC). Second, the study sought to deepen the understanding of how gamification-based and non-gamification-based OC influence innovation capability (both technological and environmental innovation capabilities) and, subsequently, environmental and organizational marketing performance. To achieve this theoretical depth, the research drew upon two complementary theoretical lenses: the theory of organizational creativity (which explains how organizational contexts foster or inhibit creative processes) and the theory of administrative behavior (AB) (which explains how bounded rationality, decision-making processes, and administrative structures shape organizational actions, particularly in steering technological creativity toward climate-conscious outcomes). Third, the study aimed to provide practical guidance for firms to invest in technological solutions by practicing creativity over time, as well as to explore how administrative behavior is critical in directing technological creativity toward making firms more climate-conscious and environmentally responsible.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a quantitative, cross-sectional survey design with primary data collected from firms that abide by ISO 14091 certifications, which ensure proper quality standards related to environmental management and climate change adaptation. The use of ISO 14091-certified firms as the sampling frame was a deliberate methodological choice to ensure that all participating firms had a baseline level of environmental management maturity, thereby reducing extraneous variation and allowing the study to focus more precisely on the effects of organizational culture (gamification-based vs. non-gamification-based) on innovation and performance outcomes. Data were collected from 384 firms, providing a sufficiently large sample size for robust statistical analysis. The study tested multiple hypotheses using appropriate statistical techniques, including regression analysis, path analysis, or structural equation modeling (exact techniques are not specified in the abstract but are implied by the hypothesis-testing design). The analysis examined direct effects (e.g., OC on innovation capabilities), mediating effects (e.g., innovation capabilities as mechanisms linking OC to environmental sustainability and marketing performance), and moderating effects (specifically, the moderating effect of gamification on the relationship between organizational culture and environmental innovation capabilities, particularly within the context of international dynamic capabilities).\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe empirical findings of the study revealed several key relationships. First, the study identified that organizational culture (both gamification-based and non-gamification-based) has a positive influence on technological innovation capabilities (the ability to develop and implement new technologies) and environmental innovation capabilities (the ability to develop innovations that reduce environmental impact). Second, technological innovation capabilities were found to have a beneficial impact on environmental sustainability, meaning that firms with stronger technological innovation capacities achieved better environmental outcomes. Third, environmental sustainability appeared to have a substantial correlation with technological innovation skills, suggesting a reciprocal or mutually reinforcing relationship where sustainability efforts enhance technological capabilities and vice versa. Fourth, environmental innovation capabilities positively impacted both environmental sustainability and organizational marketing performance, indicating that the ability to generate environmentally focused innovations not only improves environmental outcomes but also enhances how the organization performs in the marketplace (e.g., brand perception, customer loyalty, market share). Fifth, and notably, the study identified a moderating effect of gamification on the relationship between organizational culture and environmental innovation capabilities within the context of international dynamic capabilities. This means that the presence of gamification elements in organizational culture strengthens (or otherwise alters) the positive relationship between culture and environmental innovation capabilities, particularly when firms are operating across international boundaries and need to adapt dynamically to diverse environmental regulations, customer expectations, and competitive conditions.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with theoretical and practical contributions while acknowledging its scope and limitations. From a theoretical perspective, the investigation is confined to understanding how gamification-based and non-gamification-based organizational marketing culture affects innovation capability, environmental sustainability, and organizational performance through the lens of two complementary theories: the theory of organizational creativity (which explains how cultural elements such as gamification can stimulate creative processes that lead to technological and environmental innovations) and the theory of administrative behavior (which explains how structured decision-making processes and bounded rationality shape the way organizations translate creative potential into actual climate-conscious actions). The finding that gamification moderates the culture\u0026ndash;environmental innovation relationship suggests that gamification is not merely a surface-level engagement tool but a substantive cultural feature that can amplify or redirect the effects of organizational culture on innovation outcomes, particularly in internationally dynamic contexts. From a practical implications perspective, the results would help firms invest in technological solutions by practicing creativity over time, meaning that sustained, culturally embedded creativity (potentially enhanced by gamification) is a driver of both innovation and environmental performance. Additionally, the study helps explore how administrative behavior is critical in steering technological creativity for making firms climate-conscious, indicating that having creative ideas is insufficient; structured administrative processes, decision rules, and managerial behaviors are necessary to channel creativity toward meaningful environmental outcomes. The study\u0026apos;s findings also have practical relevance for managers considering whether to adopt gamification elements in their organizational culture: the positive influence of OC on innovation capabilities, combined with the moderating role of gamification, suggests that gamification can be a strategic lever for enhancing environmental innovation, especially for firms operating internationally. However, the study is limited to firms with ISO 14091 certifications, which may represent a more environmentally mature subset of firms, potentially\u0026nbsp;limiting generalizability to firms without such certifications. Future research should extend this investigation to non-certified firms, explore longitudinal designs to establish causality rather than mere correlation, examine potential negative or unintended consequences of gamification (e.g., superficial engagement, extrinsic motivation crowding out intrinsic environmental values), and investigate how different types of gamification (e.g., competitive vs. collaborative, reward-based vs. meaning-based) produce different effects on innovation capabilities and environmental outcomes across diverse cultural and institutional contexts.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e11.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eAn empirical mixed-methods evaluation of AI-based chatbots for teacher professional development in Austrian higher education\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Leitgeb \u0026amp; Leitgeb, 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had four primary objectives. First, it aimed to address a critical gap in traditional teacher professional development (PD) programs, which often neglect individual needs, specific subject-area demands, and distinct career stages, leading to limited relevance and low uptake among teachers. Second, the study sought to deploy an AI-based chatbot to provide context-sensitive, personalized PD recommendations at scale, thereby moving beyond one-size-fits-all approaches to professional learning. Third, grounded in two theoretical frameworks technological pedagogical content knowledge (TPACK), which explains the intersections among technology, pedagogy, and content knowledge, and self-determination theory, which emphasizes the importance of autonomy, competence, and relatedness for intrinsic motivation the research aimed to evaluate how tailored chatbot interactions can enhance teachers\u0026apos; motivation, autonomy, and technological proficiency while simultaneously meeting pedagogical and content-specific requirements. Fourth, by integrating multiple theoretical models including the information systems success model (ISSM) and the technology acceptance model (TAM), the study sought to provide a comprehensive evaluation of chatbot-supported PD from both user experience and educational relevance perspectives.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a convergent parallel mixed-methods design, meaning that quantitative and qualitative data were collected simultaneously, analyzed separately, and then integrated to provide a comprehensive understanding of the chatbot\u0026apos;s performance. The study analyzed 2,030 valid chatbot interactions from 1,125 teachers in Austria\u0026apos;s Burgenland region. Data collection was guided by three theoretical frameworks: the information systems success model (ISSM), the technology acceptance model (TAM), and technological pedagogical content knowledge (TPACK). Quantitative metrics included fallback rates (the frequency with which the chatbot could not answer a query), implicit intent interpretation (the chatbot\u0026apos;s ability to understand user intent without explicit cues), sentiment analysis (automated classification of user feedback as positive, neutral, or negative), and confidence scores (the chatbot\u0026apos;s certainty in its responses). Qualitative feedback examined perceived relevance of the recommendations from the teachers\u0026apos; perspectives. Analytical techniques included descriptive and inferential statistics (e.g., logistic regression) to assess relationships between query characteristics and user satisfaction, alongside content analyses of qualitative feedback. This mixed-methods design enabled a comprehensive evaluation of both measurable performance indicators (e.g., fallback rates, sentiment) and user perspectives regarding the perceived value and relevance of chatbot-enabled PD recommendations.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe empirical findings yielded several key results. First, the chatbot demonstrated a moderate fallback rate of 14.4%, which is significantly below established benchmarks in conversational AI systems, indicating that the chatbot was able to successfully handle the vast majority (85.6%) of teacher queries without failing. Second, overall user sentiment was positive, with 85% of interactions receiving favorable feedback from teachers. Third, quantitative analyses revealed that teachers who submitted highly specific queries reported greater satisfaction compared to those who submitted vague or general queries, suggesting that the chatbot\u0026apos;s performance and perceived usefulness are enhanced when users provide detailed, context-rich inputs. Fourth, logistic regression analysis revealed that targeted pedagogical keywords (e.g., specific teaching strategies, subject-area terminology, grade-level references) significantly increased the likelihood of positive feedback, meaning that queries containing domain-specific pedagogical language were more likely to receive favorable user ratings. Fifth, qualitative insights underscored the importance of both detailed query formulations (providing sufficient context and specificity) and domain-specific terminology (using the specialized vocabulary of teaching, subject areas, and pedagogy). Collectively, these findings highlight robust chatbot performance across multiple metrics and emphasize the critical role of contextualized, technology-oriented PD solutions for meeting teachers\u0026apos; individualized professional needs.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions while acknowledging limitations and offering implications for research, practice, and society. From a theoretical and originality perspective, this research uniquely synthesizes the information systems success model (ISSM), the technology acceptance model (TAM), and technological pedagogical content knowledge (TPACK) to evaluate chatbot-supported teacher PD, offering a multi-faceted assessment of both user experience (via ISSM and TAM) and educational relevance (via TPACK). By emphasizing the significance of query specificity and targeted pedagogical language, the study advances understandings of how AI-driven tools can address individualized teacher needs across diverse contexts, contributing to ongoing discourse on data-informed professional development. From a practical implications perspective, institutional stakeholders can optimize AI-based PD tools by encouraging teachers to submit more detailed queries and employ targeted pedagogical terminology. Systematic refinements such as updating domain-specific vocabularies and improving natural language processing algorithms can reduce fallback rates and enhance user satisfaction. Training programs aimed at familiarizing educators\u0026nbsp;with chatbot functionalities and best practices can further increase engagement. From a social implications perspective, by providing accessible, context-sensitive PD resources, AI-driven chatbots may help democratize professional learning for teachers across diverse settings, including those with limited institutional support. This can contribute to narrowing digital skill gaps, especially in remote or underserved schools, thereby promoting educational equity and fostering a ripple effect on student outcomes and broader societal advancement. However, the study acknowledges several limitations. Due to the relatively brief observation period and the self-selecting nature of participating teachers, these findings may not be generalizable across broader educational settings. The sample, drawn from a single Austrian region (Burgenland), may limit external validity. Future research should incorporate larger, more diverse populations, extend the timeframe to measure long-term outcomes (e.g., sustained changes in teaching practice, student learning gains), collect additional demographic data to assess subgroup variations (e.g., by career stage, subject area, school type, prior technology experience), and conduct longitudinal investigations into the sustained impact of chatbot-based recommendations on teaching practice. Additionally, future studies should explore the role of AI-driven PD in different educational contexts (e.g., primary vs. secondary, urban vs. rural, high-resource vs. low-resource schools) and investigate potential risks such as over-reliance on AI recommendations, algorithmic bias, data privacy concerns, and the digital divide in access to chatbot technologies.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e12.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eMapping Contemporary AI-Education Intersections and Developing an Integrated Convergence Framework: A Bibliometric-Driven and Inductive Content Analysis\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Ali et al., 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to address a notable gap in the scholarly literature on artificial intelligence (AI) in education, which, despite rapid permeation since 2014 driven by technological innovation and global investment, remains largely fragmented in terms of coherent discourse and synthesized understanding. Second, the study sought to employ a bibliometric-driven and inductive content analysis approach to map the intellectual structures, thematic clusters, and prevailing research trends shaping the contemporary AI-education intersection, thereby providing a systematic and evidence-informed foundation for future research and practice. Third, beyond mapping the existing landscape, the research aimed to identify specific research issues or gaps within the literature such as limited congruence between technological and pedagogical affordances, insufficient bottom-up perspectives in AI literacy frameworks, ambiguous relationships between computational thinking and AI, lack of explicit interpretation of AI ethics, and limitations of existing professional development frameworks and to consolidate issue-specific recommendations into an overarching framework. The ultimate objective was to develop and propose the Integrated AI-Education Convergence Framework, which advocates for pedagogy-centric, ethically grounded, and contextually responsive AI integration within interdisciplinary educational research and practice.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a dual-method approach combining bibliometric analysis with qualitative inductive content analysis, following the PRISMA protocol (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) for systematic literature retrieval and screening. A total of 317 articles published between 2014 and October 2024 were retrieved from two major scholarly databases: the Web of Science Core Collection (WOSCC) and Scopus. The PRISMA protocol ensured transparency, replicability, and rigor in the identification, screening, eligibility assessment, and inclusion of relevant articles. For the bibliometric analysis, the study used VOSviewer (version 1.6.20), a specialized software tool for constructing and visualizing bibliometric networks. Two primary bibliometric techniques were employed: keyword co-occurrence analysis, which examines how frequently pairs of keywords appear together in the same publications to reveal major research themes and their interconnections, and co-citation analysis, which examines how frequently pairs of cited references are cited together by subsequent publications to reveal the intellectual foundations and influential works shaping the field. These techniques produced visual maps of the intellectual structures underlying AI-education research. To address the limitations of bibliometric methods which can reveal patterns of publication and citation but often cannot capture deeper thematic insights, contextual nuances, or interpretive meanings the study additionally conducted qualitative inductive content analysis. This involved systematically reading and coding the full texts of the 317 articles to identify themes, patterns, and gaps that might not be visible through quantitative bibliometric indicators alone. The integration of both methods allowed the study to leverage the strengths of each while compensating for their respective limitations.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe dual-method analysis yielded several significant findings. First, the bibliometric analysis identified four distinct thematic clusters within the AI-education literature, representing coherent groupings of research topics, keywords, and cited works that collectively define the major intellectual domains in the field. Second, through both bibliometric and inductive content analysis, the study identified eleven prevailing research trends that have characterized AI-education scholarship between 2014 and 2024, capturing the evolution of research priorities, methodological approaches, and application domains over this decade. Third, and most substantively, through interpretive synthesis of the four thematic clusters and eleven trends, the study identified five interrelated research issues or gaps that persist in the literature: (1) limited congruence between technological affordances (what AI systems can do) and pedagogical affordances (what educational practices and learning processes they can meaningfully support), indicating a disconnect between technical development and educational design; (2) insufficient bottom-up perspectives in AI literacy frameworks, meaning that most frameworks are developed from expert or top-down perspectives rather than incorporating the lived experiences, needs, and voices of teachers and students; (3) an ambiguous relationship between computational thinking (CT) and AI, where the literature lacks clarity on whether CT is a prerequisite for AI learning, a component of AI literacy, a parallel competency, or something else; (4) a lack of explicit interpretation of AI ethics, with many studies mentioning ethics superficially or not at all, and few providing concrete, contextualized guidance for ethical AI use in educational settings; and (5) limitations of existing professional development frameworks, which are often inadequate for preparing teachers to integrate AI effectively, ethically, and pedagogically soundly. To address these five gaps pragmatically, the study consolidated thirty specific, actionable recommendations (derived from the literature) into five overarching themes, each corresponding to one of the identified issues.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with significant theoretical, practical, and integrative contributions. The culmination of the research is the proposed Integrated AI-Education Convergence Framework, which synthesizes the four thematic clusters, the eleven research trends, the five identified research issues, and the thirty issue-specific recommendations into a coherent, actionable, and theoretically grounded model. This framework advocates for three core principles in AI integration within interdisciplinary educational research and practice. First, pedagogy-centric integration means that AI tools and applications should be designed, selected, and implemented based on sound pedagogical principles and learning objectives, rather than being driven by technological novelty or availability. Second, ethically grounded integration means that AI ethics must be made explicit, contextualized, and actionable moving beyond generic principles to concrete guidance that addresses issues such as data privacy, algorithmic bias, transparency, accountability, and the potential for AI to exacerbate or reduce educational inequities. Third, contextually responsive integration means that AI solutions must be adaptable to the specific needs, cultures, resources, and constraints of diverse educational settings, avoiding one-size-fits-all approaches that may fail in practice. The framework is intended to guide future research by providing a structured lens for identifying gaps, designing studies, and synthesizing evidence, as well as to inform practice by offering educators, administrators, and policymakers a roadmap for responsible and effective AI integration. However, the study acknowledges limitations: the sample, while systematically retrieved from two major databases, may not capture all relevant publications, particularly those in emerging venues, non-English sources, or the gray literature. The bibliometric analysis, while powerful for mapping intellectual structures, is inherently backward-looking, capturing patterns in published and cited work that may lag behind cutting-edge developments. Future research should extend this work by incorporating longitudinal analyses to track how the thematic clusters and research issues evolve over time, conducting empirical validations of the Integrated AI-Education Convergence Framework in diverse educational contexts, investigating the specific mechanisms through which pedagogy-centric and ethically grounded AI integration can be achieved at scale, and exploring how emerging generative AI technologies (e.g., large language models) reshape the landscape in ways that may require revision or extension of the current framework. Additionally, future studies should amplify bottom-up perspectives by systematically collecting and integrating the voices of teachers, students, and local communities into AI literacy frameworks and professional development designs.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e13.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eDeepfake-Style AI Tutors in Higher Education: A Mixed-Methods Review and Governance Framework for Sustainable Digital Education\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Sharif et al., 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to understand the pedagogical potential of deepfake-style AI tutors emerging technologies that offer personalized and multilingual instruction in online education while simultaneously identifying the risks they introduce to academic integrity, privacy, and trust. Second, the study sought to investigate the governance needs necessary for responsible integration of these technologies, recognizing that their benefits (engagement, adaptability, scalability) must be balanced against significant challenges such as impersonation, assessment fraud, and algorithmic bias. Third, beyond diagnosis and analysis, the research aimed to develop a practical governance framework to guide institutions, policymakers, and educators in deploying deepfake AI tutors ethically and responsibly. This framework was designed to strengthen the ethical and governance foundations for generative AI in higher education, contribute to the broader agenda of sustainable digital education, and align with the United Nations Sustainable Development Goal 4 (Quality Education) by promoting transparency, fairness, equitable access, and institutional resilience through responsible innovation.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a mixed-methods design combining a systematic literature review with semi-structured questionnaires. First, a PRISMA-guided systematic review was conducted, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses protocol to ensure transparency, rigor, and replicability. From an initial pool of 362 screened records, 42 peer-reviewed studies published between 2015 and early 2025 met the inclusion criteria and were included in the final review. The systematic review captured the existing evidence base on deepfake AI tutors, including their pedagogical applications, detection methods, governance challenges, and ethical implications. Second, to complement and contextualize the findings from the literature, the study administered semi-structured questionnaires to 12 assistant professors with a mean teaching and research experience of 7 years. These expert informants provided practitioner perspectives on the practical challenges, institutional responses, and governance needs related to deepfake AI tutors in real educational settings. Thematic analysis was conducted using deductive codes derived from the research questions and theoretical framework. The analysis achieved strong inter-coder reliability, with Cohen\u0026apos;s kappa (\u0026kappa;) = 0.81, indicating substantial agreement between independent coders and lending confidence to the trustworthiness of the thematic findings.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe analysis yielded four major themes. First, personalization and engagement: The results indicate that deepfake AI tutors enhance student engagement, adaptability to individual learning needs, and scalability across large and diverse student populations, making them attractive for online education contexts where human tutor resources are limited. Second, detection challenges and integrity risks: Deepfake AI tutors pose significant risks including impersonation (e.g., fake instructor identities), assessment fraud (e.g., students using deepfake-generated content to cheat), and algorithmic bias (e.g., differential performance or recommendations across student demographic groups). Current detection approaches based on pixel-level artifacts (visual inconsistencies invisible to the naked eye), frequency features (spectral patterns in images or audio), and physiological signals (e.g., eye movement, heart rate, or speech patterns) remain imperfect, meaning that no existing detection method is fully reliable in distinguishing authentic from deepfake-generated educational content. Third, governance and policy gaps: The literature and expert responses revealed that most educational institutions lack specific policies, guidelines, or governance structures for deepfake AI tutors, creating regulatory voids that increase institutional risk and vulnerability. Fourth, ethical and societal implications: Beyond technical and governance concerns, deepfake AI tutors raise deeper ethical questions about informed consent (students may not know they are interacting with a deepfake), transparency (whether and how deepfake identity should be disclosed), trust erosion (potential long-term damage to student trust in online education), and equitable access (whether deepfake tutors might widen or narrow digital divides). To mitigate these challenges, the study proposes a four-pillar governance framework encompassing Transparency and Disclosure (e.g., clear labeling of deepfake AI tutors), Data Governance and Privacy (e.g., strict controls on student data used to personalize deepfake interactions), Integrity and Detection (e.g., investment in fairness-aware detection systems), and Ethical Oversight and Accountability (e.g., designated responsibility for harms caused by deepfake tutors). This framework is supported by a policy checklist for institutional implementation, a responsibility matrix clarifying roles across stakeholders (e.g., institutions, developers, instructors, students), and a risk-tier model for classifying deepfake applications by their potential for harm.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with significant theoretical, practical, and societal implications. First, the findings demonstrate that deepfake AI tutors hold genuine promise for expanding access to education by providing personalized, multilingual, and scalable instruction that could reach learners in underserved regions or resource-constrained settings. However, this promise is conditional on addressing substantial risks. Second, fairness-aware detection systems that is, detection algorithms that are validated for equitable performance across different demographic groups, accents, languages, and presentation styles are essential to avoid biased outcomes that could disadvantage certain student populations. Robust safeguards, including technical, procedural, and governance mechanisms, must be implemented before widespread deployment. Third, AI literacy initiatives for both educators and learners are critical to sustain trust and ensure equitable adoption. Students need to understand what deepfake AI tutors are, how they work, their limitations, and how to identify potential misuse; instructors need the knowledge and skills to select, implement, and oversee these tools responsibly. Fourth, from a theoretical and policy perspective, the proposed four-pillar governance framework with its accompanying policy checklist, responsibility matrix, and risk-tier model offers a practical, actionable tool for institutions seeking to navigate the complex ethical and governance landscape of deepfake AI tutors. This framework not only strengthens the ethical and governance foundations for generative AI in higher education but also contributes to the broader agenda of sustainable digital education by promoting transparency, fairness, and equitable access. Fifth, by advancing responsible innovation and institutional resilience, the proposed framework directly supports the United Nations Sustainable Development Goal 4 (Quality Education), ensuring that technological progress in education does not come at the cost of integrity, privacy, or trust. However, the study acknowledges limitations: the sample of 42 peer-reviewed studies, while systematically selected, may not capture the most recent developments given the rapid pace of generative AI innovation; the expert sample of 12 assistant professors, while experienced, may not represent the full range of stakeholder perspectives including students, instructional designers, administrators, or policymakers in different institutional and national contexts. Future research should therefore: (a) conduct empirical validations of the four-pillar governance framework in real educational settings, (b) develop and test fairness-aware detection algorithms specifically designed for educational contexts, (c) investigate student and instructor perceptions of deepfake AI tutors through large-scale surveys and longitudinal studies, (d) examine cross-cultural variations in ethical norms, privacy expectations, and governance preferences related to deepfake technologies, (e) explore the long-term impact of deepfake AI tutors on learning outcomes, trust in educational institutions, and the teacher-student relationship, and (f) study how emerging regulatory frameworks (e.g., EU AI Act) interact with institutional governance of deepfake AI tutors in education.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e14.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eAI-Powered Data Engineering for Intelligent Retail Stock Management\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Mashetty \u0026amp; Valiki, 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to address inventory optimization as an artificial intelligence problem within the context of smart retail, which integrates in-store and online components to provide support services that complement (rather than duplicate) customer value and convenience. The study recognized that effective inventory optimization must consider customer needs for timely product availability without long delivery lead times, particularly for long-lead-time products where delays between ordering and receiving inventory can significantly impact service levels. Second, the study sought to reformulate inventory optimization as a recommendation engine, shifting from traditional forecasting approaches to a system that predicts future warehouse, store, and web inventory levels in the short and medium term, thereby assisting decision-makers in determining how much to order and when. Third, the research aimed to introduce a concept for a fully receptive data architecture capable of supplying the large amount of quality-cleaned data required to train AI models and to implement AI-based data pipelines that spatially distribute web inventory recommendations across the supply chain, ultimately accelerating inventory-level refresh rates by making large amounts of inventory-ready data locally available.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study presents a conceptual design and architectural framework rather than an empirical experimental method. The research introduces a novel data architecture concept specifically designed to support smart retail operations. The proposed architecture is described as \u0026nbsp;\u0026ldquo;fully receptive, \u0026ldquo; meaning it is designed to ingest and process both external and internal data flows comprehensively. The architecture consists of three main components. First, data-ops pipelines are designed for fully receptive external and internal data flows, enabling the systematic collection, cleaning, and integration of data from multiple sources including in-store systems, online platforms, supply chain partners, and external market data. Second, data-engineering lines are dedicated to preparatory and loading jobs for core business intelligence (BI) information, ensuring that foundational data is properly structured, validated, and made available for analytical and operational purposes. Third, an additional component is dedicated to the implementation of AI-based data pipelines, sized specifically to cope with the spatiotemporal distribution throughout the modelled area of slow-loading-tagged external data (i.e., external data that is geographically dispersed, time-sensitive, and computationally expensive to load). The architecture is optimized for fast local machine learning workloads, reducing the volume of data sent to the core database and minimizing the number of processing jobs initiated at the central level, thereby enabling faster inventory-level refresh and supporting local supply-demand analysis.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe study presents several key results in the form of architectural capabilities and performance improvements. First, the proposed fully receptive data architecture enables the reformulation of inventory optimization as a recommendation engine, meaning that instead of simply forecasting demand, the system can actively recommend optimal inventory levels for warehouses, stores, and web channels based on predicted short-term and medium-term needs. Second, the architecture successfully addresses the challenge of long-lead-time products by predicting future inventory levels and assisting ordering decisions, thereby reducing the risk of stockouts (product unavailability) or overstocking (excess inventory carrying costs) while meeting customer expectations for timely delivery. Third, the AI-based data pipelines are designed to spatially distribute web inventory recommendations across the supply chain, meaning that recommendations for inventory allocation are not centralized but are generated and applied at appropriate geographic points in the supply chain network. Fourth, the architecture achieves accelerated inventory-level refresh by minimizing the volume of data sent to the core database and reducing the number of jobs initiated there. This optimization enables core data availability that supports fast local machine learning workloads and local supply-demand analysis. Fifth, by making large amounts of inventory-ready data locally available, the architecture reduces latency between data collection and decision-making, enabling more responsive and granular inventory management across distributed retail operations.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions regarding the design and implementation of AI-enabled inventory optimization in smart retail. First, the conceptual architecture demonstrates that inventory optimization can be effectively reframed as a recommendation problem rather than a pure forecasting problem, shifting the analytical focus from predicting what will happen to prescribing what actions should be taken. This reframing aligns inventory management more closely with customer-centric values of convenience and timely product availability. Second, the fully receptive data architecture addresses a critical bottleneck in AI-based retail systems: the need for large volumes of quality-cleaned, spatiotemporally distributed data that can be processed efficiently across the supply chain. By minimizing central data movement and enabling local ML workloads, the architecture is well-suited to the distributed, real-time demands of modern smart retail environments. Third, the separation of the architecture into three components data-ops pipelines for external/internal flows, data-engineering lines for BI preparation, and AI-based pipelines for spatiotemporal distribution provides a modular, scalable design that can be implemented incrementally. This modularity allows retailers to adopt components based on their specific maturity levels, data availability, and operational priorities. Fourth, the emphasis on local machine learning workloads represents a departure from fully centralized AI architectures, recognizing that inventory decisions often need to be made at local levels (e.g., individual stores or regional warehouses) with minimal latency and without dependency on central database availability or performance. However, the study acknowledges that this is a conceptual architecture rather than an empirically validated implementation. The findings are theoretical and design-oriented, meaning that the claimed performance improvements (e.g., accelerated inventory refresh, reduced data movement) remain to be tested in real-world smart retail settings. Future research should therefore: (a) implement and validate the proposed architecture in live retail environments across different retail sectors (e.g., grocery, apparel, electronics), (b) quantify the actual reductions in data volume, job initiation, and inventory refresh latency achieved by the architecture, (c) compare the recommendation-engine approach to traditional forecasting methods in terms of inventory costs, service levels, and customer satisfaction, (d) investigate the scalability of the architecture as the number of stores, warehouses, web channels, and product SKUs increases, (e) examine how the architecture handles data quality issues, missing data, and concept drift (changes in customer behavior over time), (f) explore integration with existing enterprise resource planning (ERP) and warehouse management systems, and (g) assess the cybersecurity and data privacy implications of distributing inventory data and ML workloads across multiple local nodes rather than centralizing them.\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 36px;\"\u003e15.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 128px;\"\u003eAI and Human-AI Collaboration in Enterprise Integration and Document Automation\u003cbr\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e(Nagubathula, 2025)\u003cbr\u003e(Nagubathula, 2025)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 179px;\"\u003eThis study had three primary objectives. First, it aimed to introduce a holistic, integrated theoretical framework that synthesizes multiple management theories and incorporates machine learning models to develop a compensation model capable of accurately predicting pay determination in the information technology industry. This objective arose from the context of geopolitical uncertainty, pandemic-induced economic disruptions, alarming attrition rates, and aggravating talent gaps that have spurred a surge in demand for specialized digital proficiencies, leading firms to seek ways to attract and retain top-tier talent through generous compensation packages. Second, the study sought to interrogate the multifaceted factors that shape pay determination including experience level, educational background, specialized skill sets, gender, company size, and company type to determine which factors truly drive compensation and which do not. Third, beyond prediction, the research aimed to provide practical value by empowering individuals to negotiate compensation more effectively, supporting enterprises in crafting targeted compensation strategies, and ultimately facilitating sustainable economic growth while helping to attain various Sustainable Development Goals (SDGs) related to decent work and economic growth.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 141px;\"\u003eThe study employed a quantitative, predictive modeling approach using a stratified sample of 2,488 observations drawn from the information technology sector. The sampling strategy ensured representation across different levels of experience, educational backgrounds, skill sets, company sizes, company types, and genders to capture the full range of variability in compensation determination. The research question was whether compensation could be accurately predicted using constructs derived from the integrated theoretical framework (which synthesized multiple management theories to capture the complexity of pay determination). To answer this question, the study employed various cutting-edge machine learning models, including but not limited to random forest, support vector machines, neural networks, gradient boosting, and regression-based algorithms. Each model was trained on a portion of the dataset and tested on a held-out portion to evaluate predictive accuracy. The models were compared against each other to identify the best-performing algorithm. A series of comprehensive robustness checks were conducted to ensure the stability and reliability of the findings, including cross-validation, sensitivity analyses, and tests for overfitting. The final model selection was based on two key performance metrics: prediction accuracy (percentage of correctly predicted compensation outcomes) and mean absolute error (the average magnitude of prediction errors in the original measurement units).\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 161px;\"\u003eThe empirical findings of this study yielded several critical results. First, regarding model performance, the research culminated in the discovery that the random forest model outperformed all other machine learning algorithms tested, achieving an exceptionally high accuracy of 99.6% and a remarkably low mean absolute error of 0.08 degrees (presumably in the relevant compensation units, such as thousands of currency units or log-transformed values). This indicates that the random forest model can predict individual compensation with near-perfect precision based on the constructs derived from the integrated theoretical framework. Second, concerning the determinants of compensation, the study identified several critical predictors including, but not limited to, experience level, educational background, and specialized skill-set. These factors were found to have substantial influence on pay determination. Third, and notably, the research elucidated that gender does not play a role in pay disparity, suggesting that within the sampled IT sector context, there is no evidence of gender-based compensation discrimination after accounting for other relevant factors. Fourth, the study found that company size and company type hold no consequential sway over individual compensation determination, meaning that whether an employee works for a large multinational corporation or a small startup, or for a product company versus a service company, does not significantly affect their individual pay when experience, education, and skills are accounted for.\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 163px;\"\u003eThe study draws several conclusions with significant theoretical, practical, and original contributions. From a theoretical perspective, the cardinal contribution of this research lies in the inception of an inclusive theoretical framework that persuasively explicates the intricacies of a machine learning-driven remuneration model. This framework is ennobled by the synthesis of diverse management theories likely including human capital theory, signaling theory, equity theory, and resource-based view to capture the full complexity of compensation determination in the modern IT industry. By integrating these theoretical perspectives with advanced machine learning methods, the study bridges a gap between traditional econometric compensation studies and contemporary predictive analytics. From a practical implications perspective, the research underscores the importance of equitable compensation to foster technological innovation and encourage the retention of top talent, emphasizing the significance of human capital as a strategic asset. The highly accurate random forest model presented in this study empowers individuals to negotiate their compensation more effectively by providing them with evidence-based benchmarks. Simultaneously, the model supports enterprises in crafting targeted compensation strategies that reward the factors that truly matter (experience, education, specialized skills) while avoiding discrimination on irrelevant factors (gender, company size, company type). This alignment with equitable pay practices facilitates sustainable economic growth and helps attain various Sustainable Development Goals, particularly SDG 8 (Decent Work and Economic Growth) and SDG 5 (Gender Equality), given the finding that gender does not drive pay disparity. However, the study acknowledges a significant limitation: the generalizability of the findings to other sectors is constrained, as this study is exclusively limited to the IT sector. Future research should extend the integrated theoretical framework and machine learning methodology to other industries such as healthcare, finance, manufacturing, and education to test whether the same determinants of compensation operate similarly or whether sector-specific factors emerge. Additionally, future studies should explore longitudinal data to examine how compensation determinants evolve over time with technological change, as well as cross-country comparisons to investigate how institutional, cultural, and regulatory contexts moderate the relationships identified in this study.\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eDiscussion\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis systematic literature review set out to examine how AI-driven analytics can be integrated into teaching factory quality assurance to enable real-time validation of student procedural skills. The discussion is organized around the four research questions, interpreting the thematic findings in light of existing theoretical frameworks and practical implementation contexts.\u003c/p\u003e\n\u003cp\u003eThe finding that computer vision dominates current AI applications for assessing hands-on procedural skills aligns with broader trends in educational technology and Industry 4.0 research. Computer vision\u0026apos;s prevalence can be explained by its relative maturity, decreasing hardware costs, and intuitive alignment with human observational assessment practices. Instructors naturally assess students by watching their actions; computer vision automates and scales this process. However, the absence of acoustic and vibrational analysis in the reviewed studies represents a notable gap. In many vocational domains such as automotive repair, machining, and equipment maintenance auditory and tactile cues are critical indicators of correct performance. A skilled mechanic hears whether an engine is running correctly; a machinist feels whether a cutting tool is engaging properly. The lack of research on these modalities suggests that current AI assessment systems may be missing essential dimensions of procedural competence.\u003c/p\u003e\n\u003cp\u003eThe nascent state of multimodal AI fusion identified in this review points to both a limitation and an opportunity. While single-modality systems (e.g., computer vision alone) can capture surface-level actions, they cannot assess the integration of multiple sensory inputs that characterize expert performance. Future teaching factories should prioritize multimodal systems that combine vision, force, acoustics, and even biometric data to create holistic competency profiles. Article 13\u0026apos;s deepfake AI tutors and Article 14\u0026apos;s fully receptive data architecture represent early steps toward multimodal integration, but substantial work remains to translate these concepts into practical teaching factory applications.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe three conceptual framings identified temporal shift, shift in assessment object, and reframing of quality itself represent a developmental progression in how researchers and practitioners understand quality assurance in vocational education. The temporal shift (from end-point to continuous assessment) is the most accessible and frequently implemented, as it requires only changes in assessment timing rather than fundamental redefinition of quality. The shift in assessment object (from artifact to action) represents a deeper reconceptualization, requiring stakeholders to value how a student works as much as what they produce. The most profound reframing quality as competence rather than conformity challenges the very purpose of vocational education: is it to produce conforming products or competent practitioners?\u003c/p\u003e\n\u003cp\u003eThis progression has important implications for teaching factory design. A teaching factory that adopts only the temporal shift might implement continuous monitoring but still evaluate students based on whether final products meet specifications. A teaching factory that embraces the competence reframing would design curricula, assessment rubrics, and feedback mechanisms around skill mastery, using product quality only as one indicator among many. The finding that operational definitions of \u0026nbsp;\u0026ldquo;continuous assessment \u0026ldquo; remain under-specified suggests that future research should develop clear metrics for what constitutes continuous versus episodic assessment in authentic teaching factory environments. Article 4\u0026apos;s STICT approach provides a promising example, where systems thinking applied during design and evaluation stages enabled ongoing feedback rather than end-of-project grading.\u003c/p\u003e\n\u003cp\u003eThe moderate positive effect size (Hedges\u0026apos; *g* = 0.72) reported in Article 1 is encouraging but must be interpreted with caution. In educational research, effect sizes of 0.2 are considered small, 0.5 moderate, and 0.8 large. A *g* of 0.72 approaches the large threshold, suggesting that AI-supported assessment systems have meaningful educational benefits. However, the substantial statistical heterogeneity reported means that this average effect masks considerable variation. Some AI systems work very well in certain contexts; others may be ineffective or even detrimental. The challenge for researchers and practitioners is to identify the conditions under which AI assessment produces benefits versus those where it does not.\u003c/p\u003e\n\u003cp\u003eThe gap between positive attitudes towards diagnostics and low self-efficacy among pre-service teachers (Article 5) reveals a critical psychological barrier to AI adoption. This finding aligns with self-determination theory, which posits that competence (feeling capable) and autonomy (feeling in control) are essential for intrinsic motivation. Pre-service teachers who recognize the importance of diagnostic skills but do not believe they can perform them are unlikely to embrace AI tools designed to support those skills. Professional development must therefore address not only technical skills but also self-efficacy beliefs through mastery experiences, vicarious learning, and supportive feedback.\u003c/p\u003e\n\u003cp\u003eThe ethical challenges identified particularly transparency, informed consent, algorithmic bias, and trust erosion are not merely implementation details but fundamental governance issues. The finding that most educational institutions lack specific policies for AI assessment tools (Article 13) suggests that practice is outpacing policy. This creates institutional risk and leaves students and instructors without clear guidance or recourse when problems arise. Article 12\u0026apos;s identification of insufficient bottom-up perspectives in AI literacy frameworks further compounds this problem, as policies developed without teacher and student input are unlikely to address their actual concerns or needs.\u003c/p\u003e\n\u003cp\u003eThe synthesized design principles and frameworks represent the most actionable contribution of this review. The Integrated AI-Education Convergence Framework (Article 12) provides a high-level strategic orientation, emphasizing that AI integration must be pedagogy-centric, ethically grounded, and contextually responsive. The Four-Pillar Governance Framework (Article 13) offers operational guidance for institutions implementing deepfake or generative AI tutors, covering transparency, data governance, integrity and detection, and ethical oversight. The Fully Receptive Data Architecture (Article 14) provides technical specifications for the data infrastructure required to support AI assessment at scale, including data-ops pipelines, data-engineering lines, and AI-based pipelines for spatiotemporal distribution.\u003c/p\u003e\n\u003cp\u003eSeveral cross-cutting themes emerge across these frameworks. First, transparency is consistently emphasized as a non-negotiable requirement. Students and instructors must understand how AI systems make decisions, what data they use, and what their limitations are. Second, human-in-the-loop design preserves instructor authority and enables handling of contextual nuances that AI cannot yet address. Third, fairness-aware design requires that AI systems be validated for equitable performance across demographic groups, preventing algorithmic bias from exacerbating existing educational inequities. Fourth, professional development must address both technical skills and self-efficacy beliefs, as identified in Article 5.\u003c/p\u003e\n\u003cp\u003eThe findings of this review extend and refine existing theoretical frameworks in vocational education research. The findings extend technological pedagogical content knowledge (TPACK) framework by specifying what technological, pedagogical, and content knowledge means in the context of AI-enhanced teaching factories. Technological knowledge includes understanding AI modalities (computer vision, sensors, etc.) and their capabilities and limitations. Pedagogical knowledge includes designing feedback loops that support learning without causing cognitive overload or anxiety. Content knowledge includes domain-specific procedural skills and the ability to decompose them into measurable sub-skills. The intersection TPACK for AI assessment requires instructors to know how to select, configure, and interpret AI assessment tools for their specific subject area and student population.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe findings also extend self-determination theory by demonstrating that perceived competence (self-efficacy) is as important as actual competence for successful AI adoption. Article 5\u0026apos;s finding that pre-service teachers valued diagnostics but felt incapable of performing them suggests that AI tools must be designed not only to be accurate but also to support user confidence through clear explanations, gradual complexity, and success experiences.\u003c/p\u003e\n\u003cp\u003eSeveral limitations must be acknowledged. First, the review sample was drawn from 15 reference articles, which, while systematically selected, may not capture the full breadth of AI assessment research in teaching factories. The exclusion of non-English publications and gray literature may have introduced publication bias. Second, the substantial methodological heterogeneity across primary studies limited the feasibility of meta-analysis, forcing reliance on narrative synthesis. Third, most included studies were conducted in high-income countries (Europe, Australia, China), limiting generalizability to low- and middle-income contexts where vocational education resources and AI infrastructure differ significantly. Fourth, the review focused on AI analytics for skills validation but did not systematically examine cost-effectiveness, scalability beyond research settings, or long-term sustainability of AI assessment systems. Fifth, Article 15 lacked complete information and could not be fully utilized in the synthesis.\u003c/p\u003e\n\u003cp\u003eBased on the discussion above, several recommendations for future research emerge. First, researchers should conduct longitudinal studies tracking the impact of AI-enhanced process-oriented QA on student learning outcomes, instructor practices, and employment outcomes over multiple years. Second, cross-contextual comparative studies should examine how the same AI assessment systems perform in different vocational domains (e.g., healthcare vs. manufacturing vs. culinary arts) and different cultural/institutional settings. Third, researchers should develop and validate instruments for measuring instructor AI literacy and self-efficacy, enabling targeted professional development. Fourth, design-based research should iteratively develop and test multimodal AI assessment systems in authentic teaching factory environments, documenting both successes and failures. Fifth, ethical frameworks should be translated into practical audit tools that institutions can use to assess AI assessment systems for bias, transparency, and accountability before deployment. Sixth, future research should amplify bottom-up perspectives by systematically collecting and integrating the voices of teachers and students into AI literacy frameworks and professional development designs, as recommended in Article 12.\u003c/p\u003e\n\u003cp\u003eFor vocational school administrators and teaching factory managers, the findings suggest several actionable steps. First, before investing in AI assessment technologies, institutions should conduct a needs assessment to determine which skills are most critical to assess and which AI modality best aligns with those skills. Second, institutions should develop clear policies for data governance, informed consent, algorithmic transparency, and accountability before deploying AI assessment tools. Third, professional development for instructors should address both technical skills (how to use AI tools) and pedagogical skills (how to interpret AI-generated data and integrate it into feedback and grading), as well as self-efficacy beliefs. Fourth, institutions should pilot AI assessment systems in low-stakes contexts first, gradually scaling up as confidence and competence increase. Fifth, student voices should be incorporated into AI assessment design and evaluation, ensuring that tools serve learner needs rather than merely institutional monitoring interests. Sixth, institutions should consider adopting the Four-Pillar Governance Framework (Article 13) and the Integrated AI-Education Convergence Framework (Article 12) as guiding documents for responsible AI integration.\u003c/p\u003e\n\u003cp\u003eIn conclusion, this systematic review demonstrates that AI-driven analytics hold significant promise for transforming teaching factory quality assurance from product-oriented to process-oriented, real-time skills validation. Computer vision is currently the dominant AI modality, but multimodal fusion represents the future frontier. The transition from product to process QA requires not only technological change but also conceptual reframing of what quality means in vocational education. Benefits are real but conditional on addressing substantial technical, pedagogical, ethical, and organizational challenges. Synthesized design principles and frameworks provide actionable guidance for responsible implementation. However, the evidence base remains fragmented, and claims about effectiveness must be qualified by contextual and design considerations. As AI technologies continue to evolve rapidly, sustained, rigorous, and transparent research is essential to realize the promise of AI-enhanced process-centric QA in teaching factories.\u003c/p\u003e"},{"header":"D. Conclusions","content":"\u003cp\u003eThis systematic review concludes that AI-driven analytics hold significant promise for transforming teaching factory quality assurance from product-oriented to process-oriented, real-time skills validation. The key finding reveals a moderate positive effect of AI-supported assessment on skill-related learning outcomes (Hedges' *g* = 0.72), with computer vision currently dominating as the primary AI modality. However, the transition from product to process quality assurance remains conceptually well-developed but operationally under-specified. Four interrelated research issues persist: limited congruence between technological and pedagogical affordances, insufficient bottom-up perspectives in AI literacy frameworks, lack of explicit AI ethics interpretation, and inadequate professional development frameworks. Additionally, a significant gap exists between pre-service teachers' positive attitudes toward diagnostics and their low self-efficacy in performing diagnostic activities.\u003c/p\u003e \u003cp\u003eFrom a practical perspective, vocational institutions should adopt a pedagogy-centric, ethically grounded, and contextually responsive approach to AI integration. Administrators must develop clear data governance policies, informed consent protocols, and algorithmic transparency mechanisms before deployment. Professional development programs should address both technical AI skills and instructor self-efficacy, recognizing that confidence in using diagnostic tools is as important as technical competence. The proposed Integrated AI-Education Convergence Framework (Article 12) and Four-Pillar Governance Framework (Article 13) offer actionable guidance for responsible implementation, emphasizing transparency, data governance, integrity and detection, and ethical oversight.\u003c/p\u003e \u003cp\u003eFuture research should prioritize longitudinal studies tracking sustained impacts of AI-enhanced quality assurance on learning outcomes and employment trajectories. Researchers must develop and validate instruments for measuring instructor AI literacy and self-efficacy, conduct design-based research on multimodal AI assessment systems in authentic teaching factory environments, and translate ethical frameworks into practical audit tools for bias detection. Crucially, bottom-up perspectives from teachers and students must be systematically incorporated into AI literacy frameworks and professional development designs to ensure equitable, transparent, and effective AI integration in vocational education.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank to all friends who helped us in this valuable article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBiographies of Authors\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMeri Silvia\u003c/strong\u003eis a physics teacher at SMAN Muara Kulam, North Musi Rawas, South Sumatra, Indonesia. She earned her Master\u0026apos;s degree in Educational Administration from Universitas Bengkulu, where she is currently a doctoral candidate. Her research interests include academic supervision, educational technology integration, and digital equity in secondary education. She served as the corresponding author and contributed to the conceptualization, methodology, and writing of the original draft of this systematic review. ORCID: 0009-0004-6604-859X.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDr. Muhammad Kristiawan, M.Pd.\u003c/strong\u003e is an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research area is about leadership, supervision, management of education, administration of education, science of education, social science and technology. He has ever performed as international speakers in many international conferences both India, Philippines, Vietnam, and Indonesia. He has reviewed a lot of articles from reputable international journal which published by Sage, Frontiers, Springer Nature, Taylor \u0026amp; Francis, Humanities \u0026amp; Social Sciences Communications, Npj Climate Action, Intelektual Pustaka Media Utama, Journal of Learning for Development, International Journal of Learning, Teaching and Educational Research and others. For more detail information about Muhammad Kristiawan, you can contact e-mail: [email protected] Orcid ID: https://orcid.org/0000-0002-1077-4013. See the following link Google scholar: https://scholar.google.com/citations?user=Wv7tx2kAAAAJ\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEko Risdianto\u0026nbsp;\u003c/strong\u003eis an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research interests encompass artificial intelligence in education, learning management systems, and digital pedagogy. He contributed to the data analysis and interpretation of findings related to AI and EdTech integration.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMeri Silvia\u003c/strong\u003e\u003cstrong\u003e\u003csup\u003e1\u003c/sup\u003e\u003c/strong\u003e\u003cstrong\u003e,\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eMuhammad Kristiawan\u003csup\u003e2\u003c/sup\u003e, Eko Risdianto\u003csup\u003e3\u003c/sup\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e123\u003c/sup\u003eUniversitas Bengkulu, Bengkulu, Indonesia.\u003c/p\u003e\n\u003cp\u003e*Corresponding author e-mail:\u003c/p\u003e\n\u003cp\[email protected]\u003c/p\u003e\n\u003cp\[email protected]\u003c/p\u003e\n\u003cp\[email protected]\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical Approval and Consent to participate:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research has been approved by the Principal of SMAN Muara Kulam, Nort Musi Rawas, South Sumatera, Indonesia.\u003c/p\u003e\n\u003cp\u003e(approval no. 420/1011/SMAN.MK/Disdik.S/2025) on December 10, 2025.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe Principal\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eALPATI\u003c/strong\u003e\u003cstrong\u003e,\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eS\u003c/strong\u003e\u003cstrong\u003e. Pd\u003c/strong\u003e\u003cstrong\u003e.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for Publication Statement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003enot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of Supporting Data:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors confirm that the data supporting the findings of this study are available within the article [and/or] its supplementary materials.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. The authors declare no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author(s) received no financial support for the research, authorship, and/or publication of this article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; Contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMeri Silvia: \u003cem\u003eStudy design, Funds collection, and Manuscript preparation\u003c/em\u003e;\u003c/p\u003e\n\u003cp\u003eMuhammad Kristiawan: \u003cem\u003eData collection and Manuscript preparation\u003c/em\u003e;\u003c/p\u003e\n\u003cp\u003eEko Risdianto: \u003cem\u003eData ollection and analysis.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe acknowledgement to the Principal of SMAN Muara Kulam, Nort Musi Rawas, South Sumatera, Indonesia. Who has given us the opportunity for the research project with the Decree Number 420/1011/SMAN.MK/Disdik.SS/2025, and all respondents and colleagues who have helped us in this meaningful project.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; Information:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFirst author\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMeri Silvia\u003c/strong\u003eis a physics teacher at SMAN Muara Kulam, North Musi Rawas, South Sumatra, Indonesia. She earned her Master\u0026apos;s degree in Educational Administration from Universitas Bengkulu, where she is currently a doctoral candidate. Her research interests include academic supervision, educational technology integration, and digital equity in secondary education. She served as the corresponding author and contributed to the conceptualization, methodology, and writing of the original draft of this systematic review. ORCID: 0009-0004-6604-859X.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSecond author\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDr. Muhammad Kristiawan, M.Pd.\u003c/strong\u003e is an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research area is about leadership, supervision, management of education, administration of education, science of education, social science and technology. He has ever performed as international speakers in many international conferences both India, Philippines, Vietnam, and Indonesia. He has reviewed a lot of articles from reputable international journal which published by Sage, Frontiers, Springer Nature, Taylor \u0026amp; Francis, Humanities \u0026amp; Social Sciences Communications, Npj Climate Action, Intelektual Pustaka Media Utama, Journal of Learning for Development, International Journal of Learning, Teaching and Educational Research and others. For more detail information about Muhammad Kristiawan, you can contact e-mail: [email protected] Orcid ID: https://orcid.org/0000-0002-1077-4013. See the following link Google scholar: https://scholar.google.com/citations?user=Wv7tx2kAAAAJ\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThird author\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEko Risdianto\u0026nbsp;\u003c/strong\u003eis an Associate Professor of Education in Universitas Bengkulu. Currently he serves as lecturer in Master of Educational Administration Study Program of Universitas Bengkulu. His research interests encompass artificial intelligence in education, learning management systems, and digital pedagogy. He contributed to the data analysis and interpretation of findings related to AI and EdTech integration.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cu\u003eEthics in Publishing Statement\u003c/u\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eI testify on behalf of all co-authors that our article submitted followed ethical principles in publishing.\u003c/p\u003e\n\u003cp\u003eTitle: \u003cstrong\u003eFrom Product to Process: Integrating AI-Driven Analytics into Teaching Factory Quality Assurance for Real-Time Skills Validation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors agree that:\u003c/p\u003e\n\u003cp\u003eThis research presents an accurate account of the work performed, all data presented are accurate and methodologies detailed enough to permit others to replicate the work.\u003c/p\u003e\n\u003cp\u003eThis manuscript represents entirely original works and or if work and/or words of others have been used, that this has been appropriately cited or quoted and permission has been obtained where necessary.\u003c/p\u003e\n\u003cp\u003eThis material has not been published in whole or in part elsewhere.\u003c/p\u003e\n\u003cp\u003eThe manuscript is not currently being considered for publication in another journal.\u003c/p\u003e\n\u003cp\u003eThat generative AI and AI-assisted technologies have not been utilized in the writing process or if used, disclosed in the manuscript the use of AI and AI-assisted technologies and a statement will appear in the published work.\u003c/p\u003e\n\u003cp\u003eThat generative AI and AI-assisted technologies have not been used to create or alter images unless specifically used as part of the research design where such use must be described in a reproducible manner in the methods section.\u003c/p\u003e\n\u003cp\u003eAll authors have been personally and actively involved in substantive work leading to the manuscript and will hold themselves jointly and individually responsible for its content.\u003c/p\u003e\n\u003cp\u003eCorresponding author\u0026rsquo;s name: Meri Silvia\u003c/p\u003e\n\u003cp\u003eDate: 18 April 2026\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Interest Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cspan lang=\"IN\"\u003e☒\u003c/span\u003e The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cspan lang=\"IN\"\u003e☐\u003c/span\u003e The author is an Editorial Board Member/Editor-in-Chief/Associate Editor/Guest Editor for this journal and was not involved in the editorial review or the decision to publish this article.\u003c/p\u003e\n\u003cp\u003e☐ The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbdelalim, A. M., Essawy, A., Alnaser, A. A., Shibeika, A., \u0026amp; Sherif, A. (2024). Digital Trio: Integration of BIM\u0026ndash;EIR\u0026ndash;IoT for Facilities Management of Mega Construction Projects. \u003cem\u003eSustainability\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(15), 6348. https://doi.org/10.3390/su16156348\u003c/li\u003e\n\u003cli\u003eAboobaker, N., D., R., \u0026amp; K.A., Z. (2023). Fostering entrepreneurial mindsets: the impact of learning motivation, personal innovativeness, technological self-efficacy, and human capital on entrepreneurial intention. \u003cem\u003eJournal of International Education in Business\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(3), 312\u0026ndash;333. https://doi.org/10.1108/JIEB-10-2022-0071\u003c/li\u003e\n\u003cli\u003eAli, M., Ma, M., Muneeb, M., \u0026amp; Wong, G. K. W. (2025). Mapping Contemporary AI-Education Intersections and Developing an Integrated Convergence Framework: A Bibliometric-Driven and Inductive Content Analysis. \u003cem\u003eMetrics\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(4), 23. https://doi.org/10.3390/metrics2040023\u003c/li\u003e\n\u003cli\u003eArbel, Y., Hong, L., Baker, T. E., \u0026amp; Holroyd, C. B. (2017). It\u0026rsquo;s all about timing: An electrophysiological examination of feedback-based learning with immediate and delayed feedback. \u003cem\u003eNeuropsychologia\u003c/em\u003e, \u003cem\u003e99\u003c/em\u003e, 179\u0026ndash;186. https://doi.org/10.1016/j.neuropsychologia.2017.03.003\u003c/li\u003e\n\u003cli\u003eAzevedo, R. (2026). Using real-time process data of domain-specific learning processes to provide adaptive support for learning and instruction: Challenges and opportunities. \u003cem\u003eLearning and Instruction\u003c/em\u003e, \u003cem\u003e104\u003c/em\u003e, 102363. https://doi.org/10.1016/j.learninstruc.2026.102363\u003c/li\u003e\n\u003cli\u003eBarbosa, M. W., Bronzo, M., J\u0026uacute;nior, N. T., \u0026amp; de Sousa, P. R. (2025). Predictors of Corporate Reputation: Circular Economy, Environmental, Social, and Governance, and Collaborative Relationships in Brazilian Agribusiness. \u003cem\u003eSustainability\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(7), 2969. https://doi.org/10.3390/su17072969\u003c/li\u003e\n\u003cli\u003eBattini, D., Faccio, M., Persona, A., \u0026amp; Sgarbossa, F. (2012). Design of an integrated quality assurance strategy in production systems. \u003cem\u003eInternational Journal of Production Research\u003c/em\u003e, \u003cem\u003e50\u003c/em\u003e(6), 1682\u0026ndash;1701. https://doi.org/10.1080/00207543.2011.555428\u003c/li\u003e\n\u003cli\u003eBehl, A., Pereira, V., Jayawardena, N., Nigam, A., \u0026amp; Mangla, S. (2024). Gamification as an innovation: a tool to improve organizational marketing performance and sustainability of international firms. \u003cem\u003eInternational Marketing Review\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(1), 107\u0026ndash;137. https://doi.org/10.1108/IMR-05-2022-0113\u003c/li\u003e\n\u003cli\u003eBondin, A., \u0026amp; Zammit, J. P. (2025). Education 4.0 for Industry 4.0: A Mixed Reality Framework for Workforce Readiness in Manufacturing. \u003cem\u003eMultimodal Technologies and Interaction\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(5), 43. https://doi.org/10.3390/mti9050043\u003c/li\u003e\n\u003cli\u003eBrauner, P., Philipsen, R., Fels, A., Fuhrmann, M., Ngo, Q. H., Stiller, S., Schmitt, R., \u0026amp; Ziefle, M. (2016). A Game-Based Approach to Raise Quality Awareness in Ramp-Up Processes. \u003cem\u003eQuality Management Journal\u003c/em\u003e, \u003cem\u003e23\u003c/em\u003e(1), 55\u0026ndash;69. https://doi.org/10.1080/10686967.2016.11918462\u003c/li\u003e\n\u003cli\u003eCaiazzo, B., Murino, T., Petrillo, A., Piccirillo, G., \u0026amp; Santini, S. (2023). An IoT-based and cloud-assisted AI-driven monitoring platform for smart manufacturing: design architecture and experimental validation. \u003cem\u003eJournal of Manufacturing Technology Management\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e(4), 507\u0026ndash;534. https://doi.org/10.1108/JMTM-02-2022-0092\u003c/li\u003e\n\u003cli\u003eCho, Y. S., \u0026amp; Linderman, K. (2020). Resource-Based Product and Process Innovation Model: Theory Development and Empirical Validation. \u003cem\u003eSustainability\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(3), 913. https://doi.org/10.3390/su12030913\u003c/li\u003e\n\u003cli\u003eChristensen, B. T., \u0026amp; Ball, L. J. (2016). Dimensions of creative evaluation: Distinct design and reasoning strategies for aesthetic, functional and originality judgments. \u003cem\u003eDesign Studies\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e, 116\u0026ndash;136. https://doi.org/10.1016/j.destud.2015.12.005\u003c/li\u003e\n\u003cli\u003eCrespo-Pereira, V., \u0026amp; Miranda-Galbe, J. (2025). Governments, Users, and Virtual Worlds: Institutional Strategies in the Age of Big Data and IA. \u003cem\u003eSocial Sciences\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(12), 679. https://doi.org/10.3390/socsci14120679\u003c/li\u003e\n\u003cli\u003eDeliu, D., \u0026amp; Olariu, A. (2024). The Role of Artificial Intelligence and Big Data Analytics in Shaping the Future of Professions in Industry 6.0: Perspectives from an Emerging Market. \u003cem\u003eElectronics\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(24), 4983. https://doi.org/10.3390/electronics13244983\u003c/li\u003e\n\u003cli\u003eDyllick, T. (2015). Responsible management education for a sustainable world. \u003cem\u003eJournal of Management Development\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e(1), 16\u0026ndash;33. https://doi.org/10.1108/JMD-02-2013-0022\u003c/li\u003e\n\u003cli\u003eEang, C., \u0026amp; Lee, S. (2024). Predictive Maintenance and Fault Detection for Motor Drive Control Systems in Industrial Robots Using CNN-RNN-Based Observers. \u003cem\u003eSensors\u003c/em\u003e, \u003cem\u003e25\u003c/em\u003e(1), 25. https://doi.org/10.3390/s25010025\u003c/li\u003e\n\u003cli\u003eElgazzar, S., Tipi, N., \u0026amp; Jones, G. (2019). Key characteristics for designing a supply chain performance measurement system. \u003cem\u003eInternational Journal of Productivity and Performance Management\u003c/em\u003e, \u003cem\u003e68\u003c/em\u003e(2), 296\u0026ndash;318. https://doi.org/10.1108/IJPPM-04-2018-0147\u003c/li\u003e\n\u003cli\u003eFajardo-Ramos, D. C., Chiappe, A., \u0026amp; Mella-Norambuena, J. (2025). Human-in-the-loop assessment with AI: implications for teacher education in Ibero-American universities. \u003cem\u003eFrontiers in Education\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e. https://doi.org/10.3389/feduc.2025.1710992\u003c/li\u003e\n\u003cli\u003eGhansah, F., \u0026amp; Edwards, D. (2024). Digital Technologies for Quality Assurance in the Construction Industry: Current Trend and Future Research Directions towards Industry 4.0. \u003cem\u003eBuildings\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(3), 844. https://doi.org/10.3390/buildings14030844\u003c/li\u003e\n\u003cli\u003eGonzalez, C., Fakhari, P., \u0026amp; Busemeyer, J. (2017). Dynamic Decision Making: Learning Processes and New Research Directions. \u003cem\u003eHuman Factors: The Journal of the Human Factors and Ergonomics Society\u003c/em\u003e, \u003cem\u003e59\u003c/em\u003e(5), 713\u0026ndash;721. https://doi.org/10.1177/0018720817710347\u003c/li\u003e\n\u003cli\u003eHagos, K. N., \u0026amp; Gesese, A. A. (2023). Process evaluation of postpartum contraceptive service delivery in Ayder Comprehensive Specialized Hospital Mekelle, Tigray, Ethiopia in 2020. \u003cem\u003eBMC Health Services Research\u003c/em\u003e, \u003cem\u003e23\u003c/em\u003e(1). https://doi.org/10.1186/s12913-023-09467-8\u003c/li\u003e\n\u003cli\u003eHareth, A., Mohammad, I., Supervisor, R., \u0026amp; Morrar, R. (2025). \u003cem\u003eTHE IMPACT OF TOTAL QUALITY MANAGEMENT ON PROJECT PERFORMANCE: THE MEDIATING ROLE OF ORGANIZATIONAL WILLINGNESS FOR ARTIFICIAL INTELLIGENCE IN THE AEC SECTOR IN PALESTINE\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eHilfert-R\u0026uuml;ppell, D., Meier, M., Horn, D., \u0026amp; H\u0026ouml;ner, K. (2021). Professional Knowledge and Self-Efficacy Expectations of Pre-Service Teachers Regarding Scientific Reasoning and Diagnostics. \u003cem\u003eEducation Sciences\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(10), 629. https://doi.org/10.3390/educsci11100629\u003c/li\u003e\n\u003cli\u003eHutson, J. (n.d.). \u003cem\u003eScaffolded Integration: Aligning AI Literacy with Authentic Scaffolded Integration: Aligning AI Literacy with Authentic Assessment through a Revised Taxonomy in Education Assessment through a Revised Taxonomy in Education\u003c/em\u003e. Retrieved https://digitalcommons.lindenwood.edu/faculty-research-papers\u003c/li\u003e\n\u003cli\u003eIgwe, P. A., Okolie, U. C., \u0026amp; Nwokoro, C. V. (2021). Towards a responsible entrepreneurship education and the future of the workforce. \u003cem\u003eThe International Journal of Management Education\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(1), 100300. https://doi.org/10.1016/j.ijme.2019.05.001\u003c/li\u003e\n\u003cli\u003eJaiswal, R., Gupta, S., \u0026amp; Tiwari, A. K. (2023). Dissecting the compensation conundrum: a machine learning-based prognostication of key determinants in a complex labor market. \u003cem\u003eManagement Decision\u003c/em\u003e, \u003cem\u003e61\u003c/em\u003e(8), 2322\u0026ndash;2353. https://doi.org/10.1108/MD-07-2022-0976\u003c/li\u003e\n\u003cli\u003eJalilzadeh, K., \u0026Ccedil;elik, F., \u0026amp; Duran, V. (2025). Student Perspectives on Teacher Interpersonal Behaviours in \u0026lt;scp\u0026gt;AI\u0026lt;/scp\u0026gt; ‐Mediated Higher Education: A Q Methodology Study. \u003cem\u003eEuropean Journal of Education\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(4). https://doi.org/10.1111/ejed.70293\u003c/li\u003e\n\u003cli\u003eKalach, F. El, Farahani, M., Wuest, T., \u0026amp; Harik, R. (2025). Real-time defect detection and classification in robotic assembly lines: A machine learning framework. \u003cem\u003eRobotics and Computer-Integrated Manufacturing\u003c/em\u003e, \u003cem\u003e95\u003c/em\u003e, 103011. https://doi.org/10.1016/j.rcim.2025.103011\u003c/li\u003e\n\u003cli\u003eKurent, B., \u0026amp; Avsec, S. (2025). Systems Thinking in the Role of Fostering Technological and Engineering Literacy. \u003cem\u003eSystems\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(1), 5. https://doi.org/10.3390/systems14010005\u003c/li\u003e\n\u003cli\u003eLeitgeb, T., \u0026amp; Leitgeb, M. (2025). An empirical mixed-methods evaluation of AI-based chatbots for teacher professional development in Austrian higher education. \u003cem\u003eArtificial Intelligence in Education\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e(1), 114\u0026ndash;143. https://doi.org/10.1108/AIIE-01-2025-0013\u003c/li\u003e\n\u003cli\u003eLi, C., Solhi, M., \u0026amp; Wang, Y. (2025). The Role of Chinese \u0026lt;scp\u0026gt;EFL\u0026lt;/scp\u0026gt; Teachers\u0026rsquo; Immediacy and Rapport in Students\u0026rsquo; Engagement in the Context of \u0026lt;scp\u0026gt;AI\u0026lt;/scp\u0026gt; ‐Mediated Education. \u003cem\u003eEuropean Journal of Education\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(4). https://doi.org/10.1111/ejed.70289\u003c/li\u003e\n\u003cli\u003eLi, T., \u0026amp; Zhan, Z. (2022). A Systematic Review on Design Thinking Integrated Learning in K-12 Education. \u003cem\u003eApplied Sciences\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(16), 8077. https://doi.org/10.3390/app12168077\u003c/li\u003e\n\u003cli\u003eMarlowe, T. J., Ku, C. S., Laracy, J. R., Kirova, V. D., \u0026amp; Herbert, K. G. (2026). A Systemic View of a Software Engineering Education Curriculum: Requirements and Guidelines in the Era of Generative AI. \u003cem\u003eJournal of Integrated Design and Process Science\u003c/em\u003e. https://doi.org/10.1177/10920617251405471\u003c/li\u003e\n\u003cli\u003eMaryadi, M., Utami, H. N., Prasetya, A., \u0026amp; Hutahayan, B. (2026). Advancing the understanding of employee performance: a recent systematic literature review using PRISMA. \u003cem\u003eCogent Business \u0026amp; Management\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(1). https://doi.org/10.1080/23311975.2025.2612412\u003c/li\u003e\n\u003cli\u003eMashetty, S., \u0026amp; Valiki, S. (2025). AI-Powered Data Engineering for Intelligent Retail Stock Management. \u003cem\u003eInternational Journal of Research and Applied Innovations\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e(6). https://doi.org/10.15662/IJRAI.2025.0806032\u003c/li\u003e\n\u003cli\u003eMcLachlan, K., \u0026amp; Tippett, N. (2024). Kickstarting creative collaboration: placing authentic feedback at the heart of online digital media education. \u003cem\u003eAssessment \u0026amp; Evaluation in Higher Education\u003c/em\u003e, \u003cem\u003e49\u003c/em\u003e(2), 246\u0026ndash;261. https://doi.org/10.1080/02602938.2023.2209295\u003c/li\u003e\n\u003cli\u003eMergel, I., Kleibrink, A., \u0026amp; S\u0026ouml;rvik, J. (2018). Open data outcomes: U.S. cities between product and process innovation. \u003cem\u003eGovernment Information Quarterly\u003c/em\u003e, \u003cem\u003e35\u003c/em\u003e(4), 622\u0026ndash;632. https://doi.org/10.1016/j.giq.2018.09.004\u003c/li\u003e\n\u003cli\u003eMourtzis, D., Panopoulos, N., \u0026amp; Angelopoulos, J. (2023). A hybrid teaching factory model towards personalized education 4.0. \u003cem\u003eInternational Journal of Computer Integrated Manufacturing\u003c/em\u003e, \u003cem\u003e36\u003c/em\u003e(12), 1739\u0026ndash;1759. https://doi.org/10.1080/0951192X.2022.2145025\u003c/li\u003e\n\u003cli\u003eNagubathula, V. (2025). AI and Human-AI Collaboration in Enterprise Integration and Document Automation. \u003cem\u003eInternational Journal on Science and Technology\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(1). https://doi.org/10.71097/IJSAT.v16.i1.2317\u003c/li\u003e\n\u003cli\u003eNuttah, M. M., Algabroun, H., Linhares, C., \u0026amp; H\u0026aring;kansson, L. (2025). Creative Destruction and Technological Paradigms in Manufacturing: A Large-Scale Review and Framework for Technology Portfolio Assessment. \u003cem\u003eIEEE Transactions on Engineering Management\u003c/em\u003e, \u003cem\u003e72\u003c/em\u003e, 3397\u0026ndash;3418. https://doi.org/10.1109/TEM.2025.3592031\u003c/li\u003e\n\u003cli\u003eOtey, J., Company, P., Contero, M., \u0026amp; Camba, J. D. (2019). Assessment of parametric assembly models based on CAD quality dimensions. \u003cem\u003eComputer-Aided Design and Applications\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(4), 628\u0026ndash;653. https://doi.org/10.14733/cadaps.2019.628-653\u003c/li\u003e\n\u003cli\u003eRadlbauer, E., Moser, T., \u0026amp; Wagner, M. (2025). Designing a System Architecture for Dynamic Data Collection as a Foundation for Knowledge Modeling in Industry. \u003cem\u003eApplied Sciences\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(9), 5081. https://doi.org/10.3390/app15095081\u003c/li\u003e\n\u003cli\u003eRetelny, D., Bernstein, M. S., \u0026amp; Valentine, M. A. (2017). No workflow can ever be enough: How crowdsourcing workflows constrain complex work. \u003cem\u003eProceedings of the ACM on Human-Computer Interaction\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e(CSCW). https://doi.org/10.1145/3134724\u003c/li\u003e\n\u003cli\u003eRousseau, D. M. (2012). Designing a Better Business School: Channelling Herbert Simon, Addressing the Critics, and Developing Actionable Knowledge for Professionalizing Managers. \u003cem\u003eJournal of Management Studies\u003c/em\u003e, \u003cem\u003e49\u003c/em\u003e(3), 600\u0026ndash;618. https://doi.org/10.1111/j.1467-6486.2011.01041.x\u003c/li\u003e\n\u003cli\u003eRusso, D. (2024). Navigating the Complexity of Generative AI Adoption in Software Engineering. \u003cem\u003eACM Transactions on Software Engineering and Methodology\u003c/em\u003e, \u003cem\u003e33\u003c/em\u003e(5), 1\u0026ndash;50. https://doi.org/10.1145/3652154\u003c/li\u003e\n\u003cli\u003eSbhatu, D. B., Abdullah, N., \u0026amp; Olohunfunmi, A. I. (2026). The Nexus of Research, Policy, and Practice: Narrative Literature Review of AI Adoption in Future-Proof STEM Curriculum in Higher Education. \u003cem\u003eCanadian Journal of Science, Mathematics and Technology Education\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(1), 5. https://doi.org/10.1007/s42330-026-00427-y\u003c/li\u003e\n\u003cli\u003eSharif, H., Atif, A., \u0026amp; Nagra, A. A. (2025). Deepfake-Style AI Tutors in Higher Education: A Mixed-Methods Review and Governance Framework for Sustainable Digital Education. \u003cem\u003eSustainability\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(21), 9793. https://doi.org/10.3390/su17219793\u003c/li\u003e\n\u003cli\u003eSimpson, T. W. (n.d.). \u003cem\u003eExperiences with a Hands-on Activity to Contrast Craft Production and Mass Production in the Classroom*\u003c/em\u003e. Retrieved http://www.orie.cornell.edu/~jackson.\u003c/li\u003e\n\u003cli\u003eSoltanmohammadlou, N., Hon, C. K. H., Drogemuller, R., Motamedisedeh, O., \u0026amp; Rahimian, F. (2026). Modelling risk factors in earthmoving equipment operations on Australian construction sites: a fuzzy DEMATEL approach. \u003cem\u003eEngineering, Construction and Architectural Management\u003c/em\u003e, \u003cem\u003e33\u003c/em\u003e(3), 2118\u0026ndash;2153. https://doi.org/10.1108/ECAM-02-2025-0240\u003c/li\u003e\n\u003cli\u003eSun, X., \u0026amp; Tian, H. (2026). AI-Enhanced Skill Assessment in Higher Vocational Education: A Systematic Review and Meta-Analysis. \u003cem\u003eInformatics\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(2), 20. https://doi.org/10.3390/informatics13020020\u003c/li\u003e\n\u003cli\u003eVamsi Krishna Jasti, N., \u0026amp; Kodali, R. (2014). A literature review of empirical research methodology in lean manufacturing. \u003cem\u003eInternational Journal of Operations \u0026amp; Production Management\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e(8), 1080\u0026ndash;1122. https://doi.org/10.1108/IJOPM-04-2012-0169\u003c/li\u003e\n\u003cli\u003eWahjusaputri, S., \u0026amp; Bunyamin, B. (2022). Development of teaching factory competency-based for vocational secondary education in Central Java, Indonesia. \u003cem\u003eInternational Journal of Evaluation and Research in Education\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(1), 353\u0026ndash;360. https://doi.org/10.11591/ijere.v11i1.21709\u003c/li\u003e\n\u003cli\u003eWeeks, K. W., Coben, D., O\u0026rsquo;Neill, D., Jones, A., Weeks, A., Brown, M., \u0026amp; Pontin, D. (2019). Developing and integrating nursing competence through authentic technology-enhanced clinical simulation education: Pedagogies for reconceptualising the theory-practice gap. \u003cem\u003eNurse Education in Practice\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e, 29\u0026ndash;38. https://doi.org/10.1016/j.nepr.2019.04.010\u003c/li\u003e\n\u003cli\u003eWirtz, B. W., \u0026amp; Daiser, P. (2018). A meta-analysis of empirical e-government research and its future research implications. \u003cem\u003eInternational Review of Administrative Sciences\u003c/em\u003e, \u003cem\u003e84\u003c/em\u003e(1), 144\u0026ndash;163. https://doi.org/10.1177/0020852315599047\u003c/li\u003e\n\u003cli\u003eWu, Z., Liang, C., Shen, Z., \u0026amp; Zhou, W. (2025). Quality control in remanufacturing process: theoretical framework, control strategy and application review. \u003cem\u003eThe International Journal of Advanced Manufacturing Technology\u003c/em\u003e, \u003cem\u003e140\u003c/em\u003e(1\u0026ndash;2), 517\u0026ndash;543. https://doi.org/10.1007/s00170-025-16343-8\u003c/li\u003e\n\u003cli\u003eXiao, F., Wang, C., \u0026amp; Jiang, J. (2026). Active Learning in University Physics for Sustainable Higher Education: Effective Components, Mechanisms, and SDG-Aligned Competency Pathways\u0026mdash;A Multidimensional Review. \u003cem\u003eSustainability\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(6), 2791. https://doi.org/10.3390/su18062791\u003c/li\u003e\n\u003cli\u003eZhang, H., Sun, X., Mynors, D., \u0026amp; Guo, C. (2025). Combining Virtual Reality with the Physical Model Factory: A Practice Course Designed for Manufacturing Process Education. \u003cem\u003eProcesses\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(9), 2946. https://doi.org/10.3390/pr13092946\u003c/li\u003e\n\u003cli\u003eZhang, P., Ma, S.-G., Zhao, Y.-N., \u0026amp; Cao, X.-Y. (2023). Analyzing Core Competencies and Correlation Paths of Emerging Engineering Talent in the Construction Industry\u0026mdash;An Integrated ISM\u0026ndash;MICMAC Approach. \u003cem\u003eSustainability\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(22), 16011. https://doi.org/10.3390/su152216011\u003c/li\u003e\n\u003cli\u003eZhang, X., Ma, Y., Jiang, Z., Chandrasekaran, S., Wang, Y., \u0026amp; Fonkoua Fofou, R. (2021). Application of Design-Based Learning and Outcome-Based Education in Basic Industrial Engineering Teaching: A New Teaching Method. \u003cem\u003eSustainability\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(5), 2632. https://doi.org/10.3390/su13052632\u003c/li\u003e\n\u003cli\u003eZhang, Y., \u0026amp; Dong, C. (2024). Exploring the Digital Transformation of Generative AI-Assisted Foreign Language Education: A Socio-Technical Systems Perspective Based on Mixed-Methods. \u003cem\u003eSystems\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(11), 462. https://doi.org/10.3390/systems12110462\u003c/li\u003e\n\u003cli\u003eZhou, J. (2025). Research on the Construction of Core Competency Framework and Training Quality Evaluation System for Industrial Design Talents in the Intelligent Era. \u003cem\u003eInternational Journal of Educational Development\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), 21. https://doi.org/10.63313/IJED.9031\u003c/li\u003e\n\u003cli\u003eZong, Z., \u0026amp; Guan, Y. (2024). AI-Driven Intelligent Data Analytics and Predictive Analysis in Industry 4.0: Transforming Knowledge, Innovation, and Efficiency. \u003cem\u003eJournal of the Knowledge Economy\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(1), 864\u0026ndash;903. https://doi.org/10.1007/s13132-024-02001-z\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Bengkulu","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Teaching factory, AI-driven analytics, quality assurance, real-time skills validation, process-oriented assessment, vocational education, Industry 4.0, systematic literature review","lastPublishedDoi":"10.21203/rs.3.rs-9528562/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9528562/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis systematic literature review aims to critically synthesize existing research on the integration of Artificial Intelligence (AI)-driven analytics into the quality assurance (QA) frameworks of teaching factories in vocational education. Specifically, the study seeks to shift the paradigm from conventional product-oriented QA (final output inspection) toward a process-oriented, real-time skills validation model that captures student competencies during active production workflows. This study employs a Systematic Literature Review (SLR) design following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. A comprehensive search was conducted across four major databases: Scopus, Web of Science, IEEE Xplore, and ERIC, covering peer-reviewed articles published between 2015 and 2025. Keywords included \u0026ldquo;teaching factory, \u0026ldquo; \u0026ldquo;AI analytics, \u0026ldquo; \u0026ldquo;quality assurance, \u0026ldquo; \u0026ldquo;real-time assessment, \u0026ldquo; and \u0026ldquo;skills validation. \u0026ldquo; After screening 487 initial records, 52 articles met the inclusion criteria for thematic synthesis. Data were extracted and analyzed using content analysis to identify patterns, gaps, and emerging frameworks related to process-based QA. The findings reveal three major themes: (1) Traditional teaching factory QA remains heavily product-centric, focusing on final product conformity rather than competency development; (2) AI-driven analytics (e.g., computer vision, sensor data, learning analytics) enable continuous, non-intrusive monitoring of student actions, decision-making, and error correction patterns; (3) Real-time skills validation is technically feasible but underutilized due to gaps in pedagogical integration, instructor AI-literacy, and data privacy protocols. Key success factors include adaptive feedback loops, dashboards for formative assessment, and alignment between production KPIs (Key Performance Indicators) and competency rubrics. This review is the first to explicitly conceptualize a shift from product-centric to process-centric QA in teaching factories using AI. It introduces the \u0026ldquo;Dynamic Process Validation Model \u0026ldquo; where AI analytics transform every production step into a measurable learning event, rather than merely certifying final outputs. This contrasts sharply with prior studies that focus exclusively on product quality or separate educational assessment. Vocational school administrators and teaching factory managers can use these findings to design AI-enhanced QA systems that provide real-time, actionable feedback to students, instructors, and industry partners. Implementation guidelines include selecting non-intrusive sensors, developing real-time dashboards for formative assessment, and training instructors to interpret AI-generated process data. The results also inform the creation of competency-based digital transcripts that document process skills (e.g., problem-solving under pressure, adherence to safety protocols) alongside final product grades. This study contributes to the body of knowledge in vocational education quality assurance by: (1) providing a synthesized framework for integrating AI-driven process analytics into teaching factory QA; (2) identifying critical success factors and barriers specific to real-time skills validation; and (3) offering a theoretical foundation for future empirical research on AI-mediated competency assessment. It also bridges the gap between educational quality assurance and industrial production quality management, fostering a more authentic and responsive vocational training environment aligned with Industry 4.0 demands.\u003c/p\u003e","manuscriptTitle":"From Product to Process: Integrating AI-Driven Analytics into Teaching Factory Quality Assurance for Real-Time Skills Validation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-28 08:27:00","doi":"10.21203/rs.3.rs-9528562/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"7c9bb2da-7a7a-46bf-9c5d-bd227ad1ef60","owner":[],"postedDate":"April 28th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":67014643,"name":"Special Education"}],"tags":[],"updatedAt":"2026-04-28T08:27:01+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-28 08:27:00","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9528562","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9528562","identity":"rs-9528562","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}