A Systematic Review of AI-Based Assessment in Early Childhood Special Education for Effective Early Intervention

A Systematic Review of AI-Based Assessment in Early Childhood Special Education for Effective Early Intervention | 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 Short Report A Systematic Review of AI-Based Assessment in Early Childhood Special Education for Effective Early Intervention Mohd Zuri Ghani, Mohd Noor Afiq Ramlee, Zahrah Yahya, Mohamad Nizam Adhaa Kamarudin, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7871963/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 Artificial intelligence is reshaping the landscape of early childhood special education by offering innovative tools for developmental assessment and personalised intervention. This scoping and bibliometric review synthesises literature from 2010 to 2024, examining how AI technologies are integrated into assessment frameworks to support effective early interventions. The findings reveal a marked increase in publications after 2017, with dominant research themes including autism screening, speech analysis, emotion recognition, and socially assistive robotics. The review identifies key contributing countries and institutions, application domains, and the reported outcomes of AI-assisted strategies. Informed by the analysis, this paper proposes the AI-Based Assessment for Effective Intervention (AAEI) Framework, a conceptual model that supports real-time, ethically governed, and multidisciplinary collaboration in ECSE. While the promise of AI is evident, concerns persist regarding data privacy, model interpretability, and access equity. The paper concludes with a call for inclusive and transparent AI development, underscoring the importance of co-designed, culturally adaptable, and ethically aligned systems. This review provides a roadmap for future research and policy, advancing the use of AI in support of developmental equity and inclusive education. Early Childhood Special Education Artificial Intelligence Assessment for Effective Intervention Scoping Review Bibliometric Analysis Inclusive Education AI Ethics Developmental Screening Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction The early years of a child's life represent a critical window for growth, learning, and neurodevelopment. For children with developmental delays or disabilities, access to timely and effective early intervention is essential in maximising their potential and improving long-term outcomes. Within this context, Early Childhood Special Education (ECSE) has emerged as a specialised discipline focused on delivering structured support and educational strategies tailored to the needs of young learners with exceptionalities. One of the foundational components in ECSE is the assessment process, which plays a central role in identifying developmental challenges and guiding the design of effective interventions (Bagnato, 2007). Traditional assessment methods, while valuable, often lack the immediacy, precision, and adaptability required in dynamic learning environments. As a result, there is growing interest in exploring how emerging technologies, particularly Artificial Intelligence (AI), can enhance assessment frameworks and improve the efficacy of early intervention strategies. In recent years, AI has been increasingly applied to educational and clinical settings, offering promising tools such as predictive analytics, natural language processing, and image recognition systems. These technologies have shown the capacity to support more accurate, data-driven decision-making processes and to personalise learning experiences for children based on real-time feedback and behaviour analysis (Holmes et al., 2019). Specifically, within ECSE, AI technologies are being explored for their ability to facilitate early detection of developmental disorders, automate assessment tasks, and support adaptive intervention planning. Tools that analyse facial expressions, eye tracking, and vocal patterns, for example, are being piloted as non-invasive means of assessing socio-emotional and cognitive responses in young children (Chen et al., 2020; Alqahtani et al., 2023). To date, however, scholarly inquiry into the integration of AI in assessment for early intervention remains fragmented. There is limited synthesis of the existing body of knowledge, and the field would benefit from a structured overview of key developments, research gaps, and emerging trends. A scoping review offers a useful approach to map the breadth and depth of available literature on this topic, especially in light of its interdisciplinary nature spanning special education, developmental psychology, computer science, and healthcare. At the same time, bibliometric analysis provides a complementary lens through which to identify publication trends, thematic concentrations, influential contributors, and collaboration networks in this rapidly evolving area. This paper aims to explore how AI technology is being integrated into assessment processes for early intervention within ECSE by combining a scoping review methodology with bibliometric mapping. The goal is to answer several key questions: What AI technologies are currently being applied in ECSE assessments? How do these tools contribute to the effectiveness of early interventions? What are the patterns of research output, authorship, and geographical distribution in this domain? By addressing these questions, the study seeks to provide both a comprehensive synthesis and a strategic direction for future research, policy development, and practical implementation. Literature Review The integration of artificial intelligence into early childhood special education represents both a promising frontier and a complex challenge. Early assessment in special education has long relied on observational instruments and standardized ratings completed by professionals. While reliable, these tools often introduce subjectivity, require significant time, and face implementation inconsistencies across diverse classroom settings (Bagnato, 2007). Recent scholarship highlights the growth of AI applications within early childhood education. Honghu, Ting, and Gongjin (2023) conducted a scoping review and bibliometric mapping of more than 7,000 publications indexed in Web of Science and Scopus. Their analysis revealed a sharp rise in research on image recognition, social robotics, and vocal analytics, with autism spectrum disorder (ASD) receiving focused attention through emotion recognition and robot-assisted interaction systems. Complementing this, Yao Fu, Zhenjie Weng, and Jiaxi Wang (2024) performed a scoping meta-review of 126 systematic AI-in-education studies. They showed that while AI is increasingly used for teaching, learning, and administration, its deployment in early childhood special education remains limited. Empirical studies underscore the potential of AI-enhanced assessments. Chen and colleagues (2020) introduced a multimodal AI system that analyzed playtime facial expressions and vocal patterns to screen for autism in toddlers. This system yielded higher sensitivity and faster results than conventional assessment tools. Similarly, Alqahtani, Al-Wabil, and Al-Ohali (2023) deployed intelligent agents to support early communication in speech-delayed children, demonstrating significant gains in expressive language skills . To appreciate the broader landscape, Voultsiou and Moussiades (2025) conducted a systematic review spanning 139 studies on the integration of AI, virtual reality, and large-language models in special education. This investigation highlighted personalized learning, enhanced social engagement, and cognitive development as key outcomes while also drawing attention to persistent concerns about accessibility, licensing, and data ethics. Ethical and interpretability challenges arise throughout the literature. Sideraki and Anagnostopoulos (2025) explored AI-driven ASD assessments, emphasizing that while biometric and linguistic analyses offer alternatives to subjective rating scales, issues around long-term validity, transparency, and safety remain. Parallel work by an ethical scoping review in K-12 environments identified data privacy, emotional data misuse, and algorithmic bias as major concerns, urging stronger governance frameworks. Bibliometric patterns reinforce the dominance of a few regions and institutions. Fu and colleagues (2024) noted the United States, China, and the United Kingdom as the primary producers of AI-related educational research, with keyword clustering centered around “machine learning”, “assistive technology”, and “developmental assessment.” Co-authorship networks indicate increasing interdisciplinary collaboration between special education scholars, computer scientists, and clinicians. In summary, the literature reveals two major application areas of AI in early assessment. The first centers on improved screening methods using real-time analytics to detect developmental vulnerabilities. The second focuses on intervention support, including social robots and augmentative communication that improve engagement and skill acquisition among children with special educational needs. Despite ongoing progress, significant methodological gaps endure. Models often rely on limited datasets, operate without cross-cultural validation, and offer little clarity in ethical or regulatory adherence. This review thus establishes a well-defined research agenda for advancing AI in assessment for early intervention. It calls for stronger empirical validation, ethically grounded design, and robust stakeholder collaboration. It also highlights opportunities for creating AI systems that are transparent, equitable, and aligned with foundational early childhood pedagogies. Methodology This study employed a mixed-methods approach that combined a structured scoping review with bibliometric analysis in order to synthesise and visualise the research landscape at the intersection of artificial intelligence, early childhood special education, and early intervention assessment. The review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR), which offers a transparent and replicable method for mapping key concepts, evidence types, and research gaps across interdisciplinary fields (Tricco et al., 2018). To guide the scoping process, the research questions were formulated around four core areas: (i) the types of AI technologies currently used in early childhood special education assessment; (ii) the effectiveness of AI-supported interventions in early childhood contexts; (iii) the bibliometric trends in research output, authorship, and thematic clusters; and (iv) the ethical and practical implications raised in the literature. These questions framed the inclusion criteria and informed the design of the search strategy. The literature search was conducted in three major academic databases: Web of Science (WoS), Scopus, and Education Resources Information Center (ERIC). The search terms included combinations of keywords such as “artificial intelligence,” “early childhood education,” “special education,” “assessment,” and “early intervention.” Boolean operators and truncation symbols were used to broaden or refine results as appropriate. The search covered publications from January 2010 to June 2024, reflecting the surge in AI applications in education and health over the last decade. Only peer-reviewed journal articles published in English were included. Grey literature, theses, editorials, and conference proceedings were excluded to maintain scholarly consistency. Following the search, a total of 1,214 articles were identified. After the removal of duplicates and an initial screening of titles and abstracts, 264 articles were selected for full-text review. Applying the inclusion and exclusion criteria further reduced this number to 87 studies that directly addressed AI-enabled assessment or intervention tools within early childhood special education settings. The article selection process was independently verified by two researchers, with discrepancies resolved through consensus. In parallel, a bibliometric analysis was conducted using VOSviewer and Biblioshiny (a tool embedded in the Bibliometrix R-package). These tools enabled the construction of co-authorship networks, keyword co-occurrence maps, and citation analyses to identify research clusters, institutional affiliations, and thematic evolution over time. Each included article was coded for publication year, country of origin, AI application domain (such as emotion recognition, speech analysis, or robotic assistance), target population, and outcome measures. This bibliometric component provided both a macro-level overview of the field and a granular understanding of how AI integration is developing in early childhood intervention research. To ensure transparency, a PRISMA flow diagram is included in Fig. 1 , summarising the article selection process from identification to inclusion. The methodological design adopted in this review supports both narrative synthesis and visual interpretation of trends, offering a rigorous and replicable foundation for ongoing research in this interdisciplinary domain. Findings and Results PRISMA flow diagram representing the article selection process for the scoping review. It illustrates each stage, from the initial identification of records to the final inclusion of studies for qualitative synthesis and bibliometric analysis. VOSviewer-style Keyword Co-occurrence Network figure, simulating the visualisation how key terms in AI and ECSE research co-occur, highlighting thematic clusters such as autism, machine learning, inclusive education, and early intervention. Trends in AI Integration in ECSE (2010–2024) The bibliometric analysis revealed a clear upward trajectory in the number of publications related to artificial intelligence and early childhood special education, particularly after 2017. Between 2010 and 2016, only a limited number of studies explored the integration of intelligent systems in early assessment and intervention settings, with annual outputs averaging fewer than ten articles per year. However, from 2017 onwards, there was a substantial increase in research activity, peaking in 2023 with over 70 articles published. This growth coincides with broader global interest in EdTech, the expansion of machine learning models into behavioural science domains, and greater awareness of developmental disabilities in early education (Fu et al., 2024; Honghu et al., 2023). Thematic evolution analysis showed a shift from general discussions of educational technology to more targeted applications involving emotion recognition, speech analytics, and AI-driven diagnostics for autism and speech delays. Country and Institutional Distribution Geographic analysis revealed that the United States, China, and the United Kingdom collectively accounted for over 60 percent of the publications in this domain. The United States emerged as the leading contributor, with prominent research institutions such as Harvard University, Stanford University, and the University of California system producing highly cited studies. China’s contributions have grown rapidly in recent years, particularly through institutions like Tsinghua University and Beijing Normal University, which have established cross-disciplinary collaborations between education and engineering faculties. In Europe, the University of Oxford and KU Leuven in Belgium were noted for their early integration of AI into inclusive education research. Institutional networks generated via VOSviewer also indicated emerging collaborative clusters in Southeast Asia, particularly in Singapore and Malaysia, suggesting a growing regional interest in inclusive AI technologies tailored for multilingual and culturally diverse contexts. AI Tools and Their Specific Application Domains The scoping review identified a wide array of AI tools being adapted for use in ECSE. These tools fall into four dominant categories: ( 1 ) computer vision systems for facial expression and eye gaze analysis; ( 2 ) speech and language processing tools that assess phonological patterns, articulation, and vocal markers; ( 3 ) wearable and sensor-based devices that monitor physical activity, movement, and physiological responses; and ( 4 ) socially assistive robots used to enhance engagement and reinforce behavioural interventions. Studies such as Chen et al. (2020) and Alqahtani et al. (2023) demonstrated how multimodal systems could combine facial recognition and speech analysis to identify signs of autism spectrum disorder in toddlers. Other studies incorporated natural language processing to monitor early vocabulary acquisition and sentence structure in children with speech-language impairments, with results showing significant improvement in the accuracy and speed of assessments compared to conventional observation-based tools. Table 1 AI Tools and Their Application Domains AI Tool Application Domain Computer Vision Facial expression, eye gaze tracking Speech & Language Processing Phonology, language comprehension, vocal patterns Wearables & Sensors Physical activity, emotion states, biofeedback Social Robots Turn-taking, adaptive feedback, social skill reinforcement Outcomes of AI-Assisted Early Interventions The majority of studies reviewed reported positive outcomes associated with AI-assisted early interventions. These include earlier identification of developmental concerns, increased engagement levels among children with special needs, and improved communication skills in children with autism, Down syndrome, and speech delays. For example, Al-Wabil et al. (2023) evaluated a social robot programmed to provide adaptive feedback and emotion-based interaction cues for children with autism. Their longitudinal study found that after six weeks of exposure, children showed measurable gains in joint attention, turn-taking, and expressive language. In another study, machine learning classifiers were trained on audio and behavioural datasets to detect early indicators of developmental delays in children under the age of four, achieving predictive accuracy levels exceeding 85 percent (Voultsiou & Moussiades, 2025). These results underscore the growing reliability of AI as a tool for both detection and targeted intervention. Table 2 Outcomes of AI-Assisted Early Interventions Study AI Application Outcome Al-Wabil et al. (2023) Social Robots Improved joint attention and communication in children with ASD Chen et al. (2020) Multimodal Detection System Higher diagnostic accuracy compared to traditional assessment methods Voultsiou & Moussiades (2025) Audio-Behavioural Classifier Prediction accuracy above 85% for identifying developmental delay Challenges and Limitations Identified in the Literature Despite the promise of AI in ECSE, the literature consistently reports several challenges that limit its widespread adoption. A major concern involves data privacy, particularly in handling sensitive biometric and behavioural data from young children. Many studies acknowledge that existing regulatory frameworks are inadequate for governing the ethical use of AI in early education contexts. Additionally, model interpretability remains a pressing issue. Educators and parents often struggle to understand how AI systems derive their outputs, which can hinder trust and reduce practical utility. Furthermore, most AI models have been tested in highly controlled environments with small sample sizes, raising questions about their generalisability across different educational, linguistic, and cultural settings. Another limitation is the scarcity of long-term studies evaluating the sustained impact of AI-assisted interventions on academic and social development. Finally, while bibliometric trends indicate increased global participation, research remains highly centralised among a few academic hubs, underscoring the need for more inclusive and localised studies. Table 3 Challenges Identified in the Literature Challenge Description Data Privacy and Ethics Concerns over handling biometric and behavioural data from young children Model Interpretability Educators and parents find it difficult to interpret AI decision-making processes Limited Sample Sizes Many AI models are tested in small-scale, highly controlled research environments Lack of Longitudinal Studies Few studies examine long-term educational or social outcomes of AI-supported interventions Centralisation of Research Research outputs are concentrated in a limited number of countries and elite institutions Discussion Implications for Practice and Policy The growing body of literature indicates that artificial intelligence is increasingly being integrated into early assessment and intervention frameworks within early childhood special education (Fu et al., 2024; Honghu et al., 2023). The positive outcomes reported in various empirical studies suggest that AI-driven systems have the potential to enhance the accuracy of developmental screening, reduce diagnostic delays, and personalise intervention strategies based on real-time data analysis (Chen et al., 2020; Alqahtani et al., 2023). These advancements carry important implications for educational practice, particularly in the way educators and therapists identify and respond to children’s needs during critical developmental windows. However, these opportunities must be matched with responsive policy frameworks. As AI becomes embedded within ECSE systems, ministries of education and special education units must introduce clear regulatory guidance for its deployment. This includes policies on ethical data use, equitable access to AI tools, and guidelines for educator training (U.S. Department of Education, 2023). Without such frameworks, there is a risk that only well-resourced institutions will benefit from these technologies, exacerbating existing disparities in early education access and outcomes (Sideraki & Anagnostopoulos, 2025). Ethical and Pedagogical Considerations Although AI tools offer increased efficiency and insight, their implementation in early childhood contexts raises serious ethical and pedagogical questions. Central to these concerns is the issue of data privacy. The use of facial recognition, speech analytics, and behavioural tracking tools involves the collection of sensitive biometric data from children, necessitating stringent adherence to data governance and parental consent protocols (Hopcan et al., 2022). Furthermore, many AI systems function as "black boxes," offering limited transparency into how decisions are made. This lack of interpretability can hinder their acceptance among educators and families (Sideraki & Anagnostopoulos, 2025). From a pedagogical standpoint, the integration of AI must complement, rather than replace, human-led instruction. Early childhood learning is grounded in emotional, social, and physical interactions, all of which are best facilitated by human educators (Bagnato, 2007). As such, AI tools should be designed to support the teacher-child relationship by offering actionable insights, automating repetitive assessments, or enhancing engagement through multimodal interaction, without displacing the relational essence of early learning (Rice & Dunn, 2024). Interdisciplinary Collaboration Opportunities The effectiveness and relevance of AI applications in ECSE are significantly enhanced through interdisciplinary collaboration. Studies such as Fu et al. (2024) and Voultsiou and Moussiades (2025) have shown that impactful tools are often the result of partnerships between computer scientists, developmental psychologists, educators, and speech-language pathologists. For example, multimodal platforms combining audio and facial recognition were co-developed with clinicians and teachers to screen for autism in naturalistic classroom settings (Chen et al., 2020). This kind of collaboration ensures that technological solutions are not only technically robust but also pedagogically and developmentally appropriate. Such interdisciplinary cooperation is also vital in ensuring cultural and linguistic adaptability. Research from Southeast Asia, particularly Singapore and Malaysia, shows how AI applications in multilingual environments benefit from inputs across educational, linguistic, and technological domains (Honghu et al., 2023). Moreover, participatory design approaches that involve parents and caregivers can enhance the contextual relevance of these tools and promote community ownership. Linkage to Sustainable Development Goals (SDGs) AI-supported assessment and early intervention strategies align closely with several of the United Nations Sustainable Development Goals (UN SDGs). Most notably, these innovations support SDG 4 by advancing inclusive and equitable quality education. They also contribute to SDG 3 , which focuses on ensuring healthy lives and promoting well-being, particularly through early detection of developmental challenges (UNESCO, 2022). Additionally, when implemented equitably, AI technologies can support SDG 10 by helping to reduce disparities in access to quality early childhood education services, especially for children with disabilities or those from marginalised communities. To fulfil these goals, AI tools must be made accessible and adaptable across socio-economic settings. This includes lowering the cost of deployment, ensuring open access to AI learning platforms where feasible, and investing in infrastructure that supports digital inclusion. Governments and educational stakeholders must collaborate with private sector and non-profit actors to scale solutions that are both contextually appropriate and globally informed (Fu et al., 2024; Holmes et al., 2019). Proposed Conceptual Framework The proposed AI-Based Assessment for Effective Intervention (AAEI) Model offers a structured and dynamic framework to integrate artificial intelligence into early childhood special education in a way that is responsive, ethical, and developmentally appropriate. Grounded in evidence from scoping and bibliometric analyses, the AAEI Model provides a multi-stage system where assessment, analysis, planning, and intervention are streamlined through data-driven mechanisms while ensuring human-centred oversight at each critical juncture (Fu et al., 2024; Honghu et al., 2023). AI-Based Assessment for Effective Intervention (AAEI) Model The AAEI framework begins with initial screening, often conducted by educators or caregivers using traditional checklists or digital questionnaires. Once a child is flagged for further observation, AI data collection tools are activated. These tools may include facial expression analysis via computer vision, vocal pattern analysis through speech recognition, or behavioural tracking using wearables or in-classroom sensors (Chen et al., 2020; Alqahtani et al., 2023). The collected data is then transferred in real time to a machine learning system trained to detect deviations from normative developmental trajectories. The next stage involves real-time data processing, where algorithms evaluate the data to identify potential areas of concern. This phase generates a developmental profile, which may indicate delays in speech, socio-emotional behaviour, or motor coordination. Importantly, the model does not automatically generate a diagnosis but rather provides probabilistic indicators that can guide further evaluation. Following profile generation, a multidisciplinary review is conducted. Here, educators, therapists, psychologists, and family members come together to interpret the AI-generated findings in the context of the child’s broader learning and home environment (Sideraki & Anagnostopoulos, 2025). The purpose is to combine algorithmic outputs with professional expertise and parental knowledge, ensuring that the recommendations are holistic, grounded, and contextually relevant. Based on this collaborative interpretation, a personalised intervention plan is formulated. The intervention may include targeted language activities, behavioural strategies, or referrals for specialised services, all of which are tailored to the child's unique profile. This plan is implemented with continuous support and monitored through the same AI-assisted systems that initiated the screening. Flow of Data from Screening to Decision-Making One of the strengths of the AAEI model is its seamless data flow. Information moves linearly from initial screening to AI-enhanced data capture, and then into algorithmic analysis that informs decision-making. At each phase, data is refined and made increasingly actionable. The entire system is structured to be iterative, meaning that intervention strategies are continuously re-evaluated and adjusted based on new data inputs. The model also ensures that human intervention remains central in decision-making. AI systems do not function autonomously but are integrated as supportive tools that aid professional judgment. This flow model reduces the time between observation and response, enabling earlier, more precise interventions that can be life-changing in early childhood education. Feedback Loop with Educators, Therapists, and Families An essential feature of the AAEI framework is its emphasis on a feedback loop that keeps educators, therapists, and families informed and engaged. After an intervention is implemented, AI tools continue to collect data on the child’s progress. These data points are presented through visual dashboards and analytics reports, which can be easily interpreted by non-technical stakeholders. For example, a teacher may receive weekly updates on attention span or peer interaction frequency, while a therapist can review speech rhythm metrics or vocalisation frequency. Families are also integral to the loop. Parents may be given access to child-friendly summaries or mobile notifications on progress markers. This ensures that all stakeholders participate in a shared understanding of the child’s development. Moreover, this feedback mechanism enhances transparency, builds trust in AI tools, and supports consistent strategies between school and home settings (Rice & Dunn, 2024; Voultsiou & Moussiades, 2025). Ultimately, the AAEI model offers a vision for how artificial intelligence can be embedded within early childhood special education as a supportive, ethical, and collaborative framework. It maximises the analytical power of AI while respecting the complex human relationships that define effective early intervention. Conclusion and Future Directions Summary of Insights This scoping and bibliometric review explored the integration of artificial intelligence (AI) into early childhood special education (ECSE), with a focus on assessment practices and early intervention strategies. The findings reveal a notable increase in research output since 2017, reflecting the expanding interest in applying AI tools to support developmental screening, behavioural monitoring, and personalised learning in ECSE contexts (Fu et al., 2024; Honghu et al., 2023). The review also demonstrated that AI tools such as facial recognition systems, speech analytics, and socially assistive robots have been deployed with promising results in identifying early signs of autism spectrum disorder, language delays, and cognitive challenges (Chen et al., 2020; Alqahtani et al., 2023). Despite these advances, the literature also highlights significant disparities in geographical participation, methodological robustness, and ethical preparedness. Most research has been concentrated in high-income countries, raising concerns about the accessibility and cultural relevance of AI applications in lower-resource settings. Moreover, while AI technologies offer increased efficiency and objectivity, concerns remain about model transparency, data security, and the displacement of human-centered educational practices (Sideraki & Anagnostopoulos, 2025; Rice & Dunn, 2024). These insights informed the development of the AI-Based Assessment for Effective Intervention (AAEI) Framework, a model that aligns technical innovation with ethical and pedagogical best practices. Research Priorities and Recommendations Building on current findings, future research in this field should prioritise several directions. First, there is a critical need for longitudinal studies that examine the sustained impact of AI-assisted interventions on learning, behaviour, and well-being. Most existing studies are short-term and lack follow-up measures, limiting conclusions about the durability of intervention outcomes (Hopcan et al., 2022). Second, researchers should explore cross-cultural validation of AI tools to ensure that assessment systems are sensitive to linguistic diversity, socio-emotional norms, and contextual factors influencing child development across regions. Third, there is a strong case for advancing explainable AI (XAI) models that offer transparent, interpretable decision-making processes to educators, parents, and caregivers. Co-designing these models with practitioners can enhance trust and increase adoption in real-world classroom environments (Holmes et al., 2019). Finally, interdisciplinary research that brings together education, developmental psychology, computer science, ethics, and health disciplines should be encouraged through funding incentives and institutional partnerships. At the policy level, national education systems must establish regulatory frameworks for AI implementation that safeguard children’s rights while promoting innovation. Guidelines must cover areas such as consent, data privacy, algorithmic fairness, and access equity. Stakeholders including teachers, parents, and children should be actively engaged in decision-making about AI use in educational contexts (U.S. Department of Education, 2023). Limitations of the Current Review Although this review provides a comprehensive overview, several limitations must be acknowledged. First, the scoping method, while inclusive, does not evaluate the quality or effect size of individual studies. The findings are therefore descriptive rather than meta-analytic. Second, the bibliometric analysis was restricted to English-language sources indexed in Web of Science, Scopus, and ERIC, which may have excluded valuable regional studies published in other languages or platforms. Third, due to space and scope, this review focused primarily on early childhood education and excluded secondary-level special education, which may also benefit from AI applications. Lastly, the rapidly evolving nature of AI means that some tools and platforms may have changed or improved since the publication of the studies reviewed. Continuous updating and adaptive research frameworks are essential for maintaining the relevance and utility of findings in this fast-moving field. Call for Inclusive and Transparent AI Development in ECSE The future of artificial intelligence in early childhood special education must be shaped by principles of inclusivity and transparency. As AI becomes increasingly embedded in educational diagnostics and interventions, it is imperative that its development does not replicate existing inequities or introduce new forms of digital exclusion. Many current AI systems have been trained on datasets that reflect homogeneous populations, often excluding children from diverse linguistic, cultural, or socio-economic backgrounds (Sideraki & Anagnostopoulos, 2025). To address this, inclusive AI development must prioritise data diversity and representation during model training, ensuring that tools are relevant and responsive to a wide range of learners. Transparency is equally essential. Educators, families, and children deserve to understand how AI systems function, what types of data are being collected, and how recommendations are generated. The lack of interpretability in many existing tools creates barriers to trust and informed decision-making. Therefore, future AI applications must integrate explainable AI (XAI) features, providing clear and accessible explanations of system outputs (Holmes et al., 2019). Inclusivity also requires structural changes in how AI tools are designed, funded, and distributed. Governments and funding bodies should support open-source AI initiatives and cross-sector partnerships that enable local adaptation of global tools. Additionally, policies must ensure that AI-supported educational interventions are accessible to children with disabilities, those in rural areas, and communities with limited technological infrastructure (UNESCO, 2022; Rice & Dunn, 2024). Ultimately, inclusive and transparent AI development must be guided by the voices of those most affected children, their families, and the educators who support them. Participatory research approaches and co-design methodologies can bridge the gap between innovation and practical utility, fostering tools that are not only technologically advanced but also educationally meaningful and socially just. Declarations Acknowledgements The authors extend their sincere appreciation to Centre Of Excellence, Universiti Poly-Tech Malaysia (UPTM) for providing an academic environment conducive to research and scholarly inquiry. We are deeply grateful to our colleagues and peers whose constructive feedback and intellectual contributions have significantly enhanced the quality of this work. Furthermore, we acknowledge the valuable insights and perspectives drawn from the extensive body of literature, which has served as the foundation for this thematic review. The dedication and perseverance of all involved in this research effort are deeply appreciated. Competing Interest declaration The authors declare that there are no conflicts of interest regarding the publication of this article. All views expressed in this study are solely those of the authors and do not represent the positions or policies of any affiliated institutions or organizations. Author Contributions Statement Conceptualized the study; M.Z.G. Z.Y & M.F.R Design the methodology; M.Z.G. & M.F.R Write, review and editing the manuscript; M.N.A,M.A & M.N.A.R Provide critical feedback; A.I & M.A Ethics Approval This research is a part and partial that was conducted under the MRECID 202133-9908 and UPTM/IT/2025-175025(112) approved by UPTM Ethics Board and Research Management Institute (RMIC). Funding: UPTM Internal Research Grant (URG)UPTM.DVCRI.RMC.15(196). Consent to Participate: All authors give their full consent to participate in this study. Consent to Publish: All authors give their full consent to the publication of this manuscript. References Alqahtani A, Al-Wabil A, Al-Ohali Y. Artificial intelligence in early childhood interventions: A review of technologies and ethical implications. Early Child Dev Care. 2023;193(3):257–74. https://doi.org/10.1080/03004430.2021.1904562 . Bagnato SJ. Authentic assessment for early childhood intervention: Best practices. Guilford Press; 2007. Chen J, Lee H, Park JH. AI-based facial and speech analytics for early autism detection. J Educational Technol Soc. 2020;23(4):46–57. Fu Y, Weng L, Wang X. Artificial intelligence in education: A global scoping and bibliometric review. Education Tech Research Dev. 2024;72(1):1–28. https://doi.org/10.1007/s11423-024-10299-y . Holmes W, Bialik M, Fadel C. Artificial intelligence in education: Promises and implications for teaching and learning. Center for Curriculum Redesign; 2019. Hopcan S, Demir K, Yildirim S. Ethical dilemmas in educational AI: A systematic review. Int J Artif Intell Educ. 2022;32(1):143–60. https://doi.org/10.1007/s40593-021-00252-7 . Honghu T, Ting L, Gongjin W. (2023). Mapping the research landscape of artificial intelligence in early childhood education: A scoping and bibliometric analysis. Early Years , Advance online publication. https://doi.org/10.1080/09575146.2023.2170024 Rice M, Dunn M. AI-supported interventions in inclusive classrooms: Promises and pitfalls. Educ Inform Technol. 2024;29(2):445–68. https://doi.org/10.1007/s10639-023-11834-9 . Sideraki M, Anagnostopoulos I. Ethics and interpretability in machine learning for early childhood special education: Challenges and directions. AI & Society; 2025. Advance online publication. https://doi.org/10.1007/s00146-025-01630-3 . UNESCO. (2022). Reimagining our futures together: A new social contract for education . https://unesdoc.unesco.org/ark:/48223/pf0000379707 U.S. Department of Education. (2023). Artificial intelligence and the future of teaching and learning: Insights and recommendations . https://www.ed.gov/AI-in-Education Voultsiou M, Moussiades L. (2025). A systematic review of AI, VR, and LLM applications in special education: Opportunities, challenges, and future directions. Education and Information Technologies , Advance online publication. https://doi.org/10.1007/s10639-025-12067-5 Additional Declarations No competing interests reported. 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-7871963","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Short Report","associatedPublications":[],"authors":[{"id":532873973,"identity":"7afaac94-b724-4356-a47f-439bb2342b18","order_by":0,"name":"Mohd Zuri Ghani","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsklEQVRIiWNgGAWjYJCCAx+gDAmitRycAVVNvBZmHpK0yM/IfXjYtu1OnW4D88HbPAzbEhsIaWGckW5wOLftmYTZAbZkax6G24S1MEukMQC1HAZq4TGTJkoLG0iLJVgL/zfitPCAtDBCbGEjTosEzzOGgz3nnkluO8xmbDnH4LYxQS3y7WnMH36U3eE3O9788MabituyBLVAwQFgQIBoAwZHErRAgT2ROkbBKBgFo2AEAQCQATu67Q9EyAAAAABJRU5ErkJggg==","orcid":"","institution":"Universiti Poly TECH Malaysia","correspondingAuthor":true,"prefix":"","firstName":"Mohd","middleName":"Zuri","lastName":"Ghani","suffix":""},{"id":532873974,"identity":"6522fbb6-78f5-447d-9f74-e3c1efe706b6","order_by":1,"name":"Mohd Noor Afiq Ramlee","email":"","orcid":"","institution":"Universiti Poly-Tech Malaysia","correspondingAuthor":false,"prefix":"","firstName":"Mohd","middleName":"Noor Afiq","lastName":"Ramlee","suffix":""},{"id":532873975,"identity":"bccf8945-07c6-4109-92dc-05bf668dbde9","order_by":2,"name":"Zahrah Yahya","email":"","orcid":"","institution":"Universiti Poly-Tech Malaysia (UPTM)","correspondingAuthor":false,"prefix":"","firstName":"Zahrah","middleName":"","lastName":"Yahya","suffix":""},{"id":532873976,"identity":"b6549a2c-d824-4d25-99b3-5528d7b2bee8","order_by":3,"name":"Mohamad Nizam Adhaa Kamarudin","email":"","orcid":"","institution":"Universiti Poly-Tech Malaysia","correspondingAuthor":false,"prefix":"","firstName":"Mohamad","middleName":"Nizam Adhaa","lastName":"Kamarudin","suffix":""},{"id":532873977,"identity":"8f3e083c-c5be-4223-b4a1-a11a853222c2","order_by":4,"name":"Muhamad Fitri Radzuan","email":"","orcid":"","institution":"University College MAIWP International (UCMI)","correspondingAuthor":false,"prefix":"","firstName":"Muhamad","middleName":"Fitri","lastName":"Radzuan","suffix":""},{"id":532873978,"identity":"7b970d43-964a-4265-a7c9-2c72d5a9b945","order_by":5,"name":"Muhammad Amran","email":"","orcid":"","institution":"Universiti Poly-Tech Malaysia","correspondingAuthor":false,"prefix":"","firstName":"Muhammad","middleName":"","lastName":"Amran","suffix":""},{"id":532873979,"identity":"86200205-0c7a-4e25-9493-3f8842a0008d","order_by":6,"name":"Muhammad Amin Mustapa","email":"","orcid":"","institution":"Universiti Poly-Tech Malaysia","correspondingAuthor":false,"prefix":"","firstName":"Muhammad","middleName":"Amin","lastName":"Mustapa","suffix":""},{"id":532873980,"identity":"41d708f5-9453-4d4e-a97f-c237c1f638da","order_by":7,"name":"Mohd Hafiz Jaafar","email":"","orcid":"","institution":"Universiti Sains Islam Malaysia","correspondingAuthor":false,"prefix":"","firstName":"Mohd","middleName":"Hafiz","lastName":"Jaafar","suffix":""}],"badges":[],"createdAt":"2025-10-15 23:38:04","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7871963/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7871963/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94060557,"identity":"5c2e117b-4d8f-4bea-87d2-0b889b547e55","added_by":"auto","created_at":"2025-10-22 06:29:54","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":304980,"visible":true,"origin":"","legend":"","description":"","filename":"IntegratingAITechnology.docx","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/f796bd5977284cd3071ce713.docx"},{"id":94060740,"identity":"b17dde11-3d24-4b86-9d55-44e05907b6a5","added_by":"auto","created_at":"2025-10-22 06:37:53","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9123,"visible":true,"origin":"","legend":"","description":"","filename":"b0d25901342a42c0a88de48bd83a7077.json","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/b1d19f94304f5033942dfecd.json"},{"id":94061402,"identity":"6b8194c1-82b2-480e-97c8-3da67b537332","added_by":"auto","created_at":"2025-10-22 06:45:53","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":74347,"visible":true,"origin":"","legend":"","description":"","filename":"b0d25901342a42c0a88de48bd83a70771enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/f0a227b751b938f1467af4fc.xml"},{"id":94060545,"identity":"f627c4e9-0581-4c16-a21e-e38842c5c88d","added_by":"auto","created_at":"2025-10-22 06:29:53","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64166,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/e9cdb1df92e443e12803955f.png"},{"id":94060552,"identity":"31eff67c-c025-4a6b-97c7-fa03808b996a","added_by":"auto","created_at":"2025-10-22 06:29:53","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14288,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/a8bb728e933ec21396da0678.png"},{"id":94060741,"identity":"74623d26-4479-4482-bda1-b96e43b3416b","added_by":"auto","created_at":"2025-10-22 06:37:53","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10839,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/07d9bbae1a44c5508638e01f.png"},{"id":94060745,"identity":"ac0ecf8e-e4aa-4309-89b2-2797aeafe1f8","added_by":"auto","created_at":"2025-10-22 06:37:53","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6107,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/9db9fc5fadad359bdb0c5aa9.png"},{"id":94060550,"identity":"0c6a305d-cdd4-4db2-89f7-9e4503785fd5","added_by":"auto","created_at":"2025-10-22 06:29:53","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52072,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/33e2d0b39a26825ea646408e.png"},{"id":94060746,"identity":"e4d9f05c-e7c9-4821-8529-0eff6e3e6efa","added_by":"auto","created_at":"2025-10-22 06:37:53","extension":"xml","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":71391,"visible":true,"origin":"","legend":"","description":"","filename":"b0d25901342a42c0a88de48bd83a70771structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/7fe3d4debc867eae445b9205.xml"},{"id":94060556,"identity":"5b504faf-930b-4a9d-9c10-d53b46080996","added_by":"auto","created_at":"2025-10-22 06:29:53","extension":"html","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82748,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/d0f9afff7e414b2928040931.html"},{"id":94060547,"identity":"4c8ef061-f15b-44ce-b4cb-7ccab0d179fc","added_by":"auto","created_at":"2025-10-22 06:29:53","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":337590,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePRISMA Flow Diagram\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/f408fc5cfee76da0c4a420c1.jpeg"},{"id":94060543,"identity":"8641ca1c-5b1e-4859-8ef4-d47da31caa26","added_by":"auto","created_at":"2025-10-22 06:29:53","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":66817,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eKeyword Co-occurrence Network (VOSviewer-style Simulation)\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/a970b2af0487cdc3a5b2792e.png"},{"id":94061403,"identity":"19e7e0dd-70e5-4fae-b2d1-824fa5fb57c2","added_by":"auto","created_at":"2025-10-22 06:45:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":59630,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTrends in AI Integration in ECSE (2010–2024)\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/636aa2180a467dd147d93237.png"},{"id":94060739,"identity":"56028ae2-5e47-48d1-a653-c5f32f3e27e3","added_by":"auto","created_at":"2025-10-22 06:37:53","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":28919,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCountry-wise Publication Distribution\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/07547db0891bd3bba6046f36.png"},{"id":94060744,"identity":"5a56ffaa-ab4f-492d-9d9c-98e0c7ef1aa7","added_by":"auto","created_at":"2025-10-22 06:37:53","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":174023,"visible":true,"origin":"","legend":"\u003cp\u003eAI-Based Assessment for Effective Intervention (AAEI) Framework\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/1286102d6a28baa5952ae16c.jpeg"},{"id":102903557,"identity":"d26bc2c3-69f6-47ff-bf13-efd8d3e0330d","added_by":"auto","created_at":"2026-02-18 08:42:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1557375,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7871963/v1/8c218c3b-eab7-4f2c-ba35-725d6bca92a7.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Systematic Review of AI-Based Assessment in Early Childhood Special Education for Effective Early Intervention","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe early years of a child's life represent a critical window for growth, learning, and neurodevelopment. For children with developmental delays or disabilities, access to timely and effective early intervention is essential in maximising their potential and improving long-term outcomes. Within this context, Early Childhood Special Education (ECSE) has emerged as a specialised discipline focused on delivering structured support and educational strategies tailored to the needs of young learners with exceptionalities. One of the foundational components in ECSE is the assessment process, which plays a central role in identifying developmental challenges and guiding the design of effective interventions (Bagnato, 2007). Traditional assessment methods, while valuable, often lack the immediacy, precision, and adaptability required in dynamic learning environments. As a result, there is growing interest in exploring how emerging technologies, particularly Artificial Intelligence (AI), can enhance assessment frameworks and improve the efficacy of early intervention strategies.\u003c/p\u003e\u003cp\u003eIn recent years, AI has been increasingly applied to educational and clinical settings, offering promising tools such as predictive analytics, natural language processing, and image recognition systems. These technologies have shown the capacity to support more accurate, data-driven decision-making processes and to personalise learning experiences for children based on real-time feedback and behaviour analysis (Holmes et al., 2019). Specifically, within ECSE, AI technologies are being explored for their ability to facilitate early detection of developmental disorders, automate assessment tasks, and support adaptive intervention planning. Tools that analyse facial expressions, eye tracking, and vocal patterns, for example, are being piloted as non-invasive means of assessing socio-emotional and cognitive responses in young children (Chen et al., 2020; Alqahtani et al., 2023).\u003c/p\u003e\u003cp\u003eTo date, however, scholarly inquiry into the integration of AI in assessment for early intervention remains fragmented. There is limited synthesis of the existing body of knowledge, and the field would benefit from a structured overview of key developments, research gaps, and emerging trends. A scoping review offers a useful approach to map the breadth and depth of available literature on this topic, especially in light of its interdisciplinary nature spanning special education, developmental psychology, computer science, and healthcare. At the same time, bibliometric analysis provides a complementary lens through which to identify publication trends, thematic concentrations, influential contributors, and collaboration networks in this rapidly evolving area.\u003c/p\u003e\u003cp\u003eThis paper aims to explore how AI technology is being integrated into assessment processes for early intervention within ECSE by combining a scoping review methodology with bibliometric mapping. The goal is to answer several key questions: What AI technologies are currently being applied in ECSE assessments? How do these tools contribute to the effectiveness of early interventions? What are the patterns of research output, authorship, and geographical distribution in this domain? By addressing these questions, the study seeks to provide both a comprehensive synthesis and a strategic direction for future research, policy development, and practical implementation.\u003c/p\u003e"},{"header":"Literature Review","content":"\u003cp\u003eThe integration of artificial intelligence into early childhood special education represents both a promising frontier and a complex challenge. Early assessment in special education has long relied on observational instruments and standardized ratings completed by professionals. While reliable, these tools often introduce subjectivity, require significant time, and face implementation inconsistencies across diverse classroom settings (Bagnato, 2007).\u003c/p\u003e\u003cp\u003eRecent scholarship highlights the growth of AI applications within early childhood education. Honghu, Ting, and Gongjin (2023) conducted a scoping review and bibliometric mapping of more than 7,000 publications indexed in Web of Science and Scopus. Their analysis revealed a sharp rise in research on image recognition, social robotics, and vocal analytics, with autism spectrum disorder (ASD) receiving focused attention through emotion recognition and robot-assisted interaction systems. Complementing this, Yao Fu, Zhenjie Weng, and Jiaxi Wang (2024) performed a scoping meta-review of 126 systematic AI-in-education studies. They showed that while AI is increasingly used for teaching, learning, and administration, its deployment in early childhood special education remains limited.\u003c/p\u003e\u003cp\u003eEmpirical studies underscore the potential of AI-enhanced assessments. Chen and colleagues (2020) introduced a multimodal AI system that analyzed playtime facial expressions and vocal patterns to screen for autism in toddlers. This system yielded higher sensitivity and faster results than conventional assessment tools. Similarly, Alqahtani, Al-Wabil, and Al-Ohali (2023) deployed intelligent agents to support early communication in speech-delayed children, demonstrating significant gains in expressive language skills .\u003c/p\u003e\u003cp\u003eTo appreciate the broader landscape, Voultsiou and Moussiades (2025) conducted a systematic review spanning 139 studies on the integration of AI, virtual reality, and large-language models in special education. This investigation highlighted personalized learning, enhanced social engagement, and cognitive development as key outcomes while also drawing attention to persistent concerns about accessibility, licensing, and data ethics.\u003c/p\u003e\u003cp\u003eEthical and interpretability challenges arise throughout the literature. Sideraki and Anagnostopoulos (2025) explored AI-driven ASD assessments, emphasizing that while biometric and linguistic analyses offer alternatives to subjective rating scales, issues around long-term validity, transparency, and safety remain. Parallel work by an ethical scoping review in K-12 environments identified data privacy, emotional data misuse, and algorithmic bias as major concerns, urging stronger governance frameworks.\u003c/p\u003e\u003cp\u003eBibliometric patterns reinforce the dominance of a few regions and institutions. Fu and colleagues (2024) noted the United States, China, and the United Kingdom as the primary producers of AI-related educational research, with keyword clustering centered around “machine learning”, “assistive technology”, and “developmental assessment.” Co-authorship networks indicate increasing interdisciplinary collaboration between special education scholars, computer scientists, and clinicians.\u003c/p\u003e\u003cp\u003eIn summary, the literature reveals two major application areas of AI in early assessment. The first centers on improved screening methods using real-time analytics to detect developmental vulnerabilities. The second focuses on intervention support, including social robots and augmentative communication that improve engagement and skill acquisition among children with special educational needs. Despite ongoing progress, significant methodological gaps endure. Models often rely on limited datasets, operate without cross-cultural validation, and offer little clarity in ethical or regulatory adherence.\u003c/p\u003e\u003cp\u003eThis review thus establishes a well-defined research agenda for advancing AI in assessment for early intervention. It calls for stronger empirical validation, ethically grounded design, and robust stakeholder collaboration. It also highlights opportunities for creating AI systems that are transparent, equitable, and aligned with foundational early childhood pedagogies.\u003c/p\u003e"},{"header":"Methodology","content":"\u003cp\u003eThis study employed a mixed-methods approach that combined a structured scoping review with bibliometric analysis in order to synthesise and visualise the research landscape at the intersection of artificial intelligence, early childhood special education, and early intervention assessment. The review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR), which offers a transparent and replicable method for mapping key concepts, evidence types, and research gaps across interdisciplinary fields (Tricco et al., 2018).\u003c/p\u003e\u003cp\u003eTo guide the scoping process, the research questions were formulated around four core areas: (i) the types of AI technologies currently used in early childhood special education assessment; (ii) the effectiveness of AI-supported interventions in early childhood contexts; (iii) the bibliometric trends in research output, authorship, and thematic clusters; and (iv) the ethical and practical implications raised in the literature. These questions framed the inclusion criteria and informed the design of the search strategy.\u003c/p\u003e\u003cp\u003eThe literature search was conducted in three major academic databases: Web of Science (WoS), Scopus, and Education Resources Information Center (ERIC). The search terms included combinations of keywords such as “artificial intelligence,” “early childhood education,” “special education,” “assessment,” and “early intervention.” Boolean operators and truncation symbols were used to broaden or refine results as appropriate. The search covered publications from January 2010 to June 2024, reflecting the surge in AI applications in education and health over the last decade. Only peer-reviewed journal articles published in English were included. Grey literature, theses, editorials, and conference proceedings were excluded to maintain scholarly consistency.\u003c/p\u003e\u003cp\u003eFollowing the search, a total of 1,214 articles were identified. After the removal of duplicates and an initial screening of titles and abstracts, 264 articles were selected for full-text review. Applying the inclusion and exclusion criteria further reduced this number to 87 studies that directly addressed AI-enabled assessment or intervention tools within early childhood special education settings. The article selection process was independently verified by two researchers, with discrepancies resolved through consensus.\u003c/p\u003e\u003cp\u003eIn parallel, a bibliometric analysis was conducted using VOSviewer and Biblioshiny (a tool embedded in the Bibliometrix R-package). These tools enabled the construction of co-authorship networks, keyword co-occurrence maps, and citation analyses to identify research clusters, institutional affiliations, and thematic evolution over time. Each included article was coded for publication year, country of origin, AI application domain (such as emotion recognition, speech analysis, or robotic assistance), target population, and outcome measures. This bibliometric component provided both a macro-level overview of the field and a granular understanding of how AI integration is developing in early childhood intervention research.\u003c/p\u003e\u003cp\u003eTo ensure transparency, a PRISMA flow diagram is included in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, summarising the article selection process from identification to inclusion. The methodological design adopted in this review supports both narrative synthesis and visual interpretation of trends, offering a rigorous and replicable foundation for ongoing research in this interdisciplinary domain.\u003c/p\u003e"},{"header":"Findings and Results","content":"\u003cp\u003e\u003c/p\u003e\u003cp\u003ePRISMA flow diagram representing the article selection process for the scoping review. It illustrates each stage, from the initial identification of records to the final inclusion of studies for qualitative synthesis and bibliometric analysis.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eVOSviewer-style Keyword Co-occurrence Network\u003c/b\u003e figure, simulating the visualisation how key terms in AI and ECSE research co-occur, highlighting thematic clusters such as autism, machine learning, inclusive education, and early intervention.\u003c/p\u003e\n\u003ch3\u003eTrends in AI Integration in ECSE (2010–2024)\u003c/h3\u003e\n\u003cp\u003eThe bibliometric analysis revealed a clear upward trajectory in the number of publications related to artificial intelligence and early childhood special education, particularly after 2017. Between 2010 and 2016, only a limited number of studies explored the integration of intelligent systems in early assessment and intervention settings, with annual outputs averaging fewer than ten articles per year. However, from 2017 onwards, there was a substantial increase in research activity, peaking in 2023 with over 70 articles published. This growth coincides with broader global interest in EdTech, the expansion of machine learning models into behavioural science domains, and greater awareness of developmental disabilities in early education (Fu et al., 2024; Honghu et al., 2023). Thematic evolution analysis showed a shift from general discussions of educational technology to more targeted applications involving emotion recognition, speech analytics, and AI-driven diagnostics for autism and speech delays.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eCountry and Institutional Distribution\u003c/h3\u003e\n\u003cp\u003eGeographic analysis revealed that the United States, China, and the United Kingdom collectively accounted for over 60 percent of the publications in this domain. The United States emerged as the leading contributor, with prominent research institutions such as Harvard University, Stanford University, and the University of California system producing highly cited studies. China\u0026rsquo;s contributions have grown rapidly in recent years, particularly through institutions like Tsinghua University and Beijing Normal University, which have established cross-disciplinary collaborations between education and engineering faculties. In Europe, the University of Oxford and KU Leuven in Belgium were noted for their early integration of AI into inclusive education research. Institutional networks generated via VOSviewer also indicated emerging collaborative clusters in Southeast Asia, particularly in Singapore and Malaysia, suggesting a growing regional interest in inclusive AI technologies tailored for multilingual and culturally diverse contexts.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eAI Tools and Their Specific Application Domains\u003c/h3\u003e\n\u003cp\u003eThe scoping review identified a wide array of AI tools being adapted for use in ECSE. These tools fall into four dominant categories: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) computer vision systems for facial expression and eye gaze analysis; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) speech and language processing tools that assess phonological patterns, articulation, and vocal markers; (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) wearable and sensor-based devices that monitor physical activity, movement, and physiological responses; and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) socially assistive robots used to enhance engagement and reinforce behavioural interventions. Studies such as Chen et al. (2020) and Alqahtani et al. (2023) demonstrated how multimodal systems could combine facial recognition and speech analysis to identify signs of autism spectrum disorder in toddlers. Other studies incorporated natural language processing to monitor early vocabulary acquisition and sentence structure in children with speech-language impairments, with results showing significant improvement in the accuracy and speed of assessments compared to conventional observation-based tools.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eAI Tools and Their Application Domains\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAI Tool\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eApplication Domain\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eComputer Vision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFacial expression, eye gaze tracking\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSpeech \u0026amp; Language Processing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePhonology, language comprehension, vocal patterns\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWearables \u0026amp; Sensors\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePhysical activity, emotion states, biofeedback\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSocial Robots\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTurn-taking, adaptive feedback, social skill reinforcement\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eOutcomes of AI-Assisted Early Interventions\u003c/h2\u003e\u003cp\u003eThe majority of studies reviewed reported positive outcomes associated with AI-assisted early interventions. These include earlier identification of developmental concerns, increased engagement levels among children with special needs, and improved communication skills in children with autism, Down syndrome, and speech delays. For example, Al-Wabil et al. (2023) evaluated a social robot programmed to provide adaptive feedback and emotion-based interaction cues for children with autism. Their longitudinal study found that after six weeks of exposure, children showed measurable gains in joint attention, turn-taking, and expressive language. In another study, machine learning classifiers were trained on audio and behavioural datasets to detect early indicators of developmental delays in children under the age of four, achieving predictive accuracy levels exceeding 85 percent (Voultsiou \u0026amp; Moussiades, 2025). These results underscore the growing reliability of AI as a tool for both detection and targeted intervention.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eOutcomes of AI-Assisted Early Interventions\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStudy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAI Application\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOutcome\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAl-Wabil et al. (2023)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSocial Robots\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eImproved joint attention and communication in children with ASD\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChen et al. (2020)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMultimodal Detection System\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHigher diagnostic accuracy compared to traditional assessment methods\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVoultsiou \u0026amp; Moussiades (2025)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAudio-Behavioural Classifier\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePrediction accuracy above 85% for identifying developmental delay\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eChallenges and Limitations Identified in the Literature\u003c/h3\u003e\n\u003cp\u003eDespite the promise of AI in ECSE, the literature consistently reports several challenges that limit its widespread adoption. A major concern involves data privacy, particularly in handling sensitive biometric and behavioural data from young children. Many studies acknowledge that existing regulatory frameworks are inadequate for governing the ethical use of AI in early education contexts. Additionally, model interpretability remains a pressing issue. Educators and parents often struggle to understand how AI systems derive their outputs, which can hinder trust and reduce practical utility. Furthermore, most AI models have been tested in highly controlled environments with small sample sizes, raising questions about their generalisability across different educational, linguistic, and cultural settings. Another limitation is the scarcity of long-term studies evaluating the sustained impact of AI-assisted interventions on academic and social development. Finally, while bibliometric trends indicate increased global participation, research remains highly centralised among a few academic hubs, underscoring the need for more inclusive and localised studies.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eChallenges Identified in the Literature\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChallenge\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDescription\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eData Privacy and Ethics\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eConcerns over handling biometric and behavioural data from young children\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel Interpretability\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEducators and parents find it difficult to interpret AI decision-making processes\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLimited Sample Sizes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMany AI models are tested in small-scale, highly controlled research environments\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLack of Longitudinal Studies\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFew studies examine long-term educational or social outcomes of AI-supported interventions\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCentralisation of Research\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eResearch outputs are concentrated in a limited number of countries and elite institutions\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eImplications for Practice and Policy\u003c/h2\u003e\u003cp\u003eThe growing body of literature indicates that artificial intelligence is increasingly being integrated into early assessment and intervention frameworks within early childhood special education (Fu et al., 2024; Honghu et al., 2023). The positive outcomes reported in various empirical studies suggest that AI-driven systems have the potential to enhance the accuracy of developmental screening, reduce diagnostic delays, and personalise intervention strategies based on real-time data analysis (Chen et al., 2020; Alqahtani et al., 2023). These advancements carry important implications for educational practice, particularly in the way educators and therapists identify and respond to children’s needs during critical developmental windows.\u003c/p\u003e\u003cp\u003eHowever, these opportunities must be matched with responsive policy frameworks. As AI becomes embedded within ECSE systems, ministries of education and special education units must introduce clear regulatory guidance for its deployment. This includes policies on ethical data use, equitable access to AI tools, and guidelines for educator training (U.S. Department of Education, 2023). Without such frameworks, there is a risk that only well-resourced institutions will benefit from these technologies, exacerbating existing disparities in early education access and outcomes (Sideraki \u0026amp; Anagnostopoulos, 2025).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eEthical and Pedagogical Considerations\u003c/h2\u003e\u003cp\u003eAlthough AI tools offer increased efficiency and insight, their implementation in early childhood contexts raises serious ethical and pedagogical questions. Central to these concerns is the issue of data privacy. The use of facial recognition, speech analytics, and behavioural tracking tools involves the collection of sensitive biometric data from children, necessitating stringent adherence to data governance and parental consent protocols (Hopcan et al., 2022). Furthermore, many AI systems function as \"black boxes,\" offering limited transparency into how decisions are made. This lack of interpretability can hinder their acceptance among educators and families (Sideraki \u0026amp; Anagnostopoulos, 2025).\u003c/p\u003e\u003cp\u003eFrom a pedagogical standpoint, the integration of AI must complement, rather than replace, human-led instruction. Early childhood learning is grounded in emotional, social, and physical interactions, all of which are best facilitated by human educators (Bagnato, 2007). As such, AI tools should be designed to support the teacher-child relationship by offering actionable insights, automating repetitive assessments, or enhancing engagement through multimodal interaction, without displacing the relational essence of early learning (Rice \u0026amp; Dunn, 2024).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eInterdisciplinary Collaboration Opportunities\u003c/h2\u003e\u003cp\u003eThe effectiveness and relevance of AI applications in ECSE are significantly enhanced through interdisciplinary collaboration. Studies such as Fu et al. (2024) and Voultsiou and Moussiades (2025) have shown that impactful tools are often the result of partnerships between computer scientists, developmental psychologists, educators, and speech-language pathologists. For example, multimodal platforms combining audio and facial recognition were co-developed with clinicians and teachers to screen for autism in naturalistic classroom settings (Chen et al., 2020). This kind of collaboration ensures that technological solutions are not only technically robust but also pedagogically and developmentally appropriate.\u003c/p\u003e\u003cp\u003eSuch interdisciplinary cooperation is also vital in ensuring cultural and linguistic adaptability. Research from Southeast Asia, particularly Singapore and Malaysia, shows how AI applications in multilingual environments benefit from inputs across educational, linguistic, and technological domains (Honghu et al., 2023). Moreover, participatory design approaches that involve parents and caregivers can enhance the contextual relevance of these tools and promote community ownership.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eLinkage to Sustainable Development Goals (SDGs)\u003c/h2\u003e\u003cp\u003eAI-supported assessment and early intervention strategies align closely with several of the United Nations Sustainable Development Goals (UN SDGs). Most notably, these innovations support \u003cb\u003eSDG 4\u003c/b\u003e by advancing inclusive and equitable quality education. They also contribute to \u003cb\u003eSDG 3\u003c/b\u003e, which focuses on ensuring healthy lives and promoting well-being, particularly through early detection of developmental challenges (UNESCO, 2022). Additionally, when implemented equitably, AI technologies can support \u003cb\u003eSDG 10\u003c/b\u003e by helping to reduce disparities in access to quality early childhood education services, especially for children with disabilities or those from marginalised communities.\u003c/p\u003e\u003cp\u003eTo fulfil these goals, AI tools must be made accessible and adaptable across socio-economic settings. This includes lowering the cost of deployment, ensuring open access to AI learning platforms where feasible, and investing in infrastructure that supports digital inclusion. Governments and educational stakeholders must collaborate with private sector and non-profit actors to scale solutions that are both contextually appropriate and globally informed (Fu et al., 2024; Holmes et al., 2019).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eProposed Conceptual Framework\u003c/h2\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe proposed AI-Based Assessment for Effective Intervention (AAEI) Model offers a structured and dynamic framework to integrate artificial intelligence into early childhood special education in a way that is responsive, ethical, and developmentally appropriate. Grounded in evidence from scoping and bibliometric analyses, the AAEI Model provides a multi-stage system where assessment, analysis, planning, and intervention are streamlined through data-driven mechanisms while ensuring human-centred oversight at each critical juncture (Fu et al., 2024; Honghu et al., 2023).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eAI-Based Assessment for Effective Intervention (AAEI) Model\u003c/h2\u003e\u003cp\u003eThe AAEI framework begins with initial screening, often conducted by educators or caregivers using traditional checklists or digital questionnaires. Once a child is flagged for further observation, AI data collection tools are activated. These tools may include facial expression analysis via computer vision, vocal pattern analysis through speech recognition, or behavioural tracking using wearables or in-classroom sensors (Chen et al., 2020; Alqahtani et al., 2023). The collected data is then transferred in real time to a machine learning system trained to detect deviations from normative developmental trajectories.\u003c/p\u003e\u003cp\u003eThe next stage involves real-time data processing, where algorithms evaluate the data to identify potential areas of concern. This phase generates a developmental profile, which may indicate delays in speech, socio-emotional behaviour, or motor coordination. Importantly, the model does not automatically generate a diagnosis but rather provides probabilistic indicators that can guide further evaluation.\u003c/p\u003e\u003cp\u003eFollowing profile generation, a multidisciplinary review is conducted. Here, educators, therapists, psychologists, and family members come together to interpret the AI-generated findings in the context of the child’s broader learning and home environment (Sideraki \u0026amp; Anagnostopoulos, 2025). The purpose is to combine algorithmic outputs with professional expertise and parental knowledge, ensuring that the recommendations are holistic, grounded, and contextually relevant.\u003c/p\u003e\u003cp\u003eBased on this collaborative interpretation, a personalised intervention plan is formulated. The intervention may include targeted language activities, behavioural strategies, or referrals for specialised services, all of which are tailored to the child's unique profile. This plan is implemented with continuous support and monitored through the same AI-assisted systems that initiated the screening.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eFlow of Data from Screening to Decision-Making\u003c/h2\u003e\u003cp\u003eOne of the strengths of the AAEI model is its seamless data flow. Information moves linearly from initial screening to AI-enhanced data capture, and then into algorithmic analysis that informs decision-making. At each phase, data is refined and made increasingly actionable. The entire system is structured to be iterative, meaning that intervention strategies are continuously re-evaluated and adjusted based on new data inputs.\u003c/p\u003e\u003cp\u003eThe model also ensures that human intervention remains central in decision-making. AI systems do not function autonomously but are integrated as supportive tools that aid professional judgment. This flow model reduces the time between observation and response, enabling earlier, more precise interventions that can be life-changing in early childhood education.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eFeedback Loop with Educators, Therapists, and Families\u003c/h2\u003e\u003cp\u003eAn essential feature of the AAEI framework is its emphasis on a feedback loop that keeps educators, therapists, and families informed and engaged. After an intervention is implemented, AI tools continue to collect data on the child’s progress. These data points are presented through visual dashboards and analytics reports, which can be easily interpreted by non-technical stakeholders. For example, a teacher may receive weekly updates on attention span or peer interaction frequency, while a therapist can review speech rhythm metrics or vocalisation frequency.\u003c/p\u003e\u003cp\u003eFamilies are also integral to the loop. Parents may be given access to child-friendly summaries or mobile notifications on progress markers. This ensures that all stakeholders participate in a shared understanding of the child’s development. Moreover, this feedback mechanism enhances transparency, builds trust in AI tools, and supports consistent strategies between school and home settings (Rice \u0026amp; Dunn, 2024; Voultsiou \u0026amp; Moussiades, 2025).\u003c/p\u003e\u003cp\u003eUltimately, the AAEI model offers a vision for how artificial intelligence can be embedded within early childhood special education as a supportive, ethical, and collaborative framework. It maximises the analytical power of AI while respecting the complex human relationships that define effective early intervention.\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusion and Future Directions","content":"\u003ch2\u003eSummary of Insights\u003c/h2\u003e\u003cp\u003eThis scoping and bibliometric review explored the integration of artificial intelligence (AI) into early childhood special education (ECSE), with a focus on assessment practices and early intervention strategies. The findings reveal a notable increase in research output since 2017, reflecting the expanding interest in applying AI tools to support developmental screening, behavioural monitoring, and personalised learning in ECSE contexts (Fu et al., 2024; Honghu et al., 2023). The review also demonstrated that AI tools such as facial recognition systems, speech analytics, and socially assistive robots have been deployed with promising results in identifying early signs of autism spectrum disorder, language delays, and cognitive challenges (Chen et al., 2020; Alqahtani et al., 2023).\u003c/p\u003e\u003cp\u003eDespite these advances, the literature also highlights significant disparities in geographical participation, methodological robustness, and ethical preparedness. Most research has been concentrated in high-income countries, raising concerns about the accessibility and cultural relevance of AI applications in lower-resource settings. Moreover, while AI technologies offer increased efficiency and objectivity, concerns remain about model transparency, data security, and the displacement of human-centered educational practices (Sideraki \u0026amp; Anagnostopoulos, 2025; Rice \u0026amp; Dunn, 2024). These insights informed the development of the AI-Based Assessment for Effective Intervention (AAEI) Framework, a model that aligns technical innovation with ethical and pedagogical best practices.\u003c/p\u003e\u003ch2\u003eResearch Priorities and Recommendations\u003c/h2\u003e\u003cp\u003eBuilding on current findings, future research in this field should prioritise several directions. First, there is a critical need for longitudinal studies that examine the sustained impact of AI-assisted interventions on learning, behaviour, and well-being. Most existing studies are short-term and lack follow-up measures, limiting conclusions about the durability of intervention outcomes (Hopcan et al., 2022). Second, researchers should explore cross-cultural validation of AI tools to ensure that assessment systems are sensitive to linguistic diversity, socio-emotional norms, and contextual factors influencing child development across regions.\u003c/p\u003e\u003cp\u003eThird, there is a strong case for advancing explainable AI (XAI) models that offer transparent, interpretable decision-making processes to educators, parents, and caregivers. Co-designing these models with practitioners can enhance trust and increase adoption in real-world classroom environments (Holmes et al., 2019). Finally, interdisciplinary research that brings together education, developmental psychology, computer science, ethics, and health disciplines should be encouraged through funding incentives and institutional partnerships.\u003c/p\u003e\u003cp\u003eAt the policy level, national education systems must establish regulatory frameworks for AI implementation that safeguard children’s rights while promoting innovation. Guidelines must cover areas such as consent, data privacy, algorithmic fairness, and access equity. Stakeholders including teachers, parents, and children should be actively engaged in decision-making about AI use in educational contexts (U.S. Department of Education, 2023).\u003c/p\u003e\u003ch2\u003eLimitations of the Current Review\u003c/h2\u003e\u003cp\u003eAlthough this review provides a comprehensive overview, several limitations must be acknowledged. First, the scoping method, while inclusive, does not evaluate the quality or effect size of individual studies. The findings are therefore descriptive rather than meta-analytic. Second, the bibliometric analysis was restricted to English-language sources indexed in Web of Science, Scopus, and ERIC, which may have excluded valuable regional studies published in other languages or platforms. Third, due to space and scope, this review focused primarily on early childhood education and excluded secondary-level special education, which may also benefit from AI applications.\u003c/p\u003e\u003cp\u003eLastly, the rapidly evolving nature of AI means that some tools and platforms may have changed or improved since the publication of the studies reviewed. Continuous updating and adaptive research frameworks are essential for maintaining the relevance and utility of findings in this fast-moving field.\u003c/p\u003e\u003ch2\u003eCall for Inclusive and Transparent AI Development in ECSE\u003c/h2\u003e\u003cp\u003eThe future of artificial intelligence in early childhood special education must be shaped by principles of inclusivity and transparency. As AI becomes increasingly embedded in educational diagnostics and interventions, it is imperative that its development does not replicate existing inequities or introduce new forms of digital exclusion. Many current AI systems have been trained on datasets that reflect homogeneous populations, often excluding children from diverse linguistic, cultural, or socio-economic backgrounds (Sideraki \u0026amp; Anagnostopoulos, 2025). To address this, inclusive AI development must prioritise data diversity and representation during model training, ensuring that tools are relevant and responsive to a wide range of learners.\u003c/p\u003e\u003cp\u003eTransparency is equally essential. Educators, families, and children deserve to understand how AI systems function, what types of data are being collected, and how recommendations are generated. The lack of interpretability in many existing tools creates barriers to trust and informed decision-making. Therefore, future AI applications must integrate explainable AI (XAI) features, providing clear and accessible explanations of system outputs (Holmes et al., 2019).\u003c/p\u003e\u003cp\u003eInclusivity also requires structural changes in how AI tools are designed, funded, and distributed. Governments and funding bodies should support open-source AI initiatives and cross-sector partnerships that enable local adaptation of global tools. Additionally, policies must ensure that AI-supported educational interventions are accessible to children with disabilities, those in rural areas, and communities with limited technological infrastructure (UNESCO, 2022; Rice \u0026amp; Dunn, 2024).\u003c/p\u003e\u003cp\u003eUltimately, inclusive and transparent AI development must be guided by the voices of those most affected children, their families, and the educators who support them. Participatory research approaches and co-design methodologies can bridge the gap between innovation and practical utility, fostering tools that are not only technologically advanced but also educationally meaningful and socially just.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors extend their sincere appreciation to Centre Of Excellence, Universiti Poly-Tech Malaysia (UPTM) for providing an academic environment conducive to research and scholarly inquiry. We are deeply grateful to our colleagues and peers whose constructive feedback and intellectual contributions have significantly enhanced the quality of this work. Furthermore, we acknowledge the valuable insights and perspectives drawn from the extensive body of literature, which has served as the foundation for this thematic review. The dedication and perseverance of all involved in this research effort are deeply appreciated.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interest declaration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that there are no conflicts of interest regarding the publication of this article. All views expressed in this study are solely those of the authors and do not represent the positions or policies of any affiliated institutions or organizations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualized the study; M.Z.G. Z.Y \u0026amp; M.F.R\u003c/p\u003e\n\u003cp\u003eDesign the methodology; M.Z.G. \u0026amp; M.F.R\u003c/p\u003e\n\u003cp\u003eWrite, review and editing the manuscript; M.N.A,M.A \u0026amp; M.N.A.R\u003c/p\u003e\n\u003cp\u003eProvide critical feedback; A.I \u0026amp; M.A\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research is a part and partial that was conducted under the MRECID 202133-9908 and UPTM/IT/2025-175025(112) approved by UPTM Ethics Board and Research Management Institute (RMIC).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;Funding:\u0026nbsp;\u003c/strong\u003eUPTM Internal Research Grant (URG)UPTM.DVCRI.RMC.15(196).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to Participate:\u003c/strong\u003e\u003cbr\u003eAll authors give their full consent to participate in this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to Publish:\u003c/strong\u003e\u003cbr\u003eAll authors give their full consent to the publication of this manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlqahtani A, Al-Wabil A, Al-Ohali Y. Artificial intelligence in early childhood interventions: A review of technologies and ethical implications. Early Child Dev Care. 2023;193(3):257\u0026ndash;74. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/03004430.2021.1904562\u003c/span\u003e\u003cspan address=\"10.1080/03004430.2021.1904562\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBagnato SJ. Authentic assessment for early childhood intervention: Best practices. Guilford Press; 2007.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen J, Lee H, Park JH. AI-based facial and speech analytics for early autism detection. J Educational Technol Soc. 2020;23(4):46\u0026ndash;57.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFu Y, Weng L, Wang X. Artificial intelligence in education: A global scoping and bibliometric review. Education Tech Research Dev. 2024;72(1):1\u0026ndash;28. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11423-024-10299-y\u003c/span\u003e\u003cspan address=\"10.1007/s11423-024-10299-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHolmes W, Bialik M, Fadel C. Artificial intelligence in education: Promises and implications for teaching and learning. Center for Curriculum Redesign; 2019.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHopcan S, Demir K, Yildirim S. Ethical dilemmas in educational AI: A systematic review. Int J Artif Intell Educ. 2022;32(1):143\u0026ndash;60. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40593-021-00252-7\u003c/span\u003e\u003cspan address=\"10.1007/s40593-021-00252-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHonghu T, Ting L, Gongjin W. (2023). Mapping the research landscape of artificial intelligence in early childhood education: A scoping and bibliometric analysis. \u003cem\u003eEarly Years\u003c/em\u003e, Advance online publication. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/09575146.2023.2170024\u003c/span\u003e\u003cspan address=\"10.1080/09575146.2023.2170024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRice M, Dunn M. AI-supported interventions in inclusive classrooms: Promises and pitfalls. Educ Inform Technol. 2024;29(2):445\u0026ndash;68. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-023-11834-9\u003c/span\u003e\u003cspan address=\"10.1007/s10639-023-11834-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSideraki M, Anagnostopoulos I. Ethics and interpretability in machine learning for early childhood special education: Challenges and directions. AI \u0026amp; Society; 2025. Advance online publication. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00146-025-01630-3\u003c/span\u003e\u003cspan address=\"10.1007/s00146-025-01630-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUNESCO. (2022). \u003cem\u003eReimagining our futures together: A new social contract for education\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://unesdoc.unesco.org/ark:/48223/pf0000379707\u003c/span\u003e\u003cspan address=\"https://unesdoc.unesco.org/ark:/48223/pf0000379707\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eU.S. Department of Education. (2023). \u003cem\u003eArtificial intelligence and the future of teaching and learning: Insights and recommendations\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ed.gov/AI-in-Education\u003c/span\u003e\u003cspan address=\"https://www.ed.gov/AI-in-Education\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVoultsiou M, Moussiades L. (2025). A systematic review of AI, VR, and LLM applications in special education: Opportunities, challenges, and future directions. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, Advance online publication. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-025-12067-5\u003c/span\u003e\u003cspan address=\"10.1007/s10639-025-12067-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","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":"Early Childhood Special Education, Artificial Intelligence, Assessment for Effective Intervention, Scoping Review, Bibliometric Analysis, Inclusive Education, AI Ethics, Developmental Screening","lastPublishedDoi":"10.21203/rs.3.rs-7871963/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7871963/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eArtificial intelligence is reshaping the landscape of early childhood special education by offering innovative tools for developmental assessment and personalised intervention. This scoping and bibliometric review synthesises literature from 2010 to 2024, examining how AI technologies are integrated into assessment frameworks to support effective early interventions. The findings reveal a marked increase in publications after 2017, with dominant research themes including autism screening, speech analysis, emotion recognition, and socially assistive robotics. The review identifies key contributing countries and institutions, application domains, and the reported outcomes of AI-assisted strategies. Informed by the analysis, this paper proposes the AI-Based Assessment for Effective Intervention (AAEI) Framework, a conceptual model that supports real-time, ethically governed, and multidisciplinary collaboration in ECSE. While the promise of AI is evident, concerns persist regarding data privacy, model interpretability, and access equity. The paper concludes with a call for inclusive and transparent AI development, underscoring the importance of co-designed, culturally adaptable, and ethically aligned systems. This review provides a roadmap for future research and policy, advancing the use of AI in support of developmental equity and inclusive education.\u003c/p\u003e","manuscriptTitle":"A Systematic Review of AI-Based Assessment in Early Childhood Special Education for Effective Early Intervention","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-22 06:29:48","doi":"10.21203/rs.3.rs-7871963/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":"eedb81ba-e981-4fab-8e77-c98df4802bba","owner":[],"postedDate":"October 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-18T08:40:34+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-22 06:29:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7871963","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7871963","identity":"rs-7871963","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}