Supporting Iranian Teachers’ Language Assessment Literacy: A Hybrid AI–Human Feedback Approach within Collaborative Action Research

Supporting Iranian Teachers’ Language Assessment Literacy: A Hybrid AI–Human Feedback Approach within Collaborative Action Research | 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 Research Article Supporting Iranian Teachers’ Language Assessment Literacy: A Hybrid AI–Human Feedback Approach within Collaborative Action Research Mohammad Hossein Arefian This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8281675/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract This study explores the influence of a hybrid AI-human feedback model, incorporated within a Collaborative Action Research (CAR) framework, on developing Language Assessment Literacy (LAL) among Iranian English language teachers. Grounded in the challenges of conventional assessment practices in Iran’s EFL context, containing overreliance on summative testing and inadequate professional development, the research engaged a qualitative CAR methodology with 15 teachers over six months. Participants involved in iterative cycles of assessment design, application, and reflection, supported by AI-driven analytics (e.g., automated scoring and natural language processing (NLP) feedback) alongside structured peer and mentor collaboration. Findings indicated that the AI-CAR synergy improved teachers’ diagnostic precision, empowered culturally responsive assessment adaptations, and nurtured sustainable shifts toward formative practices. Main results involved improved recognition of learner patterns, critical negotiation of algorithmic feedback, and the emergence of teacher-led assessment innovations. Yet, challenges appeared concerning workload intensity and contextualizing AI recommendations within Iranian instructional and cultural norms. This study addressed the research problem of underdeveloped LAL among Iranian EFL school teachers, considered by overreliance on summative testing. Implementing a qualitative CAR design with 15 teachers over six months, it explored how a hybrid AI-human feedback model influences LAL development. Key findings indicate the model improved teachers’ diagnostic precision and nurtured formative assessment practices through critical negotiation of AI feedback. The conclusion underlines that sustainable LAL development requires hybrid models where AI augments, not substitutes, teacher judgment. Language Assessment Literacy Artificial Intelligence Collaborative Action Research Teacher Professional Development Formative Assessment EFL Context Introduction Language assessment literacy (LAL) has appeared as a significant aspect of language education, underscoring the knowledge, skills, and principles needed for professional and effective assessment practices (Levi & Inbar-Lourie, 2020 ). Within Iran’s English language teaching (ELT) context, enhancing teachers’ LAL is chiefly important because of the increasing request for high-stakes language assessments and the need for instructional alignment with international standards (Chang et al., 2024 ). Nevertheless, Iranian teachers regularly face issues in conducting rigorous assessment strategies because of restricted professional development opportunities, dependence on traditional assessment methods, and insufficient involvement with existing testing methods and techniques (Derakhshan & Ghiasvand, 2022 ). This problem is specifically important in the Iranian school context (secondary and language institute levels), where large class sizes, a centralized national curriculum, and pressure from university entrance exams further limit teachers’ capacity to creative in assessment. In spite of recognized needs, conventional top-down training workshops have established ineffective for supportable LAL development in Iran (Bazvand et al., 2025 ). A persistent gap exists in models that offer continuous, contextualized support. This study thus introduces a novel incorporated approach, stating that the problem of underdeveloped LAL needs a situated, cyclical, and data-fostered professional learning model. It explores whether and how a hybrid AI and human feedback approach, rooted within a collaborative action research (CAR) framework, can address this problem by offering teachers with iterative cycles of evidence-based reflection and collaborative sensemaking. By incorporating AI with organized human feedback, the study intends to generate a dynamic, collaborative, and reflective professional development model that enables teachers to cultivate their assessment practices and literacies systematically and continuously (Bhimdiwala et al., 2022 ). CAR is selected as the essential methodological framework because it goes beyond one-off training by integrating professional development within teachers’ own contexts. It offers a structured yet flexible process for teachers to collectively explore and modify their own assessment practices, generating it ideal for nurturing maintainable, context-sensitive LAL development. The concept of LAL has advanced considerably during the past two decades, with scholars highlighting its multidimensional nature. Inbar-Lourie ( 2016 ) describes LAL as the capacity to design, conduct, and evaluate language assessments while understanding their pedagogical and ethical implications. For teachers, LAL goes beyond mere assessment production; it includes interpreting assessment data, offering constructive feedback, and adjusting teaching according to learner needs and assessment results (Levi & Inbar-Lourie, 2020 ). In spite of its acknowledged significance, LAL remains underdeveloped in many educational contexts, mainly in districts where teacher training programs focus on content knowledge over assessment literacy (Sultana, 2019 ). In Iran, where established educational policies regularly dictate assessment practices, teachers often use summative assessments with minimal formative or alternative assessment incorporation (Firoozi et al., 2019 ). This dependence on traditional testing methods confines opportunities for meaningful learning feedback and fails to resolve different learner competencies (Clark, 2012 ). CAR has been encouraged as an efficient professional development strategy to fill theory-practice gaps in education through collaboration and action research (Qing-Li et al., 2019 ). By including teachers in systematic inquiry into their own practices, CAR enhances reflective instruction and enables educators to foster contextually appropriate solutions (Manfra, 2019 ). In the territory of assessment, CAR offers a planned but flexible framework for teachers to try novel assessment strategies, evaluate outcomes, and improve their assessment approaches continuously, systematically, and collectively (Yosief et al., 2024 ). Yet, former CAR models often necessitate personalized, instant, data-driven feedback mechanisms, supporting the continuous learning process (Schildkamp, 2019 ). This study resolves this restriction by integrating AI-driven analytics combined with human mentorship, allowing real-time assessment feedback and nurturing deeper teacher engagement with LAL principles. AI has transformed educational assessment by providing tailored feedback, automated scoring, and evidence-based understandings into learner performance (Kooli & Yusuf, 2025 ). AI-powered tools, including natural language processing (NLP) systems, can assess both written and spoken communication with remarkable precision, letting teachers to gain immediate insights into student performance (Zhao, 2024 ). When incorporated into teacher professional development, AI can function as a supportive aid, recognizing patterns in assessment, uncovering potential biases, and proposing research-backed recommendations for improvement (Maurya et al., 2025 ). Though, AI alone cannot substitute the comprehensive judgment of human educators, mainly in contexts necessitating cultural sensitivity and instructional flexibility (Mohammed & ‘Nell’Watson, 2019 ). A hybrid approach that includes AI efficiency with human expertise guarantees that teachers not only gain technical feedback but also participate in critical discussions about assessment validity, fairness, and educational influence (Zhang et al., 2025 ). This study places itself where LAL, CAR, and AI-enhanced feedback intersect, suggesting a novel model for Iranian teacher development. It uses AI systems to provide instant assessment insights, which are then incorporated with planned feedback from peers and mentors. Together, these features aim to nurture an ongoing and sustainable model of professional development. The blended method is particularly appreciated in the Iranian context, where restricted resources and overcrowded classrooms often make personalized teacher support difficult to accomplish (Ghourchian, 2024 ). By means of continuous CAR cycles, teachers will design, implement, and analyze assessments while getting AI-generated data and human insights. This dual feedback mechanism is anticipated to improve teachers’ confidence in assessment design, develop their ability to interpret student performance, and support the implementation of formative assessments along with summative evaluations (Morris et al., 2021 ). The importance of this study rests in its potential to reframe LAL development in resource-limited contexts by using technology-driven and collaborative learning approaches. Previous research on LAL in Iran has generally focused on workshops and training programs, which often fail to produce lasting effects (Bazvand et al., 2025 ). Unlike earlier approaches, this study incorporates professional development openly into teachers’ everyday routines, encouraging continuous involvement and practical significance. Moreover, while AI applications in education have been widely studied for student learning, its potential to improve teacher assessment literacy is still chiefly unnoticed, particularly in non-Western contexts (Ghimire, 2025 ). By filling this gap, the study contributes to worldwide discussions on AI’s role in teacher education while offering a replicable model for similar contexts. The upcoming sections will present the theoretical underpinnings of LAL, CAR, and AI in assessment, followed by the study’s methodology and findings. This research aims to indicate that merging AI and human feedback within the CAR framework can meaningfully foster Iranian teachers’ LAL, nurturing assessment practices that are impartial, more accurate, and educationally effective. The following research question is pursued throughout the study: How does a hybrid AI–human feedback model, combined within a CAR, support the development of LAL among Iranian English language teachers? Literature review The concept of LAL has been gaining considerable attention in language education, with researchers underlining its vigorous influence on the development of effective instruction and learning strategies (Gan & Lam, 2022 ). LAL encompasses the knowledge and abilities necessary to prepare, implement, and assess language tests, along with an awareness of their broader educational impact (Bazvand et al., 2025 ). Over time, this definition has extended to incorporate not only technical expertise but also ethical responsibility, principles of fairness, and the effective use of assessment results to guide instructional practices (Coombe et al., 2020 ). The multidimensional nature of LAL underlines its significance in enhancing equitable and valid assessment practices, mainly in contexts where traditional testing methods dominate (Abrar-ul-Hassan & Nassaji, 2024 ). In Iran, the dependence on summative assessments and centralized examination systems provide limited opportunities for teachers to engage with alternative assessment methods, including formative assessments, portfolios, and dynamic assessment strategies (Mirsanjari, 2025 ). This gap underscores the urgent need for professional development initiatives that foster teachers’ LAL, empowering them to adopt more different and instructionally sound assessment practices (Cui et al., 2025 ; Lam, 2015 ). Nevertheless, a significant gap exists in the literature on LAL development specially within school-level contexts, with much current research concentrated on higher education. This study aims to address this gap by situating its inquiry explicitly in the secondary school and language institute context. The challenges of fostering LAL among Iranian English language teachers are multidimensional, coming from systemic, instructional, and training-related limitations (Asad et al., 2021 ). Researchers discovered that many Iranian teachers are not introduced to modern assessment theories throughout their pre-service training, which leads to a heavy reliance on high-stakes exams and rote learning assessments (Aliakbari et al., 2023 ). This problem is compounded by the inadequacy of in-service professional development opportunities devoted to increasing LAL (Koh, 2011 ). Moreover, the cultural and institutional attention to standardized testing as a major measure of student achievement discourages teachers to try advanced assessment techniques (Au, 2022 ). Although such challenges are not exclusive to Iran, similar problems have been observed in other educational settings where teacher preparation programs place little stress on assessment literacy (Firoozi et al., 2019 ). Given that the Iranian context poses its own specific problems, existing limitations including restricted resources, overcrowded classrooms, and fixed curricular structures make it particularly difficult to embrace formative or performance-based assessment practices (Curry et al., 2020). Addressing these issues need a personalized approach that combines theoretical knowledge with practical, contextually relevant strategies for assessment reform. CAR has been increasingly acknowledged as an appreciated model for enhancing teachers’ professional development, particularly in developing LAL (Dulfer et al., 2024 ). According to Author (X), CAR includes a joint, participatory inquiry in which teachers examine their own classroom practices, evaluate the outcomes, and make ongoing modifications. Engrained in the ethos of reflective practice, this process allows teachers to take greater responsibility for their development while directly confronting authentic classroom issues (Benade, 2015 ). Within the framework of LAL, CAR provides teachers a structured yet flexible space to experiment novel assessment strategies, examine learners’ responses, and adjust their practices using evidence from the classroom. Accordingly, CAR can support teachers’ understanding of assessment by grounding professional learning in real instructional contexts (Yosief et al., 2024 ). However, traditional CAR approaches often fall short in offering instant feedback, which can slow the process of refining assessment methods (Yin & Buck, 2019 ). This gap focuses on the value of integrating technology-supported feedback tools, including AI-powered analytics, to work alongside human collaboration, empowering faster, more informed modifications. The application of AI in assessment moves beyond automated scoring. Present models leverage NLP to offer diagnostic feedback on linguistic features (e.g., syntactic complexity, lexical diversity) and machine learning to recognize patterns in student performance (Zhao, 2024 ). In formative assessment, AI can support feedback processes as conceptualized by scholars such as Carless and Dawson (2022), who highlight feedback as a dialogic process directed at closing the learning gap. Nevertheless, AI-produced feedback risks being monologic and decontextualized. A hybrid model, thus, must incorporate AI’s analytical capacity with human mediation to guarantee feedback is usable, timely, and embedded within the particular instructional context—a core challenge this study explores. The use of AI in educational assessment marks a significant transformation in the way teachers design, administer, and analyze their assessment practices. Tools driven by AI, for example automated essay evaluation systems and NLP applications, can deliver instant feedback on learner performance. In doing so, they lighten teachers’ administrative workload while providing precise, in-depth diagnostic information. Such tools hold specific value in large-scale assessment settings, where traditional grading demands major time and is susceptible to unpredictability. Along with improving efficiency, AI has the potential to reinforce assessment validity and reliability by finding trends in learner responses, flagging possible biases, and suggesting modifications grounded in empirical evidence (Topping et al., 2025 ). Trajkovski and Hayes ( 2025 ) demonstrate how AI can foster formative assessment by generating personalized feedback for individual learners and providing teachers practical, data-informed recommendations. Yet, researchers warn against placing excessive reliance on AI, emphasizing that human judgment remains essential for exploring the nuanced teaching and ethical issues essential in assessment (Malik et al., 2025 ). A joined model, blending the efficiency of AI with the interpretive and contextual expertise of teachers, helps guarantee that assessment practices are both culturally responsive and instructionally robust (Grab, 2025 ). The merging of AI and teacher professional development provides a promising, though still relatively unexplored, pathway for fostering LAL. Whereas existing scholarship on AI in education has generally focused on its impact on student learning, comparatively little attention has been given to its capability to reinforce teachers’ assessment expertise (Cope et al., 2021 ). This research gap is particularly obvious in non-Western contexts, where access to technology and changing levels of digital competence can influence the uptake of AI-based tools. In Iran, for instance, teacher education programs often lack continuous professional development, making AI-supported feedback systems a possibly scalable and cost-efficient resource for improving LAL (Erdem Coşgun, 2025 ). Such systems could review teachers’ assessment designs, recognize possible validity concerns, and suggest alternative approaches, while human mentors lead critical dialogues about the teaching significance of these recommendations (Smith, 2010 ). This two-tiered feedback process supports sociocultural perspectives on learning, which stress the role of social interaction and scaffolded support in professional development (Mun et al., 2020 ). Incorporating AI into a CAR model would let teachers to engage in continuous cycles of exploration, reflection, and refinement, effectively narrowing the gap between assessment theory and classroom practice. Current research has begun to explore how AI can supplement collaborative learning frameworks within teacher education. For example, Yan et al. ( 2025 ) illustrates that AI-enabled professional learning networks let teachers to share assessment techniques, provide constructive feedback, and collectively build knowledge within virtual communities of practice. Similarly, Kim ( 2024 ) highlight AI’s capacity to deliver timely feedback during CAR cycles, helping teachers modify their assessment methods based on data insights. These studies show that joining AI-driven and human feedback can advance both the efficiency and impact of LAL development, while maintaining the reflective and cooperative nature of CAR. However, applying these hybrid models successfully needs overcoming obstacles for example teachers’ confidence with technology, institutional backing, and guaranteeing that AI tools are compatible with local assessment traditions (Zhang et al., 2025 ). Specially, in the Iranian setting, where concerns like digital inequality and resistance to innovation exist, a gradual and inclusive strategy for incorporating AI into teacher professional development is vital (Ghourchian, 2024 ). The theoretical and empirical evidence presented in this review highlights the capacity of incorporating AI-driven and human feedback within CAR to foster LAL among English language teachers in Iran. By drawing together perspectives from research on LAL, CAR practices, and AI applications in education, this study presents a novel framework personalized to address the specific challenges confronted in the Iranian context while also contributing to broader discussions on teacher professional development. The following section will detail the methodological approach for applying and assessing this model. Theoretical Framework This study is established in an incorporation of sociocultural and reflective practitioner paradigms. Mainly, it focuses on Vygotskian sociocultural theory, wherein tools mediate higher-order psychological processes (Vygotsky, 1978). Here, AI acts as a mediating artifact that moves teachers’ analytical capabilities, offering external, data-driven insights into student performance. Yet, learning happens through social interaction; thus, the CAR framework establishes a community of practice (Lave & Wenger, 1991) where teachers collectively interpret (‘make sense of’) AI outputs, negotiating meaning and appropriating assessment concepts. This process exemplifies situated learning, where LAL development is entrenched in authentic assessment activities and social dialogue. Moreover, the cyclical CAR process operationalizes Schön’s (1983) model of the reflective practitioner. The AI-produced data function as a catalyst for reflection-in-action (e.g., real-time adjustment during a feedback session) and reflection-on-action (e.g., post-lesson analysis in CAR meetings). Through iterative cycles of planning, action, observation, and reflection, teachers engage in continuous knowledge construction, internalizing new assessment practices. Lastly, this interplay between technology, pedagogy, and content is practically examined through the TPACK framework (Mishra & Koehler, 2006). The study probes how CAR facilitates the iterative refinement of teachers’ Technological Pedagogical Content Knowledge (TPACK) by necessitating them to critically incorporate AI tools (TK) into language assessment design (PK/CK) within their particular EFL context. Methodology Research Design This research selects a qualitative CAR approach, following the iterative process of planning, acting, observing, and reflecting collaboratively within a community of practice (Bleicher, 2014). The aim of this study to explore how joining AI-based and human feedback can support Iranian English language teachers’ LAL. CAR functions both as the guiding methodology and the planned design, underlining professional development that is collaborative, contextually grounded, and responsive to participants’ needs (Bleicher, 2014). To obtain a richer perspective, the project integrates case study techniques (Aliakbari & Sadeghi, 2022) to monitor the professional development of individual teachers, as well as ethnographic observations to place the results within the cultural and institutional context of Iran’s educational system. Over a six-month period, 15 language teachers from both public and private institutions, who were selected through purposive sampling, participated in three CAR cycles. Each cycle included the construction, implementation, and evaluation of assessment tasks, maintained by AI tools (including automated essay scoring systems and NLP-based feedback platforms) in line with structured feedback from peers and mentors. By integrating CAR with technological tools, this design intends to diminish the gap between theoretical knowledge and practical application in LAL, while also providing a scalable professional development model suitable to resource-constrained settings. The hybrid model consisted of a sequenced process. First, teachers established AI-produced feedback on their assessment tasks and student responses by means of two main tools: (1) an automated writing evaluation (AWE) system offering scores and error tags (grammar, mechanics), and (2) an NLP dashboard analyzing lexical diversity and syntactic complexity in student output. Second, this data functioned as input for structured human feedback phases: (a) peer feedback in CAR meetings, where teachers compared AI outputs with their own judgments, and (b) mentor feedback, where a researcher enhanced discussions on supporting AI diagnostics with curricular goals and cultural appropriateness. Criteria for use concentrated on AI for recognizing patterns and human collaboration for interpretation and instructional action. Intervention Procedures: Teacher Activities and AI Tools Each six-week CAR cycle followed a structured sequence. Initially, in the planning phase, teachers planned an assessment task (e.g., a paragraph writing prompt, an oral presentation rubric) concentrated on a particular language aspect (e.g., grammatical accuracy for writing, fluency for speaking). They then directed this task to their students. Second, in the acting/observing phase, teachers collected student responses and submitted written samples to an Automated Writing Evaluation (AWE) system for grammar/mechanics scores and error tagging, and/or used an NLP dashboard to analyze spoken/written output for lexical diversity and syntactic complexity. They also implemented their own initial scoring. Third, in the reflecting phase, teachers brought AI analytics and their own judgments to bi-weekly CAR meetings. Here, they engaged in structured peer feedback, relating AI data with human observations, analytically examining discrepancies (e.g., AI penalizing culturally Persian rhetorical structures), and co-interpreting results. A mentor facilitated these discussions, connecting findings to LAL principles. Lastly, teachers revised their assessment task or feedback strategy for the next cycle, completing a reflective journal entry on the process. Participants and Setting The research included 15 Iranian language teachers, purposefully selected from both public secondary schools and private language institutes in Tehran, Karaj, and Ray to capture a variety of instructional contexts. The group involved 8 women and 7 men, aged between 25 and 45 (M=33). They had three to fifteen (M=9) years of classroom experience. Preliminary survey results showed that all participants have gained minimal formal training in LAL. Eligibility for inclusion required that participants were presently instructional intermediate-level language classes and were open to taking part in CAR cycles. Teaching environments in these institutions normally included crowded classrooms, averaging 25 to 35 learners, working under Iran’s national curriculum, where assessment practices are deeply shaped by high-stakes standardized examinations. Selected schools have at least minimal infrastructure to support AI-based educational tools, including computer labs with internet access. Still, technological capacity varies notably: urban public schools may have interactive whiteboards and a small number of tablets, while private institutes are more likely to use cloud-based learning platforms. The study also identifies environmental and systemic limitations, containing occasional electricity shortages, unstable internet connections, and, in some public schools, administrative limitations on digital tool use. Over the course of the six-month program, teachers took part in bi-weekly collaborative meetings at a university laboratory in Tehran. This space functioned as an essential site for mentor-led professional development workshops and as a controlled site for regular AI tool training. In this way, the project context mirrors the broader dynamics of Iran’s language sector, where enthusiasm for incorporating educational technology interconnects with continuous structural challenges in teacher training. Data collection This study selected a comprehensive, multi-layered qualitative data collection strategy to probe the multifaceted ecosystem in which teachers’ LAL develops when AI and human feedback were joined. Across three iterative six-week CAR cycles, data was collected by means of an intersected set of complementary methods intended to capture both fine-grained insights and wider contextual patterns. Semi-structured interviews worked as the fundamental means of recording participants’ perspectives. Each of the fifteen teachers had three sequential interview sessions for one hour, each with a specific focus. The first, conducted at the project’s beginning, followed a narrative inquiry format to investigate teachers’ personal histories in assessment practice, their beliefs about evaluation, and their confidence in integrating technology. Open-ended prompts boosted in-depth, reflective storytelling. The mid-cycle interviews included a stimulated recall approach in which participants involve with their own AI-generated assessment reports in real time, exchanging their thought processes as they evaluate technological recommendations against their professional judgment. This method let researchers to observe the instant cognitive and instructional negotiations taking place. Final interviews focused on the critical incident technique to determine decisive moments of change in participants’ LAL. These transformative episodes were associated with detailed phases of the CAR process and important instances of AI interaction. All interviews were implemented in English, since all participants selected to talk in English. Classroom observations were followed a multi-perspective ethnographic design, incorporating detailed qualitative accounts. Observers kept wide field notes supporting environmental conditions, unintended interruptions, and nuanced social interactions that numerical data alone could not fully represent. A selected sample of lessons were video-recorded from several angles to allow close analysis of nonverbal behaviors, mainly how instructors engaged physically with AI interfaces throughout immediate assessment tasks, as well as students’ reactions to technology-mediated feedback. These recordings were paired with instant post-lesson, video-stimulated recall sessions, in which teachers reviewed significant segments and comment on the reasoning behind their assessment decisions. Furthermore, digital screen-capture software recorded teachers’ exploration of AI tools during lesson preparation and assessment design, generating a complete record of collaboration sequences, points of uncertainty, and workflow modifications. Teachers’ reflective practice were documented by means of a secure, custom-built digital journaling platform that mixed structured prompts with open-ended writing opportunities. Weekly questions paid attention to important features of assessment practice development, for example problems in interpreting AI-generated feedback, emotional reactions to automated recommendations, and classroom variations made to integrate such input. The platform let multimedia uploads, empowering participants to contain annotated screenshots of AI dashboards, unstructured audio reflections, and classroom video excerpts of assessment episodes they wish to discuss. Each journal entry automatically noted metadata, containing time spent writing, revision activity, and modifications in response length across the study. To reflect the collaborative ethos of the CAR process, the system contained a social annotation tool, letting peers and researchers to pose questions, provide insights, or offer supportive comments, generating a collaborative reflective environment in line with the project’s stress on collaborative professional development. The study made numerous streams of data from AI–teacher collaborations, capturing detailed system logs of all feedback offered to participants, records of examples where teachers adjusted or overrode automated recommendations, following changes in assessment quality over time. Real-time NLP applied to teachers’ written feedback on student work, monitoring shifts in language complexity, sentiment, and error detection patterns affected by AI consultation. Machine learning analyses monitored the advancement of assessment designs in response to algorithmic input, creating visual trajectory maps that demonstrate developmental developments. These metrics were collected into modified dashboards accessible to both the research team and the teachers, supporting reflective, data-informed discussions during CAR sessions and guaranteeing transparency in how AI-generated analyses are produced. During the project, both physical and digital artifacts were systematically gathered. These involved all versions of assessment instruments, from primary drafts to final classroom-ready versions, anonymized samples of student work annotated by both teachers and the AI, audio recordings of peer feedback interactions throughout CAR meetings, and photographs of collaborative outputs including whiteboard-generated designs. Each artifact classified integrating a detailed metadata framework that notes creation date, contributing teacher(s), and links to significant CAR cycle stages. Special stress was placed on capturing classroom-specific modifications of AI-generated suggestions, offering concrete evidence of how general technological outputs are localized for practical instructional contexts. A robust quality assurance framework strengthened the whole process. This included a triangulation matrix mapping each research question to its supporting evidence sources, arranged member-checking sessions where participants validate primary interpretations, and systematic explore case analysis to improve developing explanations. Additional validity safeguards included intercoder reliability testing for observation data, verification of translation accuracy for interview transcripts, and technical audits of AI-generated metrics. The phased nature of the CAR process allowed for continuous improvement of data collection strategies, with each cycle’s findings guiding modifications to following protocols in response to emergent themes or unexpected patterns. Ethical safeguards involved multi-step consent processes that give participants the option to decline particular data-gathering methods without withdrawing from the study, advanced anonymization procedures keeping both teacher and student identities, and secure storage systems that meet international data protection standards for educational research. This layered, contextually grounded approach guarantees that the research captures not only quantifiable modifications in AI-assisted LAL development, but also the detailed, situated decision-making through which Iranian EFL teachers reconcile technological affordances with instructional expertise. This study obtained approval from Farhangian University Ethics Review Board. All participants offered written informed consent, with clear explanations about data use, anonymization, and their right to withdraw. Student data processed by AI tools were anonymized earlier to analysis. Data Analysis The study adopted a layered qualitative analysis strategy, incorporating inductive thematic investigation with deductive framework to explore how AI-generated feedback and collaborative human reflection cooperate in facilitating teachers’ LAL. All textual materials, containing interview transcripts, reflective journal entries, classroom observation notes, and CAR meeting records, were processed in inductively via an iterative coding cycle. This process started with open coding to superficial developing concepts, progress to axial coding to recognize interconnections among categories, and conclude with selective coding to distill central themes. Data from AI-produced assessment analytics were examined through a content analysis to underscore recurring diagnostic suggestions and their correspondence to established LAL competencies. A hybrid coding scheme were created, integrating inductively produced codes with predetermined categories drawn from LAL scholarship, mainly Inbar-Lourie’s (2013) framework for assessment knowledge. This let for comprehensive investigation of how teachers balanced AI input with the teaching necessities of their specific instructional settings. Video footage of classroom practice were subjected to collaboration analysis to identify decision-making episodes influenced by the AI–human feedback system. Triangulation was accomplished via continuous comparative analysis across the full range of data sources, with attention given to examples where teachers’ self-reported accounts differ from observed behaviors. To boost trustworthiness, member-checking sessions were implemented at two stages of the analysis, and explore case analysis were integrated to test and refine developing interpretations. The interpretive synthesis was culminating in a grounded conceptual model illustrating the processes and conditions through which AI-assisted LAL development occurs, placing the findings within the realities of Iran’s EFL sector while also contributing to international discussions on technology-mediated professional learning for teachers. Analysis followed a systematic hybrid approach. First, all qualitative data were openly coded inductively. Second, these codes were deductively mapped onto a priori categories derived from the LAL framework (assessment design, interpretation, ethics) and the CAR cycles (planning, acting, observing, reflecting). This generated an analysis matrix. Triangulation was achieved by associating evidence for each theme across data types (e.g., a teacher’s claim about by means of AI data in an interview was checked against their journal entries and observed classroom practice). The coding tree was developed iteratively across three CAR cycles. Third, coded data were synthesized to produce the interpretive themes presented in the findings. Findings The analysis showed that the incorporation of AI-generated feedback within a CAR framework influenced Iranian EFL teachers’ LAL in multidimensional ways. Across the data, we recognized eight interconnected themes demonstrating how technological insights, when critically mediated through collaborative inquiry, renovated teachers’ diagnostic skills, professional identity, assessment practices, and collaborative capacity. Theme 1: Boosted Diagnostic Skills Through AI Feedback Teachers constantly stated that AI’s collected, visualized feedback empowered them to identify patterns they had formerly ignored. For instance, one participant clarified: “ the AI analytics dashboard indicated me that a lot of my students were making regular errors in by means of definite articles with abstract nouns, something I’d never detected in 12 years of instruction. (Teacher 5-T5)” This pattern recognition progressed their assessment beyond intuition toward evidence-driven diagnosis, openly supporting the diagnostic aspect of LAL. The AI also improved how teachers classified language aspects, offering a level of detail that is often lost from traditional classroom assessments. In one meeting of the CAR group, a teacher detected: “ We used to label these as ‘tense mistakes,’ but in reality, most were problems with feature—understanding this changed our remediation strategies. (T13)” In this role, the AI operated like a fine-focus lens, helping elucidate concepts and allowing more precise instructional modifications. Furthermore, the capacity to monitor learning over long periods improved teachers’ skills in tracking student development. In her reflective journal, one teacher noticed: “ From the AI’s development summaries, I recognized grammar accuracy increase by 35% across the term, however discourse marker use augmented by only 8%—this insight restructured my teaching priorities. (T9)” By offering a clear picture of adjustment over time, the AI empowered educators to spot weaker areas and shift their instructional efforts where they were most required. Theme 2: Mediation of Feedback Through CAR The CAR process was crucial in grounding AI-generated feedback within real-life classroom contexts. Throughout one session, teachers underscored: “ The AI noticed more than half of students for ‘overusing simple sentences,’ nevertheless we acknowledged this as a feature of Persian rhetorical tradition—so we planned a rubric that stimulated syntactic variety while honoring cultural style. (T12)” This kind of reinterpretation lies at the heart of LAL, where evaluations are directed not just by linguistic benchmarks but by cultural sensitivity as well. Participants also became skilled at noticing and counteracting algorithmic bias. T1 reminded: “ We noted that the AI commonly assigned lower scores to work containing Persian cultural references—so we presented a ‘cultural relevance’ category to counterbalance this. ” Such measures extended their understanding of fairness in assessment, allowing for contextually suitable scoring without weakening analytical accuracy. Over time, the AI–CAR partnership led to collectively built rubrics personalized to the local setting, combining automated and human evaluation. As recorded in one CAR document, “ Our final framework allocated 30% to AI-generated grammar scores, 40% to human-judged content quality, and 30% to cultural appropriateness. (T6)” This incorporation of quantitative precision with qualitative judgment reflects an assessment philosophy where technology supports, instead of substitutes, professional capability. Also, this process demonstrates a mediational model where AI output is not an endpoint but a catalyst for professional discourse, leading to co-constructed, contextually-valid assessment criteria. Theme 3: Transformation of Assessment Practices The combination of AI-generated feedback and CAR-based reflection moved teachers away from an only summative mindset toward a formative approach. As one teacher clarified: “ Instead of waiting for end-of-term exams, I now obtain weekly AI updates that identify which topics need revisiting. (T11)” This modification resonated the principles of assessment for learning, embedding feedback into the continuous instructional process rather than treating it as a final verdict. The technology also encouraged experimentation with alternative assessment methods. Observation notes documented how T15 “ substituted approximately half of conventional tests with AI-assisted portfolio tasks, which encompassed automated grammar checks on early drafts. ” These changes developed the range of evidence used to evaluate learning and supported continuous enhancement of student work. Making the feedback process more clear to learners further improved motivation. In a focus group, a teacher commented, “ When I saw students’ recurring errors in the AI report, I knew exactly where they need to focus their efforts—it was more useful than just receiving a score. ” This blended model supported students’ own assessment literacy, preparing them to interpret results and take informed action. Theme 4: Professional Identity Development Working within the AI–CAR framework reformed how teachers viewed their own professional autonomy. One teacher reflected: “ I used to rely completely on ministry-prepared tests. But, now, I produce my own assessments and feel like an expert in the process. (T4)” This move from simply administering tests to actively influencing assessment practices signaled a move toward leadership in evaluation. The incorporation of data into decision-making also supported teachers’ confidence. In one CAR meeting, a participant reported: “ When my grading was questioned by the principal, I presented the AI analytics supporting my choices—before this, I never had such evidence to protect my decisions. (T10)” In this case, AI worked as a tool for professional support, supporting teachers’ authority within their institutions. Continuous participation led some teachers to develop niche expertise. As one post in a professional forum stated: “ After reviewing and interpreting 200 AI-scored essays, I’ve become the district’s point person for writing assessment design. (T2)” This kind of acknowledgement marks an innovative stage of LAL, where skills are not only developed but also respected and shared across professional networks. Theme 5: Collaborative Learning Dynamics Within the CAR framework, teachers of different experience levels involved in genuine two-way learning. One long-serving teacher reflected: “ The newer colleagues taught me how to interpret AI-generated data, whereas I guided them in understanding the cultural background of our students. (T14)” This exchange intertwined AI combination into a bigger web of professional knowledge. Collaborative efforts also created noticeable institutional resources. Among them was a mutual repository containing 150 validated test items, each annotated with notes on common errors flagged by the AI. Such joint materials strengthened institutional assessment literacy and guaranteed that innovations could sustain beyond a single teacher’s practice. The spirit of collaboration moved beyond individual schools. A post on a provincial teachers’ blog stated: “ Our WhatsApp group now circulates AI analysis templates modified for seven diverse regional dialects. (T7)” This example demonstrates how AI–CAR approaches can scale through peer-led networks, spreading context-sensitive practices across wider instructional communities. Theme 6: Tensions and Negotiations The blended AI–CAR approach also brought essential tensions to the surface. In one meeting, a teacher remembered: “ When the AI scored a paper at 65 out of 100 and I gave it 85, we had to discuss closely why our evaluations varied—three hours of intense discussion followed. (T3)” Such moments motivated the group to examine the validity and reliability of their scoring practices in detail. Cultural values sometimes conflicted with automated judgments. As one participant described: “ The AI lowered scores for what it called ‘direct criticism’ in essays, however in our context, respectful challenge is a sign of strong reasoning—so we had to modify our framework. (T1)” These exchanges emphasized the necessity of adjusting LAL practices to local norms. Adoption was not without limitation. One teacher’s journal designed the first term as “ exhausting—mastering the AI system while contributing to CAR added roughly 20 extra hours a month, but now it saves me time. (T6)” This reflects a mutual reality in capacity building: an initial period of heavy investment before long-term efficiency gains are recognized. Theme 7: Systemic and Contextual Factors Application was influenced by the realities of local infrastructure. One rural account underlined that, with just two working computers serving 300 pupils, staff made a paper-based system to monitor errors flagged by AI. This workaround demonstrates how emerging LAL in resource-limited contexts often is contingent on blending digital and non-digital tools. Resistance at the policy level also appeared. Initially, the principal feared that AI tools might encourage teachers to cut corners, but those doubts faded after having a 15% improvement in students’ exam scores. Such shifts in institutional perception are vital for long-term adoption. Community perspectives influenced development as well. Records from parent meetings describe early “ skepticism toward automated scoring ,” which lessened after families saw that teachers persisted central to evaluation. This highlights the social dimension of LAL, where building and preserving stakeholder trust is indispensable. Synthesis: Mechanisms of LAL Development Across these themes, the hybrid AI–CAR approach influenced LAL in numerous, interconnected ways. AI tools lengthened teachers’ analytical reach, letting them to examine student work in greater depth, while the CAR process helped them interpret these insights more thoughtfully. At the same time, reflection and dialogue empowered educators to bridge global assessment standards with the nuances of local culture and language, generating a contextually grounded practice. Evidence-based decision-making supported teachers’ confidence and leadership, enabling them to take ownership of their professional development. Finally, the collaborative process fostered networks of peers, sustaining innovations and empowering successful practices to spread across schools and communities. Discussion The findings of this study underscore the transformative potential of incorporating AI-generated feedback within a CAR framework to improve Iranian EFL teachers’ LAL. The findings of this study underlines the transformative potential of integrating AI-produced feedback within a CAR framework to develop Iranian EFL teachers’ LAL. These findings can be interpreted through the incorporated sociocultural and reflective practitioner lens outlined in our theoretical framework. The AI tools functioned as powerful mediating artifacts, objectively developing patterns in student language (e.g., systematic article errors) that extended teachers’ diagnostic reach—an important aspect of LAL. Yet, consistent with Vygotskian principles, this mediation required social scaffolding. The CAR community offered the zone of proximal development where teachers, through dialogue, learned to critically interpret algorithmic feedback, contextualize it within Persian rhetorical norms, and plan instructional responses. This process of collaborative sensemaking transformed the AI’s monologic output into dialogic, formative assessment practices. The hybrid AI–human feedback model not only fostered teachers’ diagnostic precision and assessment practices but also cultivated professional identity development, collaborative learning, and context-sensitive adaptations. The hybrid AI–human feedback model not only developed teachers’ diagnostic precision and assessment practices but also nurtured professional identity development, collaborative learning, and context-sensitive adaptations. These results in line with recent research highlighting the role of AI in increasing teacher decision-making while underlining the necessity of human mediation to guarantee culturally responsive assessment (Zhang et al., 2024; Kooli & Yusuf, 2025 ). The study’s results challenge the idea that AI alone can produce significant professional development, as an alternative supporting a blended approach where technology functions as a catalyst for reflective conversation rather than a replacement for instructional expertise (Mohammed & ‘Nell’Watson, 2019 ; Maurya et al., 2025 ). A key outcome of this research lies in indicating how incorporating AI-informed feedback into CAR can help close the gap between broad, standardized evaluation measures and the particular demands of local instructional contexts. Instead of simply accepting the AI’s recommendations, the teachers, participated in collaborative reflection, critically examined the AI’s suggestions and reshaped them to fit the social and cultural traditions and the needs of their students. This supports Lam’s (2025) position that LAL development must be grounded in sociocultural realities to circumvent the pitfalls of decontextualized, excessively technical assessment models. Similarly, the study underlines the role of educators as active interpreters of AI input, an idea aligns with recent work on “augmented intelligence” in education, which highlights that human expertise should persist at the core of fair and responsible assessment (Grab, 2025 ; Trajkovski & Hayes, 2025 ). The reported shift toward formative practices support Schön’s reflection-on-action. The repetitive cycle of executing an assessment, gaining instant AI data, and collectively reflecting on its implications in CAR meetings generated structured opportunities for teachers to question the dominance of summative testing. They started to internalize the value of continuous evidence collecting, going from seeing assessment as a final judgment to viewing it as a feedback tool for regulation—both for students and their own instruction. This research also contributes to a profounder understanding of how CAR can be improved through the strategic implementation of AI. Conventional CAR approaches usually hinge on planned peer reviews for feedback, proposing little in the way of instantaneous, data-informed understandings (Schildkamp, 2019 ). In contrast, the hybrid model probed here equipped teachers with immediate analytic feedback, letting them to improve assessment practices in shorter, more responsive cycles. In doing so, it addresses a limitation recognized by Yosief et al. ( 2024 ), who observed that the lag in feedback common to CAR can slow necessary teaching modifications. At the same time, the findings bring to light continuous challenges, including augmented workload and the risk of algorithmic bias, that mirror wider discussions about the complexities of integrating AI into teacher training (Erdem Coşgun, 2025 ; Ghimire, 2025 ). These issues underline the need for robust institutional support and gradual application to support permanent, effective incorporation. Particularly, the study’s findings challenge the notion that resource-limited contexts like Iran are ill-suited for AI-driven professional development. Although infrastructure limitations continued, teachers proved significant flexibility, blending digital and non-digital tools to maximize AI’s utility. This resilience supports Ghourchian’s ( 2024 ) observations about Iranian educators’ ability for innovation despite systemic limitations. Furthermore, the study’s evidence of peer-led scaling, where teachers distributed AI-modified assessment practices across schools, supports recent calls for decentralized, teacher-owned models of professional learning (Dulfer et al., 2024 ; Kim, 2024 ). The relevance of these findings reaches beyond the Iranian context, providing a transferable model for fostering LAL in comparable contexts. By framing AI as a partner in the assessment process instead of a top-down authority on quality, the approach sidesteps ethical concerns focusing on excessive dependence on automation (Malik et al., 2025 ). Upcoming work should explore its long-term influence, mainly whether the combination of AI and CAR produces long-lasting gains in student learning. More inquiry into systemic and policy limitations, including centralized testing commands that restrict the uptake of formative assessment, could shed added light on the model’s flexibility (Au, 2022 ; Mirsanjari, 2025 ). This study was implemented in particular Iranian urban/semi-urban contexts; its findings may not transfer to rural contexts with severe technological limitations. The six-month timeframe also confines claims about long-term sustainability. Future research should explore longitudinal effects on student learning outcomes and probe the application of this hybrid model in other national settings with different policy limitations. Eventually, this research illustrates that incorporating AI-driven feedback with CAR can meaningfully reinforce LAL by encouraging assessment practices that are reflective, evidence-informed, and culturally responsive. It makes the case for professional learning designs that integrate technology not as a stand-alone solution, but as a catalyst for deepening teachers’ analytical engagement with assessment, an approach that supports broader international efforts to guarantee AI functions human-centered goals in education (Zhang et al., 2025 ; Topping et al., 2025 ). As education systems around the world probe both the promise and the complexity of AI adoption, the framework suggested here provides a compelling route for enabling educators as active interpreters and co-creators of assessment innovation. Conclusion This study shows that a hybrid AI–human feedback model, rooted within a CAR framework, considerably influences the development of LAL among Iranian EFL school teachers. The findings offer a direct answer to the research question by indicating that this influence functions through a process of augmented professional judgment. Teachers leveraged AI-driven analytics to prolong their diagnostic reach, recognizing nuanced patterns in student performance, while the structured collaboration of CAR offered the crucial space to critically interpret, culturally contextualize, and instructionally adapt these algorithmic insights. This synergy nurtured a supportable shift from a summative to a formative assessment mindset, improved teachers’ confidence and autonomy in assessment design, and nurtured collaborative networks for professional innovation. The main contribution of this research is a replicable, context-sensitive model for LAL development in resource-constrained school contexts. It extends beyond presenting AI as an automated solution by providing a real framework where technology works as a catalyst for deep, collaborative professional inquiry. The model bridges a critical gap in the literature by placing AI-mediated teacher development explicitly within the school-level context and offering empirical evidence on how teachers can become active interpreters—not just passive implementers—of assessment technology. Future research should construct directly upon these findings. First, longitudinal studies are needed to examine the long-term influence of this hybrid model on student learning outcomes and the durability of teachers’ assessment innovations. Second, exploration is needed into the systemic and policy barriers—mainly rigid, centralized testing regimes—that may restrict the scalability of such formative, teacher-led assessment approaches. In conclusion, research should concentrate on the design principles for AI tools that are more transparent and customizable, thus better supporting teachers’ agency and contextual negotiation in the assessment process. Abbreviations Collaborative Action Research (CAR) English Language Teaching (ELT) Language Assessment Literacy (LAL) Natural Language Processing (NLP) Declarations Ethics Approval and Consent to Participate This study was reviewed and approved by the appropriate institutional ethics committee. All participants provided informed written consent prior to participation. Consent for Publication All participants consented to the publication of anonymized data and findings derived from this study. Funding This research received no external funding. Author Contribution Arefian did all stages individually. Acknowledgement Arefian did all stages individually. References Abrar-ul-Hassan, S., & Nassaji, H. (2024). Rescoping language assessment literacy: An expanded perspective. System , 120 , 103195. https://doi.org/10.1016/j.system.2023.103195 Aliakbari, M., & Sadeghi, S. (2022). The professional identity of the Iranian teachers: A case of professional practices. Teacher Development , 26 (3), 411–431. https://doi.org/10.1080/13664530.2022.2076729 Aliakbari, M., Yasini, A., & Sadeghi, S. (2023). Iranian EFL Teachers' Classroom Assessment Practices: Discrepancy between Theory and Practice. International Journal of Language Testing , 13 (2), 149–169. Asad, M. M., Naz, A., Churi, P., & Tahanzadeh, M. M. (2021). Virtual reality as pedagogical tool to enhance experiential learning: a systematic literature review. Education Research International , 2021 (1), 7061623. https://doi.org/10.1155/2021/7061623 Au, W. (2022). Unequal by design: High-stakes testing and the standardization of inequality . Routledge. https://doi.org/10.4324/9781003005179 Bazvand, A. D., Jalilzadeh, K., Farhady, H., & Sabzehparvar, A. (2025). Language assessment literacy of professional language testers: Implications for teacher education reform. System , 134 , 103792. https://doi.org/10.1016/j.system.2025.103792 Benade, L. (2015). Teachers’ critical reflective practice in the context of twenty-first century learning. Open Review of Educational Research , 2 (1), 42–54. https://doi.org/10.1080/23265507.2014.998159 Bhimdiwala, A., Neri, R. C., & Gomez, L. M. (2022). Advancing the design and implementation of artificial intelligence in education through continuous improvement. International Journal of Artificial Intelligence in Education , 32 (3), 756–782. https://doi.org/10.1007/s40593-021-00278-8 Bleicher, R. E. (2014). A collaborative action research approach to professional learning. Professional development in education , 40 (5), 802–821. https://doi.org/10.1080/19415257.2013.842183 Chang, D. Y. S., Lin, M. H., & Lee, J. Y. (2024). Exploring language assessment literacy and needs of English teachers at senior high school level. Asia Pacific Journal of Education , 44 (4), 854–872. https://doi.org/10.1080/02188791.2024.2313486 Clark, I. (2012). Formative assessment: Assessment is for self-regulated learning. Educational psychology review , 24 (2), 205–249. https://doi.org/10.1007/s10648-011-9191-6 Coombe, C., Vafadar, H., & Mohebbi, H. (2020). Language assessment literacy: What do we need to learn, unlearn, and relearn? Language Testing in Asia , 10 (1), 3. https://doi.org/10.1186/s40468-020-00101-6 Cope, B., Kalantzis, M., & Searsmith, D. (2021). Artificial intelligence for education: Knowledge and its assessment in AI-enabled learning ecologies. Educational philosophy and theory , 53 (12), 1229–1245. https://doi.org/10.1080/00131857.2020.1728732 Cui, Y., Liu, Y., & Yu, H. (2025). Enhancing English teachers’ language assessment literacy through a professional learning community: A collaborative autoethnography study. System , 103799. https://doi.org/10.1016/j.system.2025.103799 Curry, M. W., & Athanases, S. Z. (2020). In pursuit of engaged learning with Latinx students: Expanding learning beyond classrooms through performance-based engagements. Teachers College Record , 122 (8), 1–49. https://doi.org/10.1177/016146812012200815 Derakhshan, A., & Ghiasvand, F. (2022). Demystifying Iranian EFL teachers’ perceptions and practices of learning-oriented assessment (LOA): Challenges and prospects in focus. Language Testing in Asia , 12 (1), 55. https://doi.org/10.1186/s40468-022-00204-2 Dulfer, N., Kriewaldt, J., & McKernan, A. (2024). Using collaborative action research to enhance differentiated instruction. International Journal of Inclusive Education , 28 (8), 1402–1416. https://doi.org/10.1080/13603116.2021.1992678 Erdem Coşgun, G. (2025). Artificial intelligence literacy in assessment: Empowering pre‐service teachers to design effective exam questions for language learning. British Educational Research Journal . https://doi.org/10.1002/berj.4177 Firoozi, T., Razavipour, K., & Ahmadi, A. (2019). The language assessment literacy needs of Iranian EFL teachers with a focus on reformed assessment policies. Language Testing in Asia , 9 (1), 2. https://doi.org/10.1186/s40468-019-0078-7 Gan, L., & Lam, R. (2022). A review on language assessment literacy: Trends, foci and contributions. Language Assessment Quarterly , 19 (5), 503–525. https://doi.org/10.1080/15434303.2022.2128802 Ghimire, A. (2025). Utilizing ChatGPT to integrate world English and diverse knowledge: A transnational perspective in critical artificial intelligence (AI) literacy. Computers and Composition, 75, 102913. https://doi.org/10.1016/j.compcom.2024.102913 Ghourchian, P. (2024). Online education development in iran during the covid-19 pandemic: Opportunities and challenges. Online Education During COVID-19 and Beyond: Opportunities, Challenges and Outlook , 341–364. https://doi.org/10.1007/978-3-031- 49353-9_18 Grab, M. O. (2025). Teaching for Equity: An Exploration of AI's Role in Culturally Responsive Teaching in Higher Education Settings. Innovative Higher Education , 1–22. https://doi.org/10.1007/s10755-025-09801-4 Inbar-Lourie, O. (2013). Guest editorial to the special issue on language assessment literacy. Language Testing , 30 (3), 301–307. https://doi.org/10.1177/0265532213480126 Inbar-Lourie, O. (2016). Language assessment literacy. In Language testing and assessment (pp. 1–14). Springer, Cham. https://doi.org/10.1007/978-3-319-02326-7_19-1 Kim, J. (2024). Leading teachers' perspective on teacher-AI collaboration in education. Education and information technologies , 29 (7), 8693–8724. https://doi.org/10.1007/s10639-023- 12109-5 Koh, K. H. (2011). Improving teachers’ assessment literacy through professional development. Teaching Education , 22 (3), 255–276. https://doi.org/10.1080/10476210.2011.593164 Kooli, C., & Yusuf, N. (2025). Transforming educational assessment: Insights into the use of ChatGPT and large language models in grading. International Journal of Human– Computer Interaction , 41 (5), 3388-3399. https://doi.org/10.1080/10447318.2024.2338330 Lam, R. (2015). Language assessment training in Hong Kong: Implications for language assessment literacy. Language Testing , 32 (2), 169–197. https://doi.org/10.1177/0265532214554321 Mirsanjari, Z. (2025). Enhancing assessment literacy in EAP instruction: the role of teacher development courses in overcoming systemic barriers. Language Testing in Asia , 15 (1), 30. https://doi.org/10.1186/s40468-025-00368-7 Levi, T., & Inbar-Lourie, O. (2020). Assessment literacy or language assessment literacy: Learning from the teachers. Language Assessment Quarterly , 17 (2), 168–182. https://doi.org/10.1080/15434303.2019.1692347 Malik, A., Khan, M. L., Hussain, K., Qadir, J., & Tarhini, A. (2025). AI in higher education: unveiling academicians’ perspectives on teaching, research, and ethics in the age of ChatGPT. Interactive Learning Environments , 33 (3), 2390–2406. https://doi.org/10.1080/10494820.2024.2409407 Manfra, M. M. (2019). Action research and systematic, intentional change in teaching practice. Review of research in education , 43 (1), 163–196. https://doi.org/10.3102/0091732X18821132 Maurya, R. K., Montesinos, S., Bogomaz, M., & DeDiego, A. C. (2025). Assessing the use of ChatGPT as a psychoeducational tool for mental health practice. Counselling and Psychotherapy Research , 25 (1), e12759. https://doi.org/10.1002/capr.12759 Mohammed, P. S., & ‘Nell’Watson, E. (2019). Towards inclusive education in the age of artificial intelligence: Perspectives, challenges, and opportunities. Artificial Intelligence and Inclusive Education: Speculative futures and emerging practices , 17–37. https://doi.org/10.1007/978-981-13-8161-4_2 Morris, R., Perry, T., & Wardle, L. (2021). Formative assessment and feedback for learning in higher education: A systematic review. Review of Education , 9 (3), e3292. https://doi.org/10.1002/rev3.3292 Mun, R. U., Ezzani, M. D., & Lee, L. E. (2020). Culturally relevant leadership in gifted education: A systematic literature review. Journal for the Education of the Gifted , 43 (2), 108–142. https://doi.org/10.1177/0162353220912009 Qing-Li, H., Torres, M. N., & Shi-Ji, F. (2019). Collaborative action research for preparing teachers as reflective practitioners. Systemic Practice and Action Research , 32 (4), 411– 427. https://doi.org/10.1007/s11213-018-9461-z Schildkamp, K. (2019). Data-based decision-making for school improvement: Research insights and gaps. Educational research , 61 (3), 257–273. https://doi.org/10.1080/00131881.2019.1625716 Smith, K. (2010). Assessing the Practicum in teacher education–Do we want candidates and mentors to agree?. Studies in educational evaluation , 36 (1-2), 36–41. https://doi.org/10.1016/j.stueduc.2010.08.001 Sultana, N. (2019). Language assessment literacy: An uncharted area for the English language teachers in Bangladesh. Language Testing in Asia , 9 (1), 1. https://doi.org/10.1186/s40468- 019-0077-8 Topping, K. J., Gehringer, E., Khosravi, H., Gudipati, S., Jadhav, K., & Susarla, S. (2025). Enhancing peer assessment with artificial intelligence. International Journal of Educational Technology in Higher Education , 22 (1), 3. https://doi.org/10.1186/s41239- 024-00501-1 Trajkovski, G., & Hayes, H. (2025). AI-Assisted Formative Assessment and Feedback. In AI- Assisted Assessment in Education: Transforming Assessment and Measuring Learning (pp. 283–312). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-88252- 4_7 Yan, L., Suleman Abdullah Alwabel, A., & Mohamad, U. H. (2025). AI‐Powered Education: Transforming Teacher‐Student Interactions and Advancing Sustainable Learning Practices. European Journal of Education , 60 (4), e70351. https://doi.org/10.1111/ejed.70351 Yin, X., & Buck, G. A. (2019). Using a collaborative action research approach to negotiate an understanding of formative assessment in an era of accountability testing. Teaching and Teacher Education , 80 , 27–38. https://doi.org/10.1016/j.tate.2018.12.018 Yosief, A., Sulieman, M. S., & Biede, T. (2024). Improving the practices of teacher educators through collaborative action research: Challenges and hopes. Educational Action Research , 32 (2), 204–221. https://doi.org/10.1080/09650792.2022.2066147 Zhang, Y., Zhang, M., Wu, L., & Li, J. (2025). Digital Transition Framework for Higher Education in AI-Assisted Engineering Teaching: Challenge, Strategy, and Initiatives in China. Science & Education , 34 (2), 933–954. https://doi.org/10.1007/s11191-024-00575- 3 Zhao, D. (2024). The impact of AI-enhanced natural language processing tools on writing proficiency: An analysis of language precision, content summarization, and creative writing facilitation. Education and Information Technologies , 1–32. https://doi.org/10.1007/s10639-024-13145-5 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 21 May, 2026 Reviewers agreed at journal 14 Apr, 2026 Reviews received at journal 31 Dec, 2025 Reviewers agreed at journal 18 Dec, 2025 Reviewers invited by journal 17 Dec, 2025 Editor assigned by journal 09 Dec, 2025 Submission checks completed at journal 09 Dec, 2025 First submitted to journal 04 Dec, 2025 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-8281675","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":563297133,"identity":"fb990305-aa32-4b44-9ed5-2de38373c79c","order_by":0,"name":"Mohammad Hossein Arefian","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA00lEQVRIiWNgGAWjYJCCDwkgkr2BgZlYHYwzwFp4DpCiBUxJJBCphX/24YMND2ruMOjOfGP4uaDChoG/vTsBrxaJc2mJDQnHnjGY3c4xlp5xJo1B4szZDfitOcNj/iCB7TBIi4E0b9thBgOJXPxa5M/wf2xI+AfUcvOM8W+itBic4WFsSASqNLvBY0acLYZn2AwbEvsO85idSSuz5jmTxkPQL3JnmB82/vh2WM7s+OHNt3kqbOT423sJeB8KeBgYOAygDOIB+wNSVI+CUTAKRsEIAgCRcEgVLPmeIAAAAABJRU5ErkJggg==","orcid":"","institution":"Imam Khomeini International University","correspondingAuthor":true,"prefix":"","firstName":"Mohammad","middleName":"Hossein","lastName":"Arefian","suffix":""}],"badges":[],"createdAt":"2025-12-04 17:38:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8281675/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8281675/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98737686,"identity":"4da31672-eebf-437b-bcc7-a6027bc0eb96","added_by":"auto","created_at":"2025-12-22 06:57:34","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":60761,"visible":true,"origin":"","legend":"","description":"","filename":"ManuscriptAIandassessmentliteracy.docx","url":"https://assets-eu.researchsquare.com/files/rs-8281675/v1/ef6f4523dd7c413a4aea852e.docx"},{"id":98737687,"identity":"825ad4de-f459-448e-9497-8764fa50ab79","added_by":"auto","created_at":"2025-12-22 06:57:34","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3702,"visible":true,"origin":"","legend":"","description":"","filename":"a26ca9e129124b96b79094c671e23e08.json","url":"https://assets-eu.researchsquare.com/files/rs-8281675/v1/d321da34d114838eb7328c6c.json"},{"id":98777035,"identity":"b802c42e-f96b-44ae-8369-4f529266dae3","added_by":"auto","created_at":"2025-12-22 12:25:09","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":134227,"visible":true,"origin":"","legend":"","description":"","filename":"a26ca9e129124b96b79094c671e23e081enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8281675/v1/b71d766f0f857aa57379839c.xml"},{"id":98737691,"identity":"94f39993-2a42-43e9-9cd0-e7ee3a8180c1","added_by":"auto","created_at":"2025-12-22 06:58:10","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":134312,"visible":true,"origin":"","legend":"","description":"","filename":"a26ca9e129124b96b79094c671e23e081structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8281675/v1/1e39ce1613a4d1fcd1997f80.xml"},{"id":98737688,"identity":"f9b6da56-28e7-4bf5-b53c-0d6e1f4d1c4e","added_by":"auto","created_at":"2025-12-22 06:57:34","extension":"html","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":142820,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8281675/v1/9bb06a1c658a35840f6fed59.html"},{"id":98783276,"identity":"c42901a4-97c0-491b-85c8-e840be5048b5","added_by":"auto","created_at":"2025-12-22 12:41:39","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":776611,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8281675/v1/477444d8-24d7-43de-b862-fb84a9b51c25.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Supporting Iranian Teachers’ Language Assessment Literacy: A Hybrid AI–Human Feedback Approach within Collaborative Action Research","fulltext":[{"header":"Introduction","content":"\u003cp\u003eLanguage assessment literacy (LAL) has appeared as a significant aspect of language education, underscoring the knowledge, skills, and principles needed for professional and effective assessment practices (Levi \u0026amp; Inbar-Lourie, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Within Iran\u0026rsquo;s English language teaching (ELT) context, enhancing teachers\u0026rsquo; LAL is chiefly important because of the increasing request for high-stakes language assessments and the need for instructional alignment with international standards (Chang et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Nevertheless, Iranian teachers regularly face issues in conducting rigorous assessment strategies because of restricted professional development opportunities, dependence on traditional assessment methods, and insufficient involvement with existing testing methods and techniques (Derakhshan \u0026amp; Ghiasvand, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This problem is specifically important in the Iranian school context (secondary and language institute levels), where large class sizes, a centralized national curriculum, and pressure from university entrance exams further limit teachers\u0026rsquo; capacity to creative in assessment.\u003c/p\u003e \u003cp\u003eIn spite of recognized needs, conventional top-down training workshops have established ineffective for supportable LAL development in Iran (Bazvand et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). A persistent gap exists in models that offer continuous, contextualized support. This study thus introduces a novel incorporated approach, stating that the problem of underdeveloped LAL needs a situated, cyclical, and data-fostered professional learning model. It explores whether and how a hybrid AI and human feedback approach, rooted within a collaborative action research (CAR) framework, can address this problem by offering teachers with iterative cycles of evidence-based reflection and collaborative sensemaking. By incorporating AI with organized human feedback, the study intends to generate a dynamic, collaborative, and reflective professional development model that enables teachers to cultivate their assessment practices and literacies systematically and continuously (Bhimdiwala et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). CAR is selected as the essential methodological framework because it goes beyond one-off training by integrating professional development within teachers\u0026rsquo; own contexts. It offers a structured yet flexible process for teachers to collectively explore and modify their own assessment practices, generating it ideal for nurturing maintainable, context-sensitive LAL development.\u003c/p\u003e \u003cp\u003eThe concept of LAL has advanced considerably during the past two decades, with scholars highlighting its multidimensional nature. Inbar-Lourie (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) describes LAL as the capacity to design, conduct, and evaluate language assessments while understanding their pedagogical and ethical implications. For teachers, LAL goes beyond mere assessment production; it includes interpreting assessment data, offering constructive feedback, and adjusting teaching according to learner needs and assessment results (Levi \u0026amp; Inbar-Lourie, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). In spite of its acknowledged significance, LAL remains underdeveloped in many educational contexts, mainly in districts where teacher training programs focus on content knowledge over assessment literacy (Sultana, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In Iran, where established educational policies regularly dictate assessment practices, teachers often use summative assessments with minimal formative or alternative assessment incorporation (Firoozi et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). This dependence on traditional testing methods confines opportunities for meaningful learning feedback and fails to resolve different learner competencies (Clark, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2012\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eCAR has been encouraged as an efficient professional development strategy to fill theory-practice gaps in education through collaboration and action research (Qing-Li et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). By including teachers in systematic inquiry into their own practices, CAR enhances reflective instruction and enables educators to foster contextually appropriate solutions (Manfra, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In the territory of assessment, CAR offers a planned but flexible framework for teachers to try novel assessment strategies, evaluate outcomes, and improve their assessment approaches continuously, systematically, and collectively (Yosief et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Yet, former CAR models often necessitate personalized, instant, data-driven feedback mechanisms, supporting the continuous learning process (Schildkamp, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). This study resolves this restriction by integrating AI-driven analytics combined with human mentorship, allowing real-time assessment feedback and nurturing deeper teacher engagement with LAL principles.\u003c/p\u003e \u003cp\u003eAI has transformed educational assessment by providing tailored feedback, automated scoring, and evidence-based understandings into learner performance (Kooli \u0026amp; Yusuf, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). AI-powered tools, including natural language processing (NLP) systems, can assess both written and spoken communication with remarkable precision, letting teachers to gain immediate insights into student performance (Zhao, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). When incorporated into teacher professional development, AI can function as a supportive aid, recognizing patterns in assessment, uncovering potential biases, and proposing research-backed recommendations for improvement (Maurya et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Though, AI alone cannot substitute the comprehensive judgment of human educators, mainly in contexts necessitating cultural sensitivity and instructional flexibility (Mohammed \u0026amp; \u0026lsquo;Nell\u0026rsquo;Watson, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). A hybrid approach that includes AI efficiency with human expertise guarantees that teachers not only gain technical feedback but also participate in critical discussions about assessment validity, fairness, and educational influence (Zhang et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study places itself where LAL, CAR, and AI-enhanced feedback intersect, suggesting a novel model for Iranian teacher development. It uses AI systems to provide instant assessment insights, which are then incorporated with planned feedback from peers and mentors. Together, these features aim to nurture an ongoing and sustainable model of professional development. The blended method is particularly appreciated in the Iranian context, where restricted resources and overcrowded classrooms often make personalized teacher support difficult to accomplish (Ghourchian, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). By means of continuous CAR cycles, teachers will design, implement, and analyze assessments while getting AI-generated data and human insights. This dual feedback mechanism is anticipated to improve teachers\u0026rsquo; confidence in assessment design, develop their ability to interpret student performance, and support the implementation of formative assessments along with summative evaluations (Morris et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe importance of this study rests in its potential to reframe LAL development in resource-limited contexts by using technology-driven and collaborative learning approaches. Previous research on LAL in Iran has generally focused on workshops and training programs, which often fail to produce lasting effects (Bazvand et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Unlike earlier approaches, this study incorporates professional development openly into teachers\u0026rsquo; everyday routines, encouraging continuous involvement and practical significance. Moreover, while AI applications in education have been widely studied for student learning, its potential to improve teacher assessment literacy is still chiefly unnoticed, particularly in non-Western contexts (Ghimire, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). By filling this gap, the study contributes to worldwide discussions on AI\u0026rsquo;s role in teacher education while offering a replicable model for similar contexts. The upcoming sections will present the theoretical underpinnings of LAL, CAR, and AI in assessment, followed by the study\u0026rsquo;s methodology and findings. This research aims to indicate that merging AI and human feedback within the CAR framework can meaningfully foster Iranian teachers\u0026rsquo; LAL, nurturing assessment practices that are impartial, more accurate, and educationally effective. The following research question is pursued throughout the study:\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e \u003cb\u003eHow does a hybrid AI\u0026ndash;human feedback model, combined within a CAR, support the development of LAL among Iranian English language teachers?\u003c/b\u003e \u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"Literature review","content":"\u003cp\u003eThe concept of LAL has been gaining considerable attention in language education, with researchers underlining its vigorous influence on the development of effective instruction and learning strategies (Gan \u0026amp; Lam, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). LAL encompasses the knowledge and abilities necessary to prepare, implement, and assess language tests, along with an awareness of their broader educational impact (Bazvand et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Over time, this definition has extended to incorporate not only technical expertise but also ethical responsibility, principles of fairness, and the effective use of assessment results to guide instructional practices (Coombe et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The multidimensional nature of LAL underlines its significance in enhancing equitable and valid assessment practices, mainly in contexts where traditional testing methods dominate (Abrar-ul-Hassan \u0026amp; Nassaji, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In Iran, the dependence on summative assessments and centralized examination systems provide limited opportunities for teachers to engage with alternative assessment methods, including formative assessments, portfolios, and dynamic assessment strategies (Mirsanjari, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This gap underscores the urgent need for professional development initiatives that foster teachers\u0026rsquo; LAL, empowering them to adopt more different and instructionally sound assessment practices (Cui et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Lam, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Nevertheless, a significant gap exists in the literature on LAL development specially within school-level contexts, with much current research concentrated on higher education. This study aims to address this gap by situating its inquiry explicitly in the secondary school and language institute context.\u003c/p\u003e \u003cp\u003eThe challenges of fostering LAL among Iranian English language teachers are multidimensional, coming from systemic, instructional, and training-related limitations (Asad et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Researchers discovered that many Iranian teachers are not introduced to modern assessment theories throughout their pre-service training, which leads to a heavy reliance on high-stakes exams and rote learning assessments (Aliakbari et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This problem is compounded by the inadequacy of in-service professional development opportunities devoted to increasing LAL (Koh, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Moreover, the cultural and institutional attention to standardized testing as a major measure of student achievement discourages teachers to try advanced assessment techniques (Au, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Although such challenges are not exclusive to Iran, similar problems have been observed in other educational settings where teacher preparation programs place little stress on assessment literacy (Firoozi et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Given that the Iranian context poses its own specific problems, existing limitations including restricted resources, overcrowded classrooms, and fixed curricular structures make it particularly difficult to embrace formative or performance-based assessment practices (Curry et al., 2020). Addressing these issues need a personalized approach that combines theoretical knowledge with practical, contextually relevant strategies for assessment reform.\u003c/p\u003e \u003cp\u003eCAR has been increasingly acknowledged as an appreciated model for enhancing teachers\u0026rsquo; professional development, particularly in developing LAL (Dulfer et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). According to Author (X), CAR includes a joint, participatory inquiry in which teachers examine their own classroom practices, evaluate the outcomes, and make ongoing modifications. Engrained in the ethos of reflective practice, this process allows teachers to take greater responsibility for their development while directly confronting authentic classroom issues (Benade, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Within the framework of LAL, CAR provides teachers a structured yet flexible space to experiment novel assessment strategies, examine learners\u0026rsquo; responses, and adjust their practices using evidence from the classroom. Accordingly, CAR can support teachers\u0026rsquo; understanding of assessment by grounding professional learning in real instructional contexts (Yosief et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). However, traditional CAR approaches often fall short in offering instant feedback, which can slow the process of refining assessment methods (Yin \u0026amp; Buck, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). This gap focuses on the value of integrating technology-supported feedback tools, including AI-powered analytics, to work alongside human collaboration, empowering faster, more informed modifications.\u003c/p\u003e \u003cp\u003eThe application of AI in assessment moves beyond automated scoring. Present models leverage NLP to offer diagnostic feedback on linguistic features (e.g., syntactic complexity, lexical diversity) and machine learning to recognize patterns in student performance (Zhao, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In formative assessment, AI can support feedback processes as conceptualized by scholars such as Carless and Dawson (2022), who highlight feedback as a dialogic process directed at closing the learning gap. Nevertheless, AI-produced feedback risks being monologic and decontextualized. A hybrid model, thus, must incorporate AI\u0026rsquo;s analytical capacity with human mediation to guarantee feedback is usable, timely, and embedded within the particular instructional context\u0026mdash;a core challenge this study explores.\u003c/p\u003e \u003cp\u003eThe use of AI in educational assessment marks a significant transformation in the way teachers design, administer, and analyze their assessment practices. Tools driven by AI, for example automated essay evaluation systems and NLP applications, can deliver instant feedback on learner performance. In doing so, they lighten teachers\u0026rsquo; administrative workload while providing precise, in-depth diagnostic information. Such tools hold specific value in large-scale assessment settings, where traditional grading demands major time and is susceptible to unpredictability. Along with improving efficiency, AI has the potential to reinforce assessment validity and reliability by finding trends in learner responses, flagging possible biases, and suggesting modifications grounded in empirical evidence (Topping et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Trajkovski and Hayes (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) demonstrate how AI can foster formative assessment by generating personalized feedback for individual learners and providing teachers practical, data-informed recommendations. Yet, researchers warn against placing excessive reliance on AI, emphasizing that human judgment remains essential for exploring the nuanced teaching and ethical issues essential in assessment (Malik et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). A joined model, blending the efficiency of AI with the interpretive and contextual expertise of teachers, helps guarantee that assessment practices are both culturally responsive and instructionally robust (Grab, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe merging of AI and teacher professional development provides a promising, though still relatively unexplored, pathway for fostering LAL. Whereas existing scholarship on AI in education has generally focused on its impact on student learning, comparatively little attention has been given to its capability to reinforce teachers\u0026rsquo; assessment expertise (Cope et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). This research gap is particularly obvious in non-Western contexts, where access to technology and changing levels of digital competence can influence the uptake of AI-based tools. In Iran, for instance, teacher education programs often lack continuous professional development, making AI-supported feedback systems a possibly scalable and cost-efficient resource for improving LAL (Erdem Coşgun, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Such systems could review teachers\u0026rsquo; assessment designs, recognize possible validity concerns, and suggest alternative approaches, while human mentors lead critical dialogues about the teaching significance of these recommendations (Smith, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). This two-tiered feedback process supports sociocultural perspectives on learning, which stress the role of social interaction and scaffolded support in professional development (Mun et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Incorporating AI into a CAR model would let teachers to engage in continuous cycles of exploration, reflection, and refinement, effectively narrowing the gap between assessment theory and classroom practice.\u003c/p\u003e \u003cp\u003eCurrent research has begun to explore how AI can supplement collaborative learning frameworks within teacher education. For example, Yan et al. (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) illustrates that AI-enabled professional learning networks let teachers to share assessment techniques, provide constructive feedback, and collectively build knowledge within virtual communities of practice. Similarly, Kim (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) highlight AI\u0026rsquo;s capacity to deliver timely feedback during CAR cycles, helping teachers modify their assessment methods based on data insights. These studies show that joining AI-driven and human feedback can advance both the efficiency and impact of LAL development, while maintaining the reflective and cooperative nature of CAR. However, applying these hybrid models successfully needs overcoming obstacles for example teachers\u0026rsquo; confidence with technology, institutional backing, and guaranteeing that AI tools are compatible with local assessment traditions (Zhang et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Specially, in the Iranian setting, where concerns like digital inequality and resistance to innovation exist, a gradual and inclusive strategy for incorporating AI into teacher professional development is vital (Ghourchian, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe theoretical and empirical evidence presented in this review highlights the capacity of incorporating AI-driven and human feedback within CAR to foster LAL among English language teachers in Iran. By drawing together perspectives from research on LAL, CAR practices, and AI applications in education, this study presents a novel framework personalized to address the specific challenges confronted in the Iranian context while also contributing to broader discussions on teacher professional development. The following section will detail the methodological approach for applying and assessing this model.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eTheoretical Framework\u003c/h2\u003e \u003cp\u003eThis study is established in an incorporation of sociocultural and reflective practitioner paradigms. Mainly, it focuses on Vygotskian sociocultural theory, wherein tools mediate higher-order psychological processes (Vygotsky, 1978). Here, AI acts as a mediating artifact that moves teachers\u0026rsquo; analytical capabilities, offering external, data-driven insights into student performance. Yet, learning happens through social interaction; thus, the CAR framework establishes a community of practice (Lave \u0026amp; Wenger, 1991) where teachers collectively interpret (\u0026lsquo;make sense of\u0026rsquo;) AI outputs, negotiating meaning and appropriating assessment concepts. This process exemplifies situated learning, where LAL development is entrenched in authentic assessment activities and social dialogue.\u003c/p\u003e \u003cp\u003eMoreover, the cyclical CAR process operationalizes Sch\u0026ouml;n\u0026rsquo;s (1983) model of the reflective practitioner. The AI-produced data function as a catalyst for reflection-in-action (e.g., real-time adjustment during a feedback session) and reflection-on-action (e.g., post-lesson analysis in CAR meetings). Through iterative cycles of planning, action, observation, and reflection, teachers engage in continuous knowledge construction, internalizing new assessment practices. Lastly, this interplay between technology, pedagogy, and content is practically examined through the TPACK framework (Mishra \u0026amp; Koehler, 2006). The study probes how CAR facilitates the iterative refinement of teachers\u0026rsquo; Technological Pedagogical Content Knowledge (TPACK) by necessitating them to critically incorporate AI tools (TK) into language assessment design (PK/CK) within their particular EFL context.\u003c/p\u003e \u003c/div\u003e"},{"header":"Methodology","content":"\u003cp\u003e\u003cstrong\u003eResearch Design\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research selects a qualitative CAR approach, following the iterative process of planning, acting, observing, and reflecting collaboratively within a community of practice (Bleicher, 2014). The aim of this study to explore how joining AI-based and human feedback can support Iranian English language teachers\u0026rsquo; LAL. CAR functions both as the guiding methodology and the planned design, underlining professional development that is collaborative, contextually grounded, and responsive to participants\u0026rsquo; needs (Bleicher, 2014). To obtain a richer perspective, the project integrates case study techniques (Aliakbari \u0026amp; Sadeghi, 2022) to monitor the professional development of individual teachers, as well as ethnographic observations to place the results within the cultural and institutional context of Iran\u0026rsquo;s educational system. Over a six-month period, 15 language teachers from both public and private institutions, who were selected through purposive sampling, participated in three CAR cycles. Each cycle included the construction, implementation, and evaluation of assessment tasks, maintained by AI tools (including automated essay scoring systems and NLP-based feedback platforms) in line with structured feedback from peers and mentors. By integrating CAR with technological tools, this design intends to diminish the gap between theoretical knowledge and practical application in LAL, while also providing a scalable professional development model suitable to resource-constrained settings.\u003c/p\u003e\n\u003cp\u003eThe hybrid model consisted of a sequenced process. First, teachers established\u0026nbsp;AI-produced feedback\u0026nbsp;on their assessment tasks and student responses by means of two main tools: (1) an automated writing evaluation (AWE) system offering scores and error tags (grammar, mechanics), and (2) an NLP dashboard analyzing lexical diversity and syntactic complexity in student output. Second, this data functioned as input for\u0026nbsp;structured human feedback\u0026nbsp;phases: (a)\u0026nbsp;peer feedback\u0026nbsp;in CAR meetings, where teachers compared AI outputs with their own judgments, and (b)\u0026nbsp;mentor feedback, where a researcher enhanced discussions on supporting AI diagnostics with curricular goals and cultural appropriateness. Criteria for use concentrated on AI for recognizing\u0026nbsp;patterns\u0026nbsp;and human collaboration for\u0026nbsp;interpretation\u0026nbsp;and\u0026nbsp;instructional action.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntervention Procedures: Teacher Activities and AI Tools\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEach six-week CAR cycle followed a structured sequence. Initially, in the planning phase, teachers planned an assessment task (e.g., a paragraph writing prompt, an oral presentation rubric) concentrated on a particular language aspect (e.g., grammatical accuracy for writing, fluency for speaking). They then directed this task to their students. Second, in the acting/observing phase, teachers collected student responses and submitted written samples to an Automated Writing Evaluation (AWE) system for grammar/mechanics scores and error tagging, and/or used an NLP dashboard to analyze spoken/written output for lexical diversity and syntactic complexity. They also implemented their own initial scoring. Third, in the \u003cem\u003ereflecting\u003c/em\u003e phase, teachers brought AI analytics and their own judgments to bi-weekly CAR meetings. Here, they engaged in structured peer feedback, relating AI data with human observations, analytically examining discrepancies (e.g., AI penalizing culturally Persian rhetorical structures), and co-interpreting results. A mentor facilitated these discussions, connecting findings to LAL principles. Lastly, teachers revised their assessment task or feedback strategy for the next cycle, completing a reflective journal entry on the process.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eParticipants and Setting\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe research included 15 Iranian language teachers, purposefully selected from both public secondary schools and private language institutes in Tehran, Karaj, and Ray to capture a variety of instructional contexts. The group involved 8 women and 7 men, aged between 25 and 45 (M=33). They had three to fifteen (M=9) years of classroom experience. Preliminary survey results showed that all participants have gained minimal formal training in LAL. Eligibility for inclusion required that participants were presently instructional intermediate-level language classes and were open to taking part in CAR cycles.\u003c/p\u003e\n\u003cp\u003eTeaching environments in these institutions normally included crowded classrooms, averaging 25 to 35 learners, working under Iran\u0026rsquo;s national curriculum, where assessment practices are deeply shaped by high-stakes standardized examinations. Selected schools have at least minimal infrastructure to support AI-based educational tools, including computer labs with internet access. Still, technological capacity varies notably: urban public schools may have interactive whiteboards and a small number of tablets, while private institutes are more likely to use cloud-based learning platforms.\u003c/p\u003e\n\u003cp\u003eThe study also identifies environmental and systemic limitations, containing occasional electricity shortages, unstable internet connections, and, in some public schools, administrative limitations on digital tool use. Over the course of the six-month program, teachers took part in bi-weekly collaborative meetings at a university laboratory in Tehran. This space functioned as an essential site for mentor-led professional development workshops and as a controlled site for regular AI tool training. In this way, the project context mirrors the broader dynamics of Iran\u0026rsquo;s language sector, where enthusiasm for incorporating educational technology interconnects with continuous structural challenges in teacher training.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study selected a comprehensive, multi-layered qualitative data collection strategy to probe the multifaceted ecosystem in which teachers\u0026rsquo; LAL develops when AI and human feedback were joined. Across three iterative six-week CAR cycles, data was collected by means of an intersected set of complementary methods intended to capture both fine-grained insights and wider contextual patterns. Semi-structured interviews worked as the fundamental means of recording participants\u0026rsquo; perspectives. Each of the fifteen teachers had three sequential interview sessions for one hour, each with a specific focus. The first, conducted at the project\u0026rsquo;s beginning, followed a narrative inquiry format to investigate teachers\u0026rsquo; personal histories in assessment practice, their beliefs about evaluation, and their confidence in integrating technology. Open-ended prompts boosted in-depth, reflective storytelling.\u003c/p\u003e\n\u003cp\u003eThe mid-cycle interviews included a stimulated recall approach in which participants involve with their own AI-generated assessment reports in real time, exchanging their thought processes as they evaluate technological recommendations against their professional judgment. This method let researchers to observe the instant cognitive and instructional negotiations taking place. Final interviews focused on the critical incident technique to determine decisive moments of change in participants\u0026rsquo; LAL. These transformative episodes were associated with detailed phases of the CAR process and important instances of AI interaction.\u003c/p\u003e\n\u003cp\u003eAll interviews were implemented in English, since all participants selected to talk in English. Classroom observations were followed a multi-perspective ethnographic design, incorporating detailed qualitative accounts. Observers kept wide field notes supporting environmental conditions, unintended interruptions, and nuanced social interactions that numerical data alone could not fully represent. A selected sample of lessons were video-recorded from several angles to allow close analysis of nonverbal behaviors, mainly how instructors engaged physically with AI interfaces throughout immediate assessment tasks, as well as students\u0026rsquo; reactions to technology-mediated feedback. These recordings were paired with instant post-lesson, video-stimulated recall sessions, in which teachers reviewed significant segments and comment on the reasoning behind their assessment decisions.\u003c/p\u003e\n\u003cp\u003eFurthermore, digital screen-capture software recorded teachers\u0026rsquo; exploration of AI tools during lesson preparation and assessment design, generating a complete record of collaboration sequences, points of uncertainty, and workflow modifications. Teachers\u0026rsquo; reflective practice were documented by means of a secure, custom-built digital journaling platform that mixed structured prompts with open-ended writing opportunities. Weekly questions paid attention to important features of assessment practice development, for example problems in interpreting AI-generated feedback, emotional reactions to automated recommendations, and classroom variations made to integrate such input. The platform let multimedia uploads, empowering participants to contain annotated screenshots of AI dashboards, unstructured audio reflections, and classroom video excerpts of assessment episodes they wish to discuss.\u003c/p\u003e\n\u003cp\u003eEach journal entry automatically noted metadata, containing time spent writing, revision activity, and modifications in response length across the study. To reflect the collaborative ethos of the CAR process, the system contained a social annotation tool, letting peers and researchers to pose questions, provide insights, or offer supportive comments, generating a collaborative reflective environment in line with the project\u0026rsquo;s stress on collaborative professional development. The study made numerous streams of data from AI\u0026ndash;teacher collaborations, capturing detailed system logs of all feedback offered to participants, records of examples where teachers adjusted or overrode automated recommendations, following changes in assessment quality over time. Real-time NLP applied to teachers\u0026rsquo; written feedback on student work, monitoring shifts in language complexity, sentiment, and error detection patterns affected by AI consultation. Machine learning analyses monitored the advancement of assessment designs in response to algorithmic input, creating visual trajectory maps that demonstrate developmental developments. These metrics were collected into modified dashboards accessible to both the research team and the teachers, supporting reflective, data-informed discussions during CAR sessions and guaranteeing transparency in how AI-generated analyses are produced.\u003c/p\u003e\n\u003cp\u003eDuring the project, both physical and digital artifacts were systematically gathered. These involved all versions of assessment instruments, from primary drafts to final classroom-ready versions, anonymized samples of student work annotated by both teachers and the AI, audio recordings of peer feedback interactions throughout CAR meetings, and photographs of collaborative outputs including whiteboard-generated designs. Each artifact classified integrating a detailed metadata framework that notes creation date, contributing teacher(s), and links to significant CAR cycle stages. Special stress was placed on capturing classroom-specific modifications of AI-generated suggestions, offering concrete evidence of how general technological outputs are localized for practical instructional contexts.\u003c/p\u003e\n\u003cp\u003eA robust quality assurance framework strengthened the whole process. This included a triangulation matrix mapping each research question to its supporting evidence sources, arranged member-checking sessions where participants validate primary interpretations, and systematic explore case analysis to improve developing explanations. Additional validity safeguards included intercoder reliability testing for observation data, verification of translation accuracy for interview transcripts, and technical audits of AI-generated metrics. The phased nature of the CAR process allowed for continuous improvement of data collection strategies, with each cycle\u0026rsquo;s findings guiding modifications to following protocols in response to emergent themes or unexpected patterns. Ethical safeguards involved multi-step consent processes that give participants the option to decline particular data-gathering methods without withdrawing from the study, advanced anonymization procedures keeping both teacher and student identities, and secure storage systems that meet international data protection standards for educational research. This layered, contextually grounded approach guarantees that the research captures not only quantifiable modifications in AI-assisted LAL development, but also the detailed, situated decision-making through which Iranian EFL teachers reconcile technological affordances with instructional expertise.\u003c/p\u003e\n\u003cp\u003eThis study obtained approval from Farhangian University Ethics Review Board. All participants offered written informed consent, with clear explanations about data use, anonymization, and their right to withdraw. Student data processed by AI tools were anonymized earlier to analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study adopted a layered qualitative analysis strategy, incorporating inductive thematic investigation with deductive framework to explore how AI-generated feedback and collaborative human reflection cooperate in facilitating teachers\u0026rsquo; LAL. All textual materials, containing interview transcripts, reflective journal entries, classroom observation notes, and CAR meeting records, were processed in inductively via an iterative coding cycle. This process started with open coding to superficial developing concepts, progress to axial coding to recognize interconnections among categories, and conclude with selective coding to distill central themes. Data from AI-produced assessment analytics were examined through a content analysis to underscore recurring diagnostic suggestions and their correspondence to established LAL competencies. A hybrid coding scheme were created, integrating inductively produced codes with predetermined categories drawn from LAL scholarship, mainly Inbar-Lourie\u0026rsquo;s (2013) framework for assessment knowledge. This let for comprehensive investigation of how teachers balanced AI input with the teaching necessities of their specific instructional settings.\u003c/p\u003e\n\u003cp\u003eVideo footage of classroom practice were subjected to collaboration analysis to identify decision-making episodes influenced by the AI\u0026ndash;human feedback system. Triangulation was accomplished via continuous comparative analysis across the full range of data sources, with attention given to examples where teachers\u0026rsquo; self-reported accounts differ from observed behaviors. To boost trustworthiness, member-checking sessions were implemented at two stages of the analysis, and explore case analysis were integrated to test and refine developing interpretations. The interpretive synthesis was culminating in a grounded conceptual model illustrating the processes and conditions through which AI-assisted LAL development occurs, placing the findings within the realities of Iran\u0026rsquo;s EFL sector while also contributing to international discussions on technology-mediated professional learning for teachers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAnalysis followed a systematic hybrid approach. First, all qualitative data were openly coded inductively. Second, these codes were deductively mapped onto a priori categories derived from the LAL framework (assessment design, interpretation, ethics) and the CAR cycles (planning, acting, observing, reflecting). This generated an analysis matrix. Triangulation was achieved by associating evidence for each theme across data types (e.g., a teacher\u0026rsquo;s claim about by means of AI data in an interview was checked against their journal entries and observed classroom practice). The coding tree was developed iteratively across three CAR cycles. Third, coded data were synthesized to produce the interpretive themes presented in the findings.\u003c/p\u003e"},{"header":"Findings","content":"\u003cp\u003eThe analysis showed that the incorporation of AI-generated feedback within a CAR framework influenced Iranian EFL teachers\u0026rsquo; LAL in multidimensional ways. Across the data, we recognized eight interconnected themes demonstrating how technological insights, when critically mediated through collaborative inquiry, renovated teachers\u0026rsquo; diagnostic skills, professional identity, assessment practices, and collaborative capacity.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eTheme 1: Boosted Diagnostic Skills Through AI Feedback\u003c/h2\u003e \u003cp\u003eTeachers constantly stated that AI\u0026rsquo;s collected, visualized feedback empowered them to identify patterns they had formerly ignored. For instance, one participant clarified: \u0026ldquo;\u003cem\u003ethe AI analytics dashboard indicated me that a lot of my students were making regular errors in by means of definite articles with abstract nouns, something I\u0026rsquo;d never detected in 12 years of instruction.\u003c/em\u003e (Teacher 5-T5)\u0026rdquo; This pattern recognition progressed their assessment beyond intuition toward evidence-driven diagnosis, openly supporting the diagnostic aspect of LAL.\u003c/p\u003e \u003cp\u003eThe AI also improved how teachers classified language aspects, offering a level of detail that is often lost from traditional classroom assessments. In one meeting of the CAR group, a teacher detected: \u0026ldquo;\u003cem\u003eWe used to label these as \u0026lsquo;tense mistakes,\u0026rsquo; but in reality, most were problems with feature\u0026mdash;understanding this changed our remediation strategies.\u003c/em\u003e (T13)\u0026rdquo; In this role, the AI operated like a fine-focus lens, helping elucidate concepts and allowing more precise instructional modifications. Furthermore, the capacity to monitor learning over long periods improved teachers\u0026rsquo; skills in tracking student development. In her reflective journal, one teacher noticed: \u0026ldquo;\u003cem\u003eFrom the AI\u0026rsquo;s development summaries, I recognized grammar accuracy increase by 35% across the term, however discourse marker use augmented by only 8%\u0026mdash;this insight restructured my teaching priorities.\u003c/em\u003e (T9)\u0026rdquo; By offering a clear picture of adjustment over time, the AI empowered educators to spot weaker areas and shift their instructional efforts where they were most required.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eTheme 2: Mediation of Feedback Through CAR\u003c/h2\u003e \u003cp\u003eThe CAR process was crucial in grounding AI-generated feedback within real-life classroom contexts. Throughout one session, teachers underscored: \u0026ldquo;\u003cem\u003eThe AI noticed more than half of students for \u0026lsquo;overusing simple sentences,\u0026rsquo; nevertheless we acknowledged this as a feature of Persian rhetorical tradition\u0026mdash;so we planned a rubric that stimulated syntactic variety while honoring cultural style.\u003c/em\u003e (T12)\u0026rdquo; This kind of reinterpretation lies at the heart of LAL, where evaluations are directed not just by linguistic benchmarks but by cultural sensitivity as well. Participants also became skilled at noticing and counteracting algorithmic bias. T1 reminded: \u0026ldquo;\u003cem\u003eWe noted that the AI commonly assigned lower scores to work containing Persian cultural references\u0026mdash;so we presented a \u0026lsquo;cultural relevance\u0026rsquo; category to counterbalance this.\u003c/em\u003e\u0026rdquo; Such measures extended their understanding of fairness in assessment, allowing for contextually suitable scoring without weakening analytical accuracy.\u003c/p\u003e \u003cp\u003eOver time, the AI\u0026ndash;CAR partnership led to collectively built rubrics personalized to the local setting, combining automated and human evaluation. As recorded in one CAR document, \u0026ldquo;\u003cem\u003eOur final framework allocated 30% to AI-generated grammar scores, 40% to human-judged content quality, and 30% to cultural appropriateness.\u003c/em\u003e (T6)\u0026rdquo; This incorporation of quantitative precision with qualitative judgment reflects an assessment philosophy where technology supports, instead of substitutes, professional capability. Also, this process demonstrates a mediational model where AI output is not an endpoint but a catalyst for professional discourse, leading to co-constructed, contextually-valid assessment criteria.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eTheme 3: Transformation of Assessment Practices\u003c/h2\u003e \u003cp\u003eThe combination of AI-generated feedback and CAR-based reflection moved teachers away from an only summative mindset toward a formative approach. As one teacher clarified: \u0026ldquo;\u003cem\u003eInstead of waiting for end-of-term exams, I now obtain weekly AI updates that identify which topics need revisiting.\u003c/em\u003e (T11)\u0026rdquo; This modification resonated the principles of assessment for learning, embedding feedback into the continuous instructional process rather than treating it as a final verdict. The technology also encouraged experimentation with alternative assessment methods. Observation notes documented how T15 \u0026ldquo;\u003cem\u003esubstituted approximately half of conventional tests with AI-assisted portfolio tasks, which encompassed automated grammar checks on early drafts.\u003c/em\u003e\u0026rdquo; These changes developed the range of evidence used to evaluate learning and supported continuous enhancement of student work. Making the feedback process more clear to learners further improved motivation. In a focus group, a teacher commented, \u0026ldquo;\u003cem\u003eWhen I saw students\u0026rsquo; recurring errors in the AI report, I knew exactly where they need to focus their efforts\u0026mdash;it was more useful than just receiving a score.\u003c/em\u003e\u0026rdquo; This blended model supported students\u0026rsquo; own assessment literacy, preparing them to interpret results and take informed action.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eTheme 4: Professional Identity Development\u003c/h2\u003e \u003cp\u003eWorking within the AI\u0026ndash;CAR framework reformed how teachers viewed their own professional autonomy. One teacher reflected: \u0026ldquo;\u003cem\u003eI used to rely completely on ministry-prepared tests. But, now, I produce my own assessments and feel like an expert in the process.\u003c/em\u003e (T4)\u0026rdquo; This move from simply administering tests to actively influencing assessment practices signaled a move toward leadership in evaluation. The incorporation of data into decision-making also supported teachers\u0026rsquo; confidence. In one CAR meeting, a participant reported: \u0026ldquo;\u003cem\u003eWhen my grading was questioned by the principal, I presented the AI analytics supporting my choices\u0026mdash;before this, I never had such evidence to protect my decisions.\u003c/em\u003e (T10)\u0026rdquo; In this case, AI worked as a tool for professional support, supporting teachers\u0026rsquo; authority within their institutions. Continuous participation led some teachers to develop niche expertise. As one post in a professional forum stated: \u0026ldquo;\u003cem\u003eAfter reviewing and interpreting 200 AI-scored essays, I\u0026rsquo;ve become the district\u0026rsquo;s point person for writing assessment design.\u003c/em\u003e (T2)\u0026rdquo; This kind of acknowledgement marks an innovative stage of LAL, where skills are not only developed but also respected and shared across professional networks.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eTheme 5: Collaborative Learning Dynamics\u003c/h2\u003e \u003cp\u003eWithin the CAR framework, teachers of different experience levels involved in genuine two-way learning. One long-serving teacher reflected: \u0026ldquo;\u003cem\u003eThe newer colleagues taught me how to interpret AI-generated data, whereas I guided them in understanding the cultural background of our students.\u003c/em\u003e (T14)\u0026rdquo; This exchange intertwined AI combination into a bigger web of professional knowledge. Collaborative efforts also created noticeable institutional resources. Among them was a mutual repository containing 150 validated test items, each annotated with notes on common errors flagged by the AI. Such joint materials strengthened institutional assessment literacy and guaranteed that innovations could sustain beyond a single teacher\u0026rsquo;s practice. The spirit of collaboration moved beyond individual schools. A post on a provincial teachers\u0026rsquo; blog stated: \u0026ldquo;\u003cem\u003eOur WhatsApp group now circulates AI analysis templates modified for seven diverse regional dialects.\u003c/em\u003e (T7)\u0026rdquo; This example demonstrates how AI\u0026ndash;CAR approaches can scale through peer-led networks, spreading context-sensitive practices across wider instructional communities.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eTheme 6: Tensions and Negotiations\u003c/h2\u003e \u003cp\u003eThe blended AI\u0026ndash;CAR approach also brought essential tensions to the surface. In one meeting, a teacher remembered: \u0026ldquo;\u003cem\u003eWhen the AI scored a paper at 65 out of 100 and I gave it 85, we had to discuss closely why our evaluations varied\u0026mdash;three hours of intense discussion followed.\u003c/em\u003e (T3)\u0026rdquo; Such moments motivated the group to examine the validity and reliability of their scoring practices in detail. Cultural values sometimes conflicted with automated judgments. As one participant described: \u0026ldquo;\u003cem\u003eThe AI lowered scores for what it called \u0026lsquo;direct criticism\u0026rsquo; in essays, however in our context, respectful challenge is a sign of strong reasoning\u0026mdash;so we had to modify our framework.\u003c/em\u003e (T1)\u0026rdquo; These exchanges emphasized the necessity of adjusting LAL practices to local norms. Adoption was not without limitation. One teacher\u0026rsquo;s journal designed the first term as \u0026ldquo;\u003cem\u003eexhausting\u0026mdash;mastering the AI system while contributing to CAR added roughly 20 extra hours a month, but now it saves me time.\u003c/em\u003e (T6)\u0026rdquo; This reflects a mutual reality in capacity building: an initial period of heavy investment before long-term efficiency gains are recognized.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eTheme 7: Systemic and Contextual Factors\u003c/h2\u003e \u003cp\u003eApplication was influenced by the realities of local infrastructure. One rural account underlined that, with just two working computers serving 300 pupils, staff made a paper-based system to monitor errors flagged by AI. This workaround demonstrates how emerging LAL in resource-limited contexts often is contingent on blending digital and non-digital tools. Resistance at the policy level also appeared. Initially, the principal feared that AI tools might encourage teachers to cut corners, but those doubts faded after having a 15% improvement in students\u0026rsquo; exam scores. Such shifts in institutional perception are vital for long-term adoption. Community perspectives influenced development as well. Records from parent meetings describe early \u0026ldquo;\u003cem\u003eskepticism toward automated scoring\u003c/em\u003e,\u0026rdquo; which lessened after families saw that teachers persisted central to evaluation. This highlights the social dimension of LAL, where building and preserving stakeholder trust is indispensable.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eSynthesis: Mechanisms of LAL Development\u003c/h2\u003e \u003cp\u003eAcross these themes, the hybrid AI\u0026ndash;CAR approach influenced LAL in numerous, interconnected ways. AI tools lengthened teachers\u0026rsquo; analytical reach, letting them to examine student work in greater depth, while the CAR process helped them interpret these insights more thoughtfully. At the same time, reflection and dialogue empowered educators to bridge global assessment standards with the nuances of local culture and language, generating a contextually grounded practice. Evidence-based decision-making supported teachers\u0026rsquo; confidence and leadership, enabling them to take ownership of their professional development. Finally, the collaborative process fostered networks of peers, sustaining innovations and empowering successful practices to spread across schools and communities.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe findings of this study underscore the transformative potential of incorporating AI-generated feedback within a CAR framework to improve Iranian EFL teachers\u0026rsquo; LAL. The findings of this study underlines the transformative potential of integrating AI-produced feedback within a CAR framework to develop Iranian EFL teachers\u0026rsquo; LAL. These findings can be interpreted through the incorporated sociocultural and reflective practitioner lens outlined in our theoretical framework. The AI tools functioned as powerful mediating artifacts, objectively developing patterns in student language (e.g., systematic article errors) that extended teachers\u0026rsquo; diagnostic reach\u0026mdash;an important aspect of LAL. Yet, consistent with Vygotskian principles, this mediation required social scaffolding. The CAR community offered the zone of proximal development where teachers, through dialogue, learned to critically interpret algorithmic feedback, contextualize it within Persian rhetorical norms, and plan instructional responses. This process of collaborative sensemaking transformed the AI\u0026rsquo;s monologic output into dialogic, formative assessment practices. The hybrid AI\u0026ndash;human feedback model not only fostered teachers\u0026rsquo; diagnostic precision and assessment practices but also cultivated professional identity development, collaborative learning, and context-sensitive adaptations.\u003c/p\u003e \u003cp\u003eThe hybrid AI\u0026ndash;human feedback model not only developed teachers\u0026rsquo; diagnostic precision and assessment practices but also nurtured professional identity development, collaborative learning, and context-sensitive adaptations. These results in line with recent research highlighting the role of AI in increasing teacher decision-making while underlining the necessity of human mediation to guarantee culturally responsive assessment (Zhang et al., 2024; Kooli \u0026amp; Yusuf, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The study\u0026rsquo;s results challenge the idea that AI alone can produce significant professional development, as an alternative supporting a blended approach where technology functions as a catalyst for reflective conversation rather than a replacement for instructional expertise (Mohammed \u0026amp; \u0026lsquo;Nell\u0026rsquo;Watson, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Maurya et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eA key outcome of this research lies in indicating how incorporating AI-informed feedback into CAR can help close the gap between broad, standardized evaluation measures and the particular demands of local instructional contexts. Instead of simply accepting the AI\u0026rsquo;s recommendations, the teachers, participated in collaborative reflection, critically examined the AI\u0026rsquo;s suggestions and reshaped them to fit the social and cultural traditions and the needs of their students. This supports Lam\u0026rsquo;s (2025) position that LAL development must be grounded in sociocultural realities to circumvent the pitfalls of decontextualized, excessively technical assessment models. Similarly, the study underlines the role of educators as active interpreters of AI input, an idea aligns with recent work on \u0026ldquo;augmented intelligence\u0026rdquo; in education, which highlights that human expertise should persist at the core of fair and responsible assessment (Grab, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Trajkovski \u0026amp; Hayes, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe reported shift toward formative practices support Sch\u0026ouml;n\u0026rsquo;s reflection-on-action. The repetitive cycle of executing an assessment, gaining instant AI data, and collectively reflecting on its implications in CAR meetings generated structured opportunities for teachers to question the dominance of summative testing. They started to internalize the value of continuous evidence collecting, going from seeing assessment as a final judgment to viewing it as a feedback tool for regulation\u0026mdash;both for students and their own instruction. This research also contributes to a profounder understanding of how CAR can be improved through the strategic implementation of AI. Conventional CAR approaches usually hinge on planned peer reviews for feedback, proposing little in the way of instantaneous, data-informed understandings (Schildkamp, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In contrast, the hybrid model probed here equipped teachers with immediate analytic feedback, letting them to improve assessment practices in shorter, more responsive cycles. In doing so, it addresses a limitation recognized by Yosief et al. (\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), who observed that the lag in feedback common to CAR can slow necessary teaching modifications. At the same time, the findings bring to light continuous challenges, including augmented workload and the risk of algorithmic bias, that mirror wider discussions about the complexities of integrating AI into teacher training (Erdem Coşgun, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Ghimire, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). These issues underline the need for robust institutional support and gradual application to support permanent, effective incorporation.\u003c/p\u003e \u003cp\u003eParticularly, the study\u0026rsquo;s findings challenge the notion that resource-limited contexts like Iran are ill-suited for AI-driven professional development. Although infrastructure limitations continued, teachers proved significant flexibility, blending digital and non-digital tools to maximize AI\u0026rsquo;s utility. This resilience supports Ghourchian\u0026rsquo;s (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) observations about Iranian educators\u0026rsquo; ability for innovation despite systemic limitations. Furthermore, the study\u0026rsquo;s evidence of peer-led scaling, where teachers distributed AI-modified assessment practices across schools, supports recent calls for decentralized, teacher-owned models of professional learning (Dulfer et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kim, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe relevance of these findings reaches beyond the Iranian context, providing a transferable model for fostering LAL in comparable contexts. By framing AI as a partner in the assessment process instead of a top-down authority on quality, the approach sidesteps ethical concerns focusing on excessive dependence on automation (Malik et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Upcoming work should explore its long-term influence, mainly whether the combination of AI and CAR produces long-lasting gains in student learning. More inquiry into systemic and policy limitations, including centralized testing commands that restrict the uptake of formative assessment, could shed added light on the model\u0026rsquo;s flexibility (Au, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Mirsanjari, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study was implemented in particular Iranian urban/semi-urban contexts; its findings may not transfer to rural contexts with severe technological limitations. The six-month timeframe also confines claims about long-term sustainability. Future research should explore longitudinal effects on student learning outcomes and probe the application of this hybrid model in other national settings with different policy limitations. Eventually, this research illustrates that incorporating AI-driven feedback with CAR can meaningfully reinforce LAL by encouraging assessment practices that are reflective, evidence-informed, and culturally responsive. It makes the case for professional learning designs that integrate technology not as a stand-alone solution, but as a catalyst for deepening teachers\u0026rsquo; analytical engagement with assessment, an approach that supports broader international efforts to guarantee AI functions human-centered goals in education (Zhang et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Topping et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). As education systems around the world probe both the promise and the complexity of AI adoption, the framework suggested here provides a compelling route for enabling educators as active interpreters and co-creators of assessment innovation.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study shows that a hybrid AI\u0026ndash;human feedback model, rooted within a CAR framework, considerably influences the development of LAL among Iranian EFL school teachers. The findings offer a direct answer to the research question by indicating that this influence functions through a process of augmented professional judgment. Teachers leveraged AI-driven analytics to prolong their diagnostic reach, recognizing nuanced patterns in student performance, while the structured collaboration of CAR offered the crucial space to critically interpret, culturally contextualize, and instructionally adapt these algorithmic insights. This synergy nurtured a supportable shift from a summative to a formative assessment mindset, improved teachers\u0026rsquo; confidence and autonomy in assessment design, and nurtured collaborative networks for professional innovation.\u003c/p\u003e \u003cp\u003eThe main contribution of this research is a replicable, context-sensitive model for LAL development in resource-constrained school contexts. It extends beyond presenting AI as an automated solution by providing a real framework where technology works as a catalyst for deep, collaborative professional inquiry. The model bridges a critical gap in the literature by placing AI-mediated teacher development explicitly within the school-level context and offering empirical evidence on how teachers can become active interpreters\u0026mdash;not just passive implementers\u0026mdash;of assessment technology.\u003c/p\u003e \u003cp\u003eFuture research should construct directly upon these findings. First, longitudinal studies are needed to examine the long-term influence of this hybrid model on student learning outcomes and the durability of teachers\u0026rsquo; assessment innovations. Second, exploration is needed into the systemic and policy barriers\u0026mdash;mainly rigid, centralized testing regimes\u0026mdash;that may restrict the scalability of such formative, teacher-led assessment approaches. In conclusion, research should concentrate on the design principles for AI tools that are more transparent and customizable, thus better supporting teachers\u0026rsquo; agency and contextual negotiation in the assessment process.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eCollaborative Action Research (CAR)\u003c/p\u003e\n\u003cp\u003eEnglish Language Teaching (ELT)\u003c/p\u003e\n\u003cp\u003eLanguage Assessment Literacy (LAL)\u003c/p\u003e\n\u003cp\u003eNatural Language Processing (NLP)\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eEthics Approval and Consent to Participate\u003c/h2\u003e\n\u003cp\u003eThis study was reviewed and approved by the appropriate institutional ethics committee. All participants provided informed written consent prior to participation.\u003c/p\u003e\n\u003ch2\u003eConsent for Publication\u003c/h2\u003e\n\u003cp\u003eAll participants consented to the publication of anonymized data and findings derived from this study.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThis research received no external funding.\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eArefian did all stages individually.\u003c/p\u003e\n\u003ch2\u003eAcknowledgement\u003c/h2\u003e\n\u003cp\u003eArefian did all stages individually.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbrar-ul-Hassan, S., \u0026amp; Nassaji, H. (2024). Rescoping language assessment literacy: An expanded perspective. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e120\u003c/em\u003e, 103195. https://doi.org/10.1016/j.system.2023.103195\u003c/li\u003e\n\u003cli\u003eAliakbari, M., \u0026amp; Sadeghi, S. (2022). The professional identity of the Iranian teachers: A case of professional practices. \u003cem\u003eTeacher Development\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(3), 411\u0026ndash;431. https://doi.org/10.1080/13664530.2022.2076729\u003c/li\u003e\n\u003cli\u003eAliakbari, M., Yasini, A., \u0026amp; Sadeghi, S. (2023). Iranian EFL Teachers\u0026apos; Classroom Assessment Practices: Discrepancy between Theory and Practice. \u003cem\u003eInternational Journal of Language Testing\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(2), 149\u0026ndash;169.\u003c/li\u003e\n\u003cli\u003eAsad, M. M., Naz, A., Churi, P., \u0026amp; Tahanzadeh, M. M. (2021). Virtual reality as pedagogical tool to enhance experiential learning: a systematic literature review. \u003cem\u003eEducation Research International\u003c/em\u003e, \u003cem\u003e2021\u003c/em\u003e(1), 7061623. https://doi.org/10.1155/2021/7061623\u003c/li\u003e\n\u003cli\u003eAu, W. (2022). \u003cem\u003eUnequal by design: High-stakes testing and the standardization of inequality\u003c/em\u003e. Routledge. https://doi.org/10.4324/9781003005179\u003c/li\u003e\n\u003cli\u003eBazvand, A. D., Jalilzadeh, K., Farhady, H., \u0026amp; Sabzehparvar, A. (2025). Language assessment literacy of professional language testers: Implications for teacher education reform. \u003cem\u003eSystem\u003c/em\u003e, \u003cem\u003e134\u003c/em\u003e, 103792. https://doi.org/10.1016/j.system.2025.103792\u003c/li\u003e\n\u003cli\u003eBenade, L. (2015). Teachers\u0026rsquo; critical reflective practice in the context of twenty-first century learning. \u003cem\u003eOpen Review of Educational Research\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(1), 42\u0026ndash;54. https://doi.org/10.1080/23265507.2014.998159\u003c/li\u003e\n\u003cli\u003eBhimdiwala, A., Neri, R. C., \u0026amp; Gomez, L. M. (2022). Advancing the design and implementation of artificial intelligence in education through continuous improvement. \u003cem\u003eInternational Journal of Artificial Intelligence in Education\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(3), 756\u0026ndash;782. https://doi.org/10.1007/s40593-021-00278-8\u003c/li\u003e\n\u003cli\u003eBleicher, R. E. (2014). A collaborative action research approach to professional learning. \u003cem\u003eProfessional development in education\u003c/em\u003e, \u003cem\u003e40\u003c/em\u003e(5), 802\u0026ndash;821. https://doi.org/10.1080/19415257.2013.842183\u003c/li\u003e\n\u003cli\u003eChang, D. Y. S., Lin, M. H., \u0026amp; Lee, J. Y. (2024). Exploring language assessment literacy and needs of English teachers at senior high school level. \u003cem\u003eAsia Pacific Journal of Education\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(4), 854\u0026ndash;872. https://doi.org/10.1080/02188791.2024.2313486\u003c/li\u003e\n\u003cli\u003eClark, I. (2012). Formative assessment: Assessment is for self-regulated learning. \u003cem\u003eEducational psychology review\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(2), 205\u0026ndash;249. https://doi.org/10.1007/s10648-011-9191-6\u003c/li\u003e\n\u003cli\u003eCoombe, C., Vafadar, H., \u0026amp; Mohebbi, H. (2020). Language assessment literacy: What do we need to learn, unlearn, and relearn? \u003cem\u003eLanguage Testing in Asia\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(1), 3. https://doi.org/10.1186/s40468-020-00101-6\u003c/li\u003e\n\u003cli\u003eCope, B., Kalantzis, M., \u0026amp; Searsmith, D. (2021). Artificial intelligence for education: Knowledge and its assessment in AI-enabled learning ecologies. \u003cem\u003eEducational philosophy and theory\u003c/em\u003e, \u003cem\u003e53\u003c/em\u003e(12), 1229\u0026ndash;1245. https://doi.org/10.1080/00131857.2020.1728732\u003c/li\u003e\n\u003cli\u003eCui, Y., Liu, Y., \u0026amp; Yu, H. (2025). Enhancing English teachers\u0026rsquo; language assessment literacy through a professional learning community: A collaborative autoethnography study. \u003cem\u003eSystem\u003c/em\u003e, 103799. https://doi.org/10.1016/j.system.2025.103799\u003c/li\u003e\n\u003cli\u003eCurry, M. W., \u0026amp; Athanases, S. Z. (2020). In pursuit of engaged learning with Latinx students: Expanding learning beyond classrooms through performance-based engagements. \u003cem\u003eTeachers College Record\u003c/em\u003e, \u003cem\u003e122\u003c/em\u003e(8), 1\u0026ndash;49. https://doi.org/10.1177/016146812012200815\u003c/li\u003e\n\u003cli\u003eDerakhshan, A., \u0026amp; Ghiasvand, F. (2022). Demystifying Iranian EFL teachers\u0026rsquo; perceptions and practices of learning-oriented assessment (LOA): Challenges and prospects in focus. \u003cem\u003eLanguage Testing in Asia\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(1), 55. https://doi.org/10.1186/s40468-022-00204-2\u003c/li\u003e\n\u003cli\u003eDulfer, N., Kriewaldt, J., \u0026amp; McKernan, A. (2024). Using collaborative action research to enhance differentiated instruction. \u003cem\u003eInternational Journal of Inclusive Education\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(8), 1402\u0026ndash;1416. https://doi.org/10.1080/13603116.2021.1992678\u003c/li\u003e\n\u003cli\u003eErdem Coşgun, G. (2025). Artificial intelligence literacy in assessment: Empowering pre‐service teachers to design effective exam questions for language learning. \u003cem\u003eBritish Educational Research Journal\u003c/em\u003e. https://doi.org/10.1002/berj.4177\u003c/li\u003e\n\u003cli\u003eFiroozi, T., Razavipour, K., \u0026amp; Ahmadi, A. (2019). The language assessment literacy needs of Iranian EFL teachers with a focus on reformed assessment policies. \u003cem\u003eLanguage Testing in Asia\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(1), 2. https://doi.org/10.1186/s40468-019-0078-7\u003c/li\u003e\n\u003cli\u003eGan, L., \u0026amp; Lam, R. (2022). A review on language assessment literacy: Trends, foci and contributions. \u003cem\u003eLanguage Assessment Quarterly\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(5), 503\u0026ndash;525. https://doi.org/10.1080/15434303.2022.2128802\u003c/li\u003e\n\u003cli\u003eGhimire, A. (2025). Utilizing ChatGPT to integrate world English and diverse knowledge: A transnational perspective in critical artificial intelligence (AI) literacy. Computers and Composition, 75, 102913. https://doi.org/10.1016/j.compcom.2024.102913\u003c/li\u003e\n\u003cli\u003eGhourchian, P. (2024). Online education development in iran during the covid-19 pandemic: Opportunities and challenges. \u003cem\u003eOnline Education During COVID-19 and Beyond: Opportunities, Challenges and Outlook\u003c/em\u003e, 341\u0026ndash;364. https://doi.org/10.1007/978-3-031- 49353-9_18\u003c/li\u003e\n\u003cli\u003eGrab, M. O. (2025). Teaching for Equity: An Exploration of AI\u0026apos;s Role in Culturally Responsive Teaching in Higher Education Settings. \u003cem\u003eInnovative Higher Education\u003c/em\u003e, 1\u0026ndash;22. https://doi.org/10.1007/s10755-025-09801-4\u003c/li\u003e\n\u003cli\u003eInbar-Lourie, O. (2013). Guest editorial to the special issue on language assessment literacy. \u003cem\u003eLanguage Testing\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e(3), 301\u0026ndash;307. https://doi.org/10.1177/0265532213480126\u003c/li\u003e\n\u003cli\u003eInbar-Lourie, O. (2016). Language assessment literacy. In \u003cem\u003eLanguage testing and assessment\u003c/em\u003e (pp. 1\u0026ndash;14). Springer, Cham. https://doi.org/10.1007/978-3-319-02326-7_19-1\u003c/li\u003e\n\u003cli\u003eKim, J. (2024). Leading teachers\u0026apos; perspective on teacher-AI collaboration in education. \u003cem\u003eEducation and information technologies\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e(7), 8693\u0026ndash;8724. https://doi.org/10.1007/s10639-023- 12109-5\u003c/li\u003e\n\u003cli\u003eKoh, K. H. (2011). Improving teachers\u0026rsquo; assessment literacy through professional development. \u003cem\u003eTeaching Education\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(3), 255\u0026ndash;276. https://doi.org/10.1080/10476210.2011.593164\u003c/li\u003e\n\u003cli\u003eKooli, C., \u0026amp; Yusuf, N. (2025). Transforming educational assessment: Insights into the use of ChatGPT and large language models in grading. \u003cem\u003eInternational Journal of Human\u0026ndash; Computer Interaction\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(5), 3388-3399. https://doi.org/10.1080/10447318.2024.2338330\u003c/li\u003e\n\u003cli\u003eLam, R. (2015). Language assessment training in Hong Kong: Implications for language assessment literacy. \u003cem\u003eLanguage Testing\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(2), 169\u0026ndash;197. https://doi.org/10.1177/0265532214554321\u003c/li\u003e\n\u003cli\u003eMirsanjari, Z. (2025). Enhancing assessment literacy in EAP instruction: the role of teacher development courses in overcoming systemic barriers. \u003cem\u003eLanguage Testing in Asia\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(1), 30. https://doi.org/10.1186/s40468-025-00368-7\u003c/li\u003e\n\u003cli\u003eLevi, T., \u0026amp; Inbar-Lourie, O. (2020). Assessment literacy or language assessment literacy: Learning from the teachers. \u003cem\u003eLanguage Assessment Quarterly\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(2), 168\u0026ndash;182. https://doi.org/10.1080/15434303.2019.1692347\u003c/li\u003e\n\u003cli\u003eMalik, A., Khan, M. L., Hussain, K., Qadir, J., \u0026amp; Tarhini, A. (2025). AI in higher education: unveiling academicians\u0026rsquo; perspectives on teaching, research, and ethics in the age of ChatGPT. \u003cem\u003eInteractive Learning Environments\u003c/em\u003e, \u003cem\u003e33\u003c/em\u003e(3), 2390\u0026ndash;2406. https://doi.org/10.1080/10494820.2024.2409407\u003c/li\u003e\n\u003cli\u003eManfra, M. M. (2019). Action research and systematic, intentional change in teaching practice. \u003cem\u003eReview of research in education\u003c/em\u003e, \u003cem\u003e43\u003c/em\u003e(1), 163\u0026ndash;196. https://doi.org/10.3102/0091732X18821132\u003c/li\u003e\n\u003cli\u003eMaurya, R. K., Montesinos, S., Bogomaz, M., \u0026amp; DeDiego, A. C. (2025). Assessing the use of ChatGPT as a psychoeducational tool for mental health practice. \u003cem\u003eCounselling and Psychotherapy Research\u003c/em\u003e, \u003cem\u003e25\u003c/em\u003e(1), e12759. https://doi.org/10.1002/capr.12759\u003c/li\u003e\n\u003cli\u003eMohammed, P. S., \u0026amp; \u0026lsquo;Nell\u0026rsquo;Watson, E. (2019). Towards inclusive education in the age of artificial intelligence: Perspectives, challenges, and opportunities. \u003cem\u003eArtificial Intelligence and Inclusive Education: Speculative futures and emerging practices\u003c/em\u003e, 17\u0026ndash;37. https://doi.org/10.1007/978-981-13-8161-4_2\u003c/li\u003e\n\u003cli\u003eMorris, R., Perry, T., \u0026amp; Wardle, L. (2021). Formative assessment and feedback for learning in higher education: A systematic review. \u003cem\u003eReview of Education\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(3), e3292. https://doi.org/10.1002/rev3.3292\u003c/li\u003e\n\u003cli\u003eMun, R. U., Ezzani, M. D., \u0026amp; Lee, L. E. (2020). Culturally relevant leadership in gifted education: A systematic literature review. \u003cem\u003eJournal for the Education of the Gifted\u003c/em\u003e, \u003cem\u003e43\u003c/em\u003e(2), 108\u0026ndash;142. https://doi.org/10.1177/0162353220912009\u003c/li\u003e\n\u003cli\u003eQing-Li, H., Torres, M. N., \u0026amp; Shi-Ji, F. (2019). Collaborative action research for preparing teachers as reflective practitioners. \u003cem\u003eSystemic Practice and Action Research\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(4), 411\u0026ndash; 427. https://doi.org/10.1007/s11213-018-9461-z\u003c/li\u003e\n\u003cli\u003eSchildkamp, K. (2019). Data-based decision-making for school improvement: Research insights and gaps. \u003cem\u003eEducational research\u003c/em\u003e, \u003cem\u003e61\u003c/em\u003e(3), 257\u0026ndash;273. https://doi.org/10.1080/00131881.2019.1625716\u003c/li\u003e\n\u003cli\u003eSmith, K. (2010). Assessing the Practicum in teacher education\u0026ndash;Do we want candidates and mentors to agree?. \u003cem\u003eStudies in educational evaluation\u003c/em\u003e, \u003cem\u003e36\u003c/em\u003e(1-2), 36\u0026ndash;41. https://doi.org/10.1016/j.stueduc.2010.08.001\u003c/li\u003e\n\u003cli\u003eSultana, N. (2019). Language assessment literacy: An uncharted area for the English language teachers in Bangladesh. \u003cem\u003eLanguage Testing in Asia\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(1), 1. https://doi.org/10.1186/s40468- 019-0077-8\u003c/li\u003e\n\u003cli\u003eTopping, K. J., Gehringer, E., Khosravi, H., Gudipati, S., Jadhav, K., \u0026amp; Susarla, S. (2025). Enhancing peer assessment with artificial intelligence. \u003cem\u003eInternational Journal of Educational Technology in Higher Education\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(1), 3. https://doi.org/10.1186/s41239- 024-00501-1\u003c/li\u003e\n\u003cli\u003eTrajkovski, G., \u0026amp; Hayes, H. (2025). AI-Assisted Formative Assessment and Feedback. In \u003cem\u003eAI- Assisted Assessment in Education: Transforming Assessment and Measuring Learning\u003c/em\u003e (pp. 283\u0026ndash;312). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-88252- 4_7\u003c/li\u003e\n\u003cli\u003eYan, L., Suleman Abdullah Alwabel, A., \u0026amp; Mohamad, U. H. (2025). AI‐Powered Education: Transforming Teacher‐Student Interactions and Advancing Sustainable Learning Practices. \u003cem\u003eEuropean Journal of Education\u003c/em\u003e, \u003cem\u003e60\u003c/em\u003e(4), e70351. https://doi.org/10.1111/ejed.70351\u003c/li\u003e\n\u003cli\u003eYin, X., \u0026amp; Buck, G. A. (2019). Using a collaborative action research approach to negotiate an understanding of formative assessment in an era of accountability testing. \u003cem\u003eTeaching and Teacher Education\u003c/em\u003e, \u003cem\u003e80\u003c/em\u003e, 27\u0026ndash;38. https://doi.org/10.1016/j.tate.2018.12.018\u003c/li\u003e\n\u003cli\u003eYosief, A., Sulieman, M. S., \u0026amp; Biede, T. (2024). Improving the practices of teacher educators through collaborative action research: Challenges and hopes. \u003cem\u003eEducational Action Research\u003c/em\u003e, \u003cem\u003e32\u003c/em\u003e(2), 204\u0026ndash;221. https://doi.org/10.1080/09650792.2022.2066147\u003c/li\u003e\n\u003cli\u003eZhang, Y., Zhang, M., Wu, L., \u0026amp; Li, J. (2025). Digital Transition Framework for Higher Education in AI-Assisted Engineering Teaching: Challenge, Strategy, and Initiatives in China. \u003cem\u003eScience \u0026amp; Education\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e(2), 933\u0026ndash;954. https://doi.org/10.1007/s11191-024-00575- 3\u003c/li\u003e\n\u003cli\u003eZhao, D. (2024). The impact of AI-enhanced natural language processing tools on writing proficiency: An analysis of language precision, content summarization, and creative writing facilitation. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, 1\u0026ndash;32. https://doi.org/10.1007/s10639-024-13145-5\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"language-testing-in-asia","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ltia","sideBox":"Learn more about [Language Testing in Asia](http://languagetestingasia.springeropen.com)","snPcode":"40468","submissionUrl":"https://submission.springernature.com/new-submission/40468/3","title":"Language Testing in Asia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Language Assessment Literacy, Artificial Intelligence, Collaborative Action Research, Teacher Professional Development, Formative Assessment, EFL Context","lastPublishedDoi":"10.21203/rs.3.rs-8281675/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8281675/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study explores the influence of a hybrid AI-human feedback model, incorporated within a Collaborative Action Research (CAR) framework, on developing Language Assessment Literacy (LAL) among Iranian English language teachers. Grounded in the challenges of conventional assessment practices in Iran\u0026rsquo;s EFL context, containing overreliance on summative testing and inadequate professional development, the research engaged a qualitative CAR methodology with 15 teachers over six months. Participants involved in iterative cycles of assessment design, application, and reflection, supported by AI-driven analytics (e.g., automated scoring and natural language processing (NLP) feedback) alongside structured peer and mentor collaboration. Findings indicated that the AI-CAR synergy improved teachers\u0026rsquo; diagnostic precision, empowered culturally responsive assessment adaptations, and nurtured sustainable shifts toward formative practices. Main results involved improved recognition of learner patterns, critical negotiation of algorithmic feedback, and the emergence of teacher-led assessment innovations. Yet, challenges appeared concerning workload intensity and contextualizing AI recommendations within Iranian instructional and cultural norms. This study addressed the research problem of underdeveloped LAL among Iranian EFL school teachers, considered by overreliance on summative testing. Implementing a qualitative CAR design with 15 teachers over six months, it explored how a hybrid AI-human feedback model influences LAL development. Key findings indicate the model improved teachers\u0026rsquo; diagnostic precision and nurtured formative assessment practices through critical negotiation of AI feedback. The conclusion underlines that sustainable LAL development requires hybrid models where AI augments, not substitutes, teacher judgment.\u003c/p\u003e","manuscriptTitle":"Supporting Iranian Teachers’ Language Assessment Literacy: A Hybrid AI–Human Feedback Approach within Collaborative Action Research","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-22 06:57:28","doi":"10.21203/rs.3.rs-8281675/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-21T16:31:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"22803491968316610074111410486433051340","date":"2026-04-14T19:32:43+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-31T18:28:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"87241244668591001251121840197403853138","date":"2025-12-18T05:38:41+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-12-18T02:51:29+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-09T07:44:55+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-12-09T07:41:47+00:00","index":"","fulltext":""},{"type":"submitted","content":"Language Testing in Asia","date":"2025-12-04T17:18:23+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"language-testing-in-asia","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ltia","sideBox":"Learn more about [Language Testing in Asia](http://languagetestingasia.springeropen.com)","snPcode":"40468","submissionUrl":"https://submission.springernature.com/new-submission/40468/3","title":"Language Testing in Asia","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Open","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"024ec1ea-fe02-4866-95ff-e9a6a7b927a8","owner":[],"postedDate":"December 22nd, 2025","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-21T16:31:18+00:00","index":21,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-12-22T06:57:28+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-22 06:57:28","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8281675","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8281675","identity":"rs-8281675","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}