Computers and Education Artificial Intelligence最新文献

{"title":"Potential risks of generative artificial intelligence integration into K-12 education: A scoping review","authors":"Sisi Tao ,&nbsp;Min Lan ,&nbsp;Minjuan Wang ,&nbsp;Hui Li","doi":"10.1016/j.caeai.2026.100561","DOIUrl":"10.1016/j.caeai.2026.100561","url":null,"abstract":"<div><div>The rapid integration of Generative AI (GenAI) into K-12 education has outpaced our understanding of its potential risks and unintended consequences. Existing reviews have prioritized higher education and technical promise, overlooking the potential developmental risks it poses to children and adolescents. This scoping review synthesizes 22 empirical studies from K-12 contexts to map the types of risks and concerns reported in research, compile mitigation strategies, and identify priority gaps that could guide more developmentally informed future studies. Across the included studies, reported risks clustered into three domains: (a) risks to psychological wellbeing, with evidence of emotional disconnection and social isolation; (b) risks to intellectual agency, comprising cognitive dependency, distorted self-assessment, and the erosion of creative authorship; and (c) risks to ecological environments, including limited institutional readiness, unclear governance, equity gaps, and privacy concerns that complicate safe and consistent student engagement with GenAI. Promising mitigation strategies identified include designing tasks that value process over product, using GenAI to generate scaffolding (hints) rather than direct solutions, and embedding tools within critical AI literacy curricula. Ultimately, this review suggests that without intentional, developmentally responsive governance, GenAI risks displacing the productive struggle and authentic expression necessary for learning and identity formation. We conclude that safe integration requires shifting the focus from technical adoption to ecological protection, ensuring that tools function as transparent scaffolds for human cognition rather than opaque substitutes for student agency in the K-12 context.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100561"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147385023","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Modeling generative AI adoption in higher education: An integrated TAM–TPB–SDT framework with SEM validation","authors":"Dina Tbaishat ,&nbsp;Omar AlFandi ,&nbsp;Faten Hamad ,&nbsp;Syed Muhammad Salman Bukhari ,&nbsp;Suha Al Muhaissen","doi":"10.1016/j.caeai.2026.100541","DOIUrl":"10.1016/j.caeai.2026.100541","url":null,"abstract":"<div><div>This study investigates the determinants of university students' adoption of generative artificial intelligence (GAI) tools in higher education. Integrating the Technology Acceptance Model (TAM), the Theory of Planned Behavior (TPB), and Self-Determination Theory (SDT), it develops and tests a complete model that captures cognitive, social, and motivational influences on adoption. A cross-sectional survey was conducted among 517 undergraduate and postgraduate students at Jordanian universities. The data were analyzed using structural equation modeling (SEM) with a two-step approach: confirmatory factor analysis (CFA) to validate the measurement model, followed by SEM to test the hypothesized structural relationships. Reliability, validity, measurement invariance across gender, and mediation effects were assessed. The integrated model showed excellent fit and substantial explanatory power, accounting for 83 % of the variance in behavioral intention and 81.6 % in actual AI use. Relatedness, perceived usefulness, attitude, and autonomy emerged as significant predictors of intention, while behavioral intention and competence predicted actual use. The ease of use strongly influenced usefulness, and mediation analysis confirmed indirect effects through usefulness and attitude. The model was invariant across gender groups, supporting its generalizability. This research extends TAM and TPB by integrating SDT's psychological needs, highlighting relatedness and competence as novel drivers of adoption. It provides the first empirical evidence from Jordan, a region underrepresented in the literature, highlighting that motivational dynamics carry greater weight than social norms in collectivist educational contexts. The study advances theoretical models of technology adoption and offers practical insights for universities and policymakers on promoting responsible and sustainable integration of AI in education.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100541"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926725","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Towards responsible AI in education: A Delphi-AHP-based framework for evaluating educational large language models","authors":"Pingrong Lin,&nbsp;Qin Deng,&nbsp;Yanbian Zhou","doi":"10.1016/j.caeai.2025.100534","DOIUrl":"10.1016/j.caeai.2025.100534","url":null,"abstract":"<div><div>As large language models (LLMs) become deeply integrated into the educational landscape, evaluation criteria focusing solely on performance are insufficient to mitigate the risks of value misalignment and socioethical concerns. To steer educational LLMs towards responsible and beneficial development, this study aims to construct a multidimensional evaluation framework grounded in educational theory. Initially, a preliminary pool of evaluation indicators was established on the basis of a review of the literature and pedagogical theories. The Delphi method was subsequently employed to refine the indicator structure by integrating opinions from 21 cross-disciplinary experts. The analytic hierarchy process (AHP) was then applied to weigh these indicators and determine their priorities. The final framework comprises 5 first-level indicators and 21 s-level indicators. Learning effectiveness, knowledge construction capability, and social alignment are assigned critical weights, whereas intelligent interaction capability is less prioritized. Among the second-level indicators, information veracity was weighted the highest, while educational equity had the weakest influence. This study not only provides direction for the development and optimization of educational LLMs but also offers a reference for establishing a responsible artificial intelligence in education ecosystem.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100534"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145791849","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Coauthorship integrity: Reconceptualising assessment validity for the age of generative artificial intelligence","authors":"Mohsen Ebrahimzadeh ,&nbsp;Antonette Shibani ,&nbsp;Simon Buckingham Shum","doi":"10.1016/j.caeai.2026.100609","DOIUrl":"10.1016/j.caeai.2026.100609","url":null,"abstract":"<div><div>We consider how a future of pervasive human/AI coauthorship challenges current notions of validity and academic integrity. Specifically, we address widespread concerns that in non-proctored contexts, students are using generative artificial intelligence (GenAI) to submit texts they do not understand. Adopting an assessment validity lens, we show how GenAI undermines the integrity of multiple forms of validity evidence, leading us to propose <em>Coauthorship Integrity</em> as a new conceptual source of validity evidence for addressing these threats. Coauthorship Integrity is violated when students submit AI-generated content that they do not understand. To hold students accountable in this regard, a university needs to explore practical methods to gather evidence that could potentially be implemented at scale. We report progress in the development of an “AI Viva”, a conversational agent that engages students in a hybrid experience, combining elements of a viva voce with comprehension questions of controllable type and complexity, engaging the learner with quantitative and dialogic feedback. To provide preliminary expert validation for this prototype, we report qualitative and quantitative evidence from in-depth evaluations with expert educators and assessment experts. By shifting assessment design towards verifying students' epistemic ownership of their writing, this work suggests a potential new pathway for integrating GenAI into pedagogically meaningful, validity-aligned assessment practices. This paper's contributions are thus threefold: <em>conceptual</em> (Coauthorship Integrity), <em>technical</em> (LLM-generated questions of controllable difficulty and type, from arbitrary texts) and <em>empirical</em> (preliminary expert evaluation of the concept and initial implementation of the prototype's potential for eliciting such evidence).</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100609"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148230838","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A deep learning approach to estimating interaction levels in face-to-face lessons","authors":"Danilo Valdes-Ramirez ,&nbsp;Jesus Alfonso Beltran-Sanchez ,&nbsp;Santiago Enrique Conant-Pablos ,&nbsp;Roberto Ponce Lopez ,&nbsp;Angeles Dominguez ,&nbsp;Claudia Camacho-Zuñiga ,&nbsp;Genaro Zavala","doi":"10.1016/j.caeai.2025.100528","DOIUrl":"10.1016/j.caeai.2025.100528","url":null,"abstract":"<div><div>Interaction is a key component of effective learning, fostering active participation and deeper understanding. Real-time insights into classroom interaction help instructors adapt their teaching strategies, while longitudinal data inform the design of improved learning activities. However, many educators remain unaware of the interaction levels during lessons. This study proposes an AI-powered algorithm capable of estimating classroom interaction levels every few seconds during face-to-face instruction. The method relies on a fine-tuned YOLOv8 model for detecting nonverbal interaction cues, followed by postprocessing and data fusion to compute interaction proportions relative to the number of students, ensuring full anonymity. The algorithm classifies behaviors into four interaction categories—student–professor, student–student, student–object, and no-interaction—and outputs a composite interaction score. The detection model achieved high accuracy (average precision, [email protected]&gt; 92 % across all categories and mean average precision [email protected] = 96 %). Validation with ten experienced professors who rated 100 classroom images revealed poor intraclass correlation (ICC = 0.209), underscoring the subjective nature of “interaction.” Statistical comparisons showed no significant differences <span><math><mo>(</mo><mi>p</mi><mo>&gt;</mo><mn>0.05</mn><mo>)</mo></math></span> between the algorithm’s estimates and expert ratings, though equivalence testing (TOST) did not confirm statistical equivalence <span><math><mo>(</mo><mi>p</mi><mo>&gt;</mo><mn>0.05</mn><mo>)</mo></math></span> for any comparison. The closest alignment occurred between the system’s student–student ratio and the experts’ median evaluations. Two case studies further illustrated the algorithm’s sensitivity to pedagogical context, capturing higher interaction levels during teamwork and workshop sessions. These findings demonstrate the potential of data-driven analytics to support reflective teaching and adaptive learning design.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100528"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145791851","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"What undergraduate students need to know and actually know about generative AI","authors":"Sina Rismanchian,&nbsp;Eesha Tur Razia Babar,&nbsp;Shayan Doroudi","doi":"10.1016/j.caeai.2026.100554","DOIUrl":"10.1016/j.caeai.2026.100554","url":null,"abstract":"<div><div>In November 2022, the release of ChatGPT sparked widespread adoption of generative AI chatbots among students, yet little is known about what undergraduate students actually understand about these tools or how accurately they perceive their capabilities. To address this gap, we propose a theoretically grounded framework for Generative AI (GenAI) literacy that integrates three forms of conceptual knowledge about large language models (LLMs)—foundations, capabilities, and limitations, and societal impact—with students' perceptions and folk theories of AI chatbots. We then developed and validated a GenAI literacy survey, including multiple-choice knowledge items and perception items, using expert review and item response theory (IRT) modeling. To examine GenAI literacy across different educational contexts, we conducted two complementary studies: Study 1 surveyed students enrolled in courses at a large public R1 university, and Study 2 surveyed a national sample of U.S. undergraduate students recruited via Prolific. Across both samples, approximately 60% of students reported using AI chatbots weekly or daily. However, many students overestimated chatbots’ capabilities, particularly on tasks involving reasoning and counting, and often anthropomorphized or treated chatbots as search engines. Knowledge scores were higher among computer science students at the R1 university and frequent chatbot users, while perception accuracy varied by group. Critically, greater conceptual knowledge was associated with less overestimation of chatbot abilities, suggesting that knowledge supports more calibrated and responsible use. This work introduces a validated framework and instrument for assessing GenAI literacy and highlights the need for AI literacy initiatives that move beyond tool usage to address misconceptions, beliefs, and responsible engagement with generative AI in higher education.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100554"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147385020","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Improving instructional design proficiency of master's students in mathematics education through intelligent educational technologies","authors":"Fusheng Zhu ,&nbsp;Qian liang ,&nbsp;Zhanmin Mao ,&nbsp;Ying Wang","doi":"10.1016/j.caeai.2026.100597","DOIUrl":"10.1016/j.caeai.2026.100597","url":null,"abstract":"<div><div>Research on smart education has predominantly focused on macro-level concepts and frameworks, such as core principles and system architectures, while offering limited guidance for discipline-specific integration and practical implementation. This study addresses two critical gaps: (1) how an M.Ed. course can be enhanced through intelligent educational technologies, and (2) whether a smart classroom teaching mode can effectively improve M.Ed. students’ instructional design competence, particularly in terms of precision and professionalism.</div><div>Adopting a pedagogically oriented approach rather than an algorithmic one, this study proposes a three-dimensional framework encompassing learning effectiveness, information and communication technology (ICT), and classroom organization. Based on this framework, a set of smart classroom characteristics aligned with M.Ed. program objectives is developed. Using a mathematics M.Ed. course as a case study, the study integrates intelligent educational technologies—such as automated scoring, personalized recommendations, and multi-AI feedback—across the pre-class, in-class, and post-class stages to construct a technology-enhanced teaching loop.</div><div>Results from a quasi-experimental design and quantitative analysis indicate that this approach significantly improves students' ability to formulate precise and professionally grounded instructional objectives. Building on these findings, the study proposes the “D–T–E Model” (Disciplinary Demand–Technological Empowerment–Evaluation Loop) as a transferable pedagogical framework for professional master's education. The findings provide practical implications for the design and implementation of AI-supported smart classrooms in teacher education contexts.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100597"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148167578","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Augmenting knowledge tracing through modeling dynamic higher-order concept interactions: A temporal hypergraph memory network","authors":"Mehrnoush Mohammadi ,&nbsp;Kamal Berahmand ,&nbsp;Shazia Sadiq ,&nbsp;Hassan Khosravi","doi":"10.1016/j.caeai.2026.100616","DOIUrl":"10.1016/j.caeai.2026.100616","url":null,"abstract":"<div><div>Knowledge Tracing (KT) aims to estimate students’ evolving knowledge states from their question–answer interactions, with performance prediction as the evaluation proxy. Memory-augmented neural networks have transformed KT by automatically extracting latent concepts and maintaining dynamic, interpretable representations of student understanding. Despite these advances, such models treat concepts as independent entities. They fail to capture how higher-order interactions among concepts shape learning trajectories - particularly when concepts co-occur across varying question compositions. To address these gaps, we propose the Temporal Hypergraph Memory Network (THyMeN), a hybrid model that augments memory-based tracing with temporal hypergraph reasoning. THyMeN captures dynamic higher-order concept interactions by representing students’ learning histories as evolving hypergraphs, where each hyperedge reflects the multi-concept structure of a question. A bidirectional message-passing mechanism enables mutual refinement between concept nodes and question hyperedges, modeling how concept dynamics shift across questions. An attention-based fusion mechanism then integrates this composition-aware information with concept-level knowledge states, enabling predictions that reflect both contextual variability and persistent learning. Additionally, an adaptive scaling strategy moderates mastery updates using the diversity of concept co-occurrences across questions, aiming to model stable trajectories consistent with learning from varied practice. Experiments on four benchmark datasets show that THyMeN outperforms seven baselines and state-of-the-art models in predictive accuracy, and generates smoother, pedagogically plausible knowledge trajectories. Ablation and structural comparison studies further validate its design innovations, underscoring the importance of modeling concept co-occurrence and practice diversity. These results establish THyMeN as an interpretable framework that advances KT toward explainable and pedagogically grounded learning support.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100616"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148167579","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Corrigendum to “Evaluating lab assistant chatbot on student learning and behaviors in a programming short course” [Computers and Education: Artificial Intelligence 10 (2026) 100527]","authors":"Thanapon Noraset,&nbsp;Akara Supratak,&nbsp;Chaiyong Ragkhitwetsagul,&nbsp;Nubthong Worathong,&nbsp;Suppawong Tuarob","doi":"10.1016/j.caeai.2026.100547","DOIUrl":"10.1016/j.caeai.2026.100547","url":null,"abstract":"","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100547"},"PeriodicalIF":23.4,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148565825","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Can students judge like experts? A large-scale study on the pedagogical quality of AI and human personalized formative feedback","authors":"Tanya Nazaretsky ,&nbsp;Hagit Gabbay ,&nbsp;Tanja Käser","doi":"10.1016/j.caeai.2025.100533","DOIUrl":"10.1016/j.caeai.2025.100533","url":null,"abstract":"<div><div>While feedback is essential for guiding student learning, providing timely and personalized guidance in large-scale educational settings remains a significant challenge. Generative AI offers a scalable solution, yet little is known about students’ perceptions of AI-generated feedback. In this paper, we aim to investigate how the identity of the feedback provider (human vs. AI) affects students’ ability to assess feedback quality and whether their judgments are biased. We propose a comprehensive rubric for assessing the pedagogical quality of formative feedback. We use it to compare the objective quality of AI-generated and human-crafted feedback (N = 979). Next, using data collected from 472 STEM students, we examine the extent to which students’ perceptions of the same feedback align with those of the experts. Our contribution is threefold. First, by introducing a structured rubric, we address the need for more standardized and reliable methods to assess the pedagogical quality of AI-generated feedback. Second, our analysis indicates that the pedagogical quality of AI-generated feedback is, in practice, comparable to that of human-authored feedback. However, both types exhibit limitations, particularly in addressing metacognitive aspects. Third, students’ evaluations are largely influenced by their perceptions of the feedback provider’s credibility rather than the actual quality of the feedback itself. This pattern is consistent across all academic levels, genders, and fields of study. Our findings underscore the need for targeted strategies to enhance students’ ability to evaluate feedback objectively and to improve the pedagogical quality of AI-generated feedback, thereby strengthening the effectiveness of AI-powered educational feedback systems.</div></div>","PeriodicalId":34469,"journal":{"name":"Computers and Education Artificial Intelligence","volume":"10 ","pages":"Article 100533"},"PeriodicalIF":0.0,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926800","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}