AI in Education

The rapid integration of generative artificial intelligence (GenAI) into education across all sectors has prompted a proliferating body of scholarship addressing the ethical, social, and environmental implications of these technologies. This narrative literature review synthesises international empirical, conceptual, and policy literature published between 2022 and 2026 to trace the evolving story of ethical concerns surrounding GenAI in education. Drawing on the moral philosophy of consequentialism, particularly the utilitarian ethics of John Stuart Mill, the review analyses six interconnected domains of ethical concern: environmental sustainability and the carbon footprint of AI infrastructure; algorithmic bias, ideological encoding, and the reproduction of misinformation; user dependency and the erosion of learner agency; the displacement of critical and creative thinking; data privacy and surveillance; and the orientation of major GenAI platforms toward profit-driven and capitalistic outcomes. Unlike systematic reviews that privilege methodological replicability, this narrative review foregrounds interpretive synthesis, tracing how the ethical discourse has shifted from early alarm and prohibition toward more nuanced frameworks for responsible integration. The review identifies a consequentialist tension at the heart of the debate: while GenAI offers measurable benefits in personalisation, accessibility, and efficiency, these gains must be weighed against distributed harms that disproportionately affect vulnerable populations, the natural environment, and the epistemic foundations of education itself. The review concludes with a set of guidelines for the ethical use of GenAI in educational contexts, grounded in the literature synthesised in the article. Full article

This systematic review examines research published between 2021 and 2025 on generative AI and chatbot use in secondary education across nine countries in Southern Europe and the Balkans: Greece, Italy, Spain, Portugal, Malta, Serbia, Croatia, Bulgaria, and Romania. Drawing on studies from IEEE Xplore, the ACM Digital Library, Google Scholar, and arXiv, this review synthesizes evidence on instructional uses, reported learning outcomes, teacher readiness, governance, and language-localization constraints. Across the region, the literature shows rapid experimentation in writing, language learning, programming, and project-based learning but limited long-term evaluation and weak cross-country comparability. Teacher interest is high, yet institutional guidance, assessment frameworks, and local-language resources remain uneven. This review argues that the next phase of adoption should move from isolated classroom experimentation to system-level implementation built around teacher AI literacy, transparent assessment, and context-sensitive design for smaller linguistic ecosystems. Full article

This paper examines the intersection between digital learning environments and multilingual education policies, with a focus on the linguistic integration of migrant students in Europe. It explores how technology, particularly mobile-assisted learning, artificial intelligence, and immersive tools, can strengthen language acquisition and promote social inclusion. Drawing on European and Greek policy frameworks, the study shows how digital pedagogies operationalize multilingualism as both an educational objective and a social justice priority. Based on a qualitative review of contemporary research and institutional reports, the findings indicate that digitally enhanced learning environments act as catalysts for equity, intercultural dialogue, and active participation when supported by coherent pedagogical design. The paper concludes by outlining policy recommendations for the development of multilingual digital ecosystems that align technological innovation with democratic, inclusive, and human-centred education. Overall, the analysis highlights that technology-mediated multilingualism can effectively reinforce participation, inclusion, and linguistic integration when embedded within robust policy structures and sound pedagogical practice. Full article

Early identification of at-risk students is crucial for timely pedagogical intervention. Determining which assessments instructors should prioritize is complicated by the fact that different eXplainable-AI (XAI) methods can produce conflicting rankings for the same predictive model. We develop a framework combining a sequential GRU model with two complementary XAI techniques, Gradient SHAP (attribution) and DiCE (counterfactuals), and evaluate it in a foundational Data Structures and Algorithms course. The framework produces predictions and explanations for every prefix length throughout the semester and quantifies inter-method agreement and intra-method stability using three disagreement metrics. Intersecting the top-k features identified by both methods isolates a compact subset of assessments whose predictive role is confirmed across two fundamentally different explanation mechanisms. We interpret this cross-method agreement as a heuristic that increases confidence in identified features relative to single-method results, though not as evidence of causal validity. For individual students, the framework uses the intersection of the two types of explanations when it is non-empty; otherwise, the instructor chooses between SHAP’s diagnostic view and DiCE’s prescriptive view, with an optional check against the top-k list. The resulting guidance is less susceptible to method-specific biases than analyses relying on a single method. Full article

Emotion is increasingly incorporated into AI-supported feedback in education, yet less is known about how learners interpret emotion-related messages once they are presented. This paper reports an exploratory classroom-based study comparing three learner-facing strategies for presenting emotion-aware feedback: inference with explanation, inference without explanation, and deliberate non-inference. Using a Wizard-of-Oz procedure embedded in a web-based classroom activity, 78 undergraduate students completed a conceptual quiz, a brief reflection task, and an applied data-analysis task during a 90-min course session. Following the activity, participants evaluated the system on six 7-point Likert outcomes: Perceived Accuracy, Interpretability, Emotional Comfort, Willingness to Reuse, Perceived Usefulness, and Trust. Significant differences were observed across all six outcomes. Across every dimension, the same ordinal pattern emerged: feedback with explanation received the highest ratings, no inference occupied an intermediate position, and inference without explanation was rated lowest. Notably, deliberate non-inference was evaluated more favorably than unexplained inference across all six outcomes. These findings suggest that the learner-facing value of emotion-aware educational AI depends not only on whether emotion is inferred, but on how such inference is presented and contextualized. The study contributes classroom-based evidence that learner interpretation should be treated as an important criterion in evaluating emotion-aware educational AI and that deliberate non-inference can function as a legitimate response strategy when affective claims cannot be presented in an intelligible and contextually grounded way. Full article

The study examines how secondary school students interacted with an AI-powered educational chatbot, MyBotBuddy, while working on a programming task, and how observed dialogue structures were associated with differences in pre- to post-test performance. Fifty students first completed an unassisted pre-test, then attempted a chatbot-supported programming task, and finally completed an unassisted post-test. Based on score change, students were grouped into learning gain, no gain, and learning loss categories. Dialogue transcripts were analyzed using Epistemic Network Analysis to identify co-occurring discourse patterns, alongside descriptive sentiment analysis to characterize lexical tone. Students in the learning gain group showed more connected multi-turn patterns involving solution attempts, feedback uptake, knowledge-related contributions, and clarification following feedback. In contrast, the no gain and learning loss groups showed less iterative and less systematically connected interaction structures. Average sentiment polarity differed only slightly across groups and is interpreted cautiously because the dialogue was technical and programming focused. The findings are associational and exploratory rather than causal and suggest that learner engagement with a chatbot may be more informative than interaction frequency alone. We discuss implications for educational chatbot design, especially the potential value of multi-turn scaffolding and reflective prompting, while outlining the need for future validation, baseline-controlled analyses, and experimental work. Full article

Rapid adoption of general-purpose generative AI (GenAI) tools, such as ChatGPT, is reshaping teaching, learning, and assessment in chemical education. In this study, we expanded the implementation of GenAI tools within an upper-level undergraduate biochemistry course, providing students access to four distinct platforms: commercial chatbots (ChatGPT and LearningClues) and in-house tools developed at the University of Michigan (U-M GPT and U-M Maizey). We analyzed student learning outcomes from GenAI-enhanced writing assignments using pre- and post-surveys. Our results show that integrating GenAI into biochemistry coursework promoted effective and responsible usage, enhanced students’ prompt literacy, built ethical awareness, and increased confidence in utilizing these tools. The study specifically examined factors influencing GenAI acceptance: familiarity, perceived usefulness, ease of use, and trust. Trust emerged as the most significant criterion, with a majority of students recommending in-house chatbots for future cohorts due to strong privacy and ethical standards. Over the last year, we observed a shift in student sentiment from excitement about efficiency to emerging concerns about creativity silencing. This highlights the importance of addressing both capabilities and risks of using AI-tools through teaching AI literacy. Full article

Learning is fundamentally human, even as Artificial Intelligence (AI) challenges human exclusivity. AI, along with Virtual Reality (VR), emerges as a powerful tool that is set to transform higher education, the institutional embodiment of this pursuit at its highest level. These technologies offer the potential not to replace the human factor, but to enhance our ability to create more adaptive, immersive, and truly human-centric learning experiences, aligning powerfully with the emerging vision of Education 5.0, which emphasizes ethical, collaborative learning ecosystems. This research maps how AI and VR tools act as a disruptive force, examining additionally their capabilities and limitations. Moreover, it explores how AI and VR interact to overcome traditional pedagogy’s constraints, fostering environments where technology serves human learning goals. Employing a comprehensive two-month audit of over 60 AI, VR, and AI-VR hybrid tools, the study assesses their functionalities and properties such as technical complexity, cost structures, integration capabilities, and compliance with ethical standards. Findings reveal that AI and VR systems provide significant opportunities for the future of education by providing personalized and captivating environments that encourage experiential learning and improve student motivation across disciplines. Nonetheless, numerous challenges limit widespread adoption, such as advanced infrastructure requirements and strategic planning. By articulating a structured evaluative framework and highlighting emerging trends, this paper provides practical guidance for educational stakeholders seeking to select and implement AI and VR tools in higher education. Full article

The widespread adoption of generative artificial intelligence in higher education has intensified debates around assessment, authorship, and academic integrity. This paper argues that such debates obscure a more fundamental pedagogical shift, namely, the decoupling of assessment performance from cognitive engagement. Historically, assessment functioned not only as a measure of learning, but also as a structural mechanism that implicitly enforced cognitive engagement. With the advent of GenAI, learners can increasingly produce assessment outputs without necessarily engaging in the cognitive processes traditionally associated with learning. As a result, cognitive engagement has shifted from being a pedagogical necessity to an intentional learner choice. This paper conceptualises this shift as the cognitive engagement gap, wherein successful assessment completion no longer reliably indicates learning or epistemic development. Through a theory-informed conceptual analysis, the paper examines how GenAI reconfigures learning processes, challenges the validity of assessment as a proxy for learning, and exposes long-standing assumptions embedded in assessment-centred pedagogies. In response, the paper proposes a Cognitive Engagement-Centred Assessment (CECA) framework, offering principled guidance for designing assessment that foregrounds cognitive processes, metacognition, and learning assurance in AI-mediated environments. The paper concludes by positioning GenAI not as a threat to assessment, but as a catalyst for more intentional, transparent, and learning-centred pedagogical design. Full article

Generative Artificial Intelligence (GenAI) is transforming higher education in terms of pedagogy, student involvement, and academic management. This systematic literature review examines 30 peer-reviewed articles published from 2019 to 2025, adhering to PRISMA 2020 and Kitchenham’s methodologies. Descriptive and thematic analyses highlight five opportunities: (a) tailored and adaptive education; (b) deliberate fostering of critical thinking; (c) enhanced accessibility for varied learners; (d) teaching innovation via multimodal content development and feedback; and (e) collaborative methods that regard AI as a co-teacher. Four ongoing challenge categories also surface: (a) risks to academic integrity; (b) excessive dependence on GenAI that may hinder learner independence; (c) inconsistent faculty preparedness and change-management abilities; and (d) differences in infrastructure and policy both regionally and globally. Intersecting ethical issues, such as data privacy, algorithmic bias, transparency, and accountability, highlight the necessity for governance that aligns with institutional risk and reflects societal values. Analyzing the recent literature, this systematic review offers four contributions: (a) a recommendation model for responsible GenAI implementation in higher education institutions; (b) a framework for sustainable integration of GenAI; (c) a highlight of the future research recommendations; and (d) an integrated policy and pedagogical recommendations roadmap. These models emphasize the integration of AI literacy, ethical considerations, and critical thinking goals into educational programs. The review advocates for a strategic, stakeholder-focused approach to implementation that enhances rather than replaces human instruction, thus connecting GenAI’s educational potential with ethical, context-aware avenues for institutional transformation. Full article

Generative artificial intelligence in higher education is typically framed as either a student productivity tool or an institutional disruption. This agenda-setting essay advances a third position: mobile generative AI functions as relational infrastructure—a persistent communicative presence that mediates identity, meaning-making, and belonging during institutional transition. Focusing on international graduate student onboarding, we abductively “think through” two complementary theoretical lenses. Constitutive Artificial Intelligence Identity Theory (CAIIT) conceptualizes AI as a co-constitutive participant in identity formation through recursive communicative feedback loops. Language Convergence/Meaning Divergence (LC/MD) theory explains how shared institutional language masks interpretive gaps across intercultural and bureaucratic contexts. Reading narrative vignettes through these frameworks, we argue that generative AI is neither simple curricular tool nor personal aid, but both relational and organizational infrastructure, redistributing translational, emotional, and interpretive labor in higher education. We outline four design principles for AI-integrated onboarding: distinguish communicative scaffolding from cognitive replacement; design systems that assume meaning divergence; center equity in AI-mediated transitions; and anticipate ethical risk. Reframing AI as relational infrastructure shifts AI-in-education research toward relational accountability and institutional care. Full article

This article contributes to current debates around the ethics of using AI in K-12 education by extending an ethical framework based on the constructs of wellbeing, autonomy and justice to examine how AI may differentially impact specific stakeholders. Data about K-12 building and district leaders’ perceptions of AI risks were collected during the 2023–24 school year in Western New York as part of an exploratory sequential mixed methods study, which included semi-structured interviews with a diverse group of 36 K-12 leaders, followed by a survey (n = 160). Survey findings confirm K-12 leaders’ widespread recognition, although at varying levels of concern, of AI risks related to (a) students cheating, (b) students’ other questionable AI uses, (c) educators’ questionable AI uses, (d) increasing inequities due to AI, (e) cybersecurity and privacy breaches, and to a much lesser extent, the (f) potential for job replacement. The ethical analysis reveals major differences in the implications of each of these six kinds of AI risk for the wellbeing, autonomy, and justice of K-12 educators, K-12 students, and society, respectively, as well as tensions between competing needs and values, which in turn call for risk-specific strategies as well as inevitable tradeoffs. A comparison with a study of musicians’ perceptions of AI using the same ethical framework reveals interesting similarities and differences in ethical concerns about AI in different fields, suggesting the value of more cross-disciplinary studies. Full article

Generative AI systems increasingly mediate how students retrieve literature and generate citations, shifting methodological rigor toward the maintenance of an auditable evidence trail. This study audits the search stage of AI-assisted literature review work, focusing on retrieval performance and citation traceability rather than downstream screening or synthesis. Four widely accessible tools were compared across two retrieval postures, and Boolean queries were executed against Scopus and evaluated against a DOI-verified librarian baseline built from Scopus, Web of Science, and Google Scholar. Using a canonical prompt and a bounded top-k capture rule (k = 20), each bibliographic record was evaluated for DOI traceability, DOI resolution integrity, metadata accuracy, and run-to-run drift. Records were screened through staged title/abstract and full-text eligibility review, and the final set included 37 studies after quality appraisal was 37 studies. Across sixteen audit runs, natural-language prompting frequently produced under-target yields, recurrent integrity failures, and low overlap with the librarian benchmark. Boolean translation improved run completion and increased the proportion of auditable records, but reproducibility remained unstable across repeated runs. These findings show that correctness at the record level does not ensure stability at the evidence-set level. Limitations include the bounded tool set, the search-stage focus, and the absence of downstream screening or synthesis evaluation. Retrieval posture, therefore, emerges as a practical governance lever for AI-assisted literature review workflows and supports the use of a student-facing verification checklist anchored in DOI verification and transparent protocol capture. This research received no external funding. OSF registration: Open Science Framework, 10.17605/OSF.IO/U8NHT. The manuscript reports the final included set as n = 37, states no external funding, and lists the OSF registration DOI. Full article

Generative artificial intelligence (GenAI) has rapidly emerged as a transformative educational technology, raising questions about how educators and pre-service teachers critically engage with AI-produced content. This case study investigates how WebQuests, a long-established, inquiry-based pedagogical model, can foster critical engagement with GenAI tools. Situated within an initial teacher education programme, a WebQuest, incorporating GenAI sources, was implemented with 24 pre-service language teachers, who engaged with curated resources alongside ChatGPT and Copilot to produce infographics for secondary school audiences. Data were collected through semi-structured interviews and were analysed using Braun and Clarke’s thematic analysis. Findings indicate that scaffolded engagement with GenAI encouraged participants to compare AI-generated outputs with trusted sources, critically evaluate accuracy and reliability, and reflect on integration into their future practice. Whilst pre-service teachers valued GenAI’s accessibility and efficiency, they expressed concerns about clarity, verbosity, and trustworthiness. The WebQuest model effectively supported synthesis of multiple information sources, fostering functional AI engagement and critical evaluation of its affordances and limitations. This case study concludes that integrating GenAI within structured, inquiry-based pedagogies advances digital and AI literacy in initial teacher education, whilst highlighting the need for institutional guidance, professional development, and further research in this area. Full article

Modeling the temporal dynamics of student learning is a central goal in educational data mining. Deep Knowledge Tracing (DKT) has emerged as a key approach, yet existing models are highly sensitive to out-of-distribution (OOD) inputs, such as those arising from curriculum changes, new assessment formats, or behavioral noise, which severely degrade predictive reliability. To address this challenge, we propose Energy-Based Out-of-Distribution Deep Knowledge Tracing (EB-OOD DKT), a unified framework that integrates energy-based uncertainty estimation and contrastive representation learning within a transformer-based DKT architecture. The model computes energy scores via the negative log-sum-exponential of prediction logits, serving as confidence indicators for detecting OOD inputs during inference. Additionally, an InfoNCE-based contrastive loss enhances representation robustness by aligning in-distribution samples and separating OOD cases in latent space. Temporal and behavioral context features, such as normalized response intervals and cumulative attempt counts, are incorporated to enrich cognitive-behavioral modeling. Experiments on four public educational datasets demonstrate consistent improvements in prediction accuracy and OOD detection. EB-OOD DKT provides a promising approach for more reliable student modeling across educational platforms with different content distributions. Full article

As artificial intelligence (AI) becomes increasingly integrated into higher education, instructors and institutions face urgent questions about its implications for teaching, learning, and scholarly practice as well as power, agency, and access. This study draws on a critical AI media literacy framework to analyze user-generated discussions in the two largest higher education subreddits on Reddit.com. Through thematic content analysis, I explore faculty perceptions, pedagogical tensions, and imaginative possibilities surrounding AI’s academic role in shaping the current and future landscape of higher education. Findings reveal that discussions of student cheating, AI policies, writing practices, and faculty labor are not merely technical debates but sites where surveillance regimes, accountability structures, and academic precarity are negotiated in real time. Ultimately, I argue that AI in higher education is not simply a technological shift but a structural transformation requiring deliberate, critically informed governance grounded in equity and human agency. Full article

Open-source large language models (LLMs) are increasingly explored in educational contexts due to their transparency, adaptability, and alignment with institutional governance and equity considerations. Despite growing interest, empirical research on how open-source LLMs are deployed in education and what evidence currently supports their integration remains limited and fragmented. This paper presents a state-of-the-art narrative review of peer-reviewed, human empirical studies examining the use of open-source LLMs in education. Guided by three questions, the review synthesizes how open-source LLMs are deployed across instructional contexts, what learner-related evidence is reported, and how teachers engage in human–AI collaboration. The reviewed literature is concentrated in higher education, particularly within computer science and programming domains, with applications focused on post-class tutoring, guidance, and formative feedback. Learner perceptions are generally positive, but evidence linking open-source LLM use to measurable learning outcomes remains emerging and inconsistent. Through interpretive synthesis, the review articulates a four-role model—Designer, Facilitator, Monitor, and Evaluator—that captures how teacher agency is enacted across AI-supported instructional workflows. This review maps recurring orchestration dimensions, decision points, and tensions that characterize early implementations, and it proposes a minimal orchestration reporting scaffold (configuration, boundaries, logging, adjudication) intended to support auditability and cross-study comparison as the empirical base develops. Full article

With the rise of social media and the sharing of information, as well as the use of AI tools like ChatGPT in education, the ability to evaluate information credibility has become a crucial skill. The CREDIBLE framework, standing for Credibility, Reliability, Evidence, Date, Intent, Bias, Logic, and Expertise, offers a practical, student-friendly approach to source evaluation, especially suited for secondary and postsecondary learners. Unlike models and frameworks designed for higher education, CREDIBLE helps learners critically assess both online and AI-generated content. This paper introduces the framework and explores how educators can embed it into instruction to foster critical thinking, academic integrity, and responsible digital literacy. Full article

This study provides direct evidence from university professors’ experiences regarding the key features they use to identify artificial intelligence (AI)–generated texts and ranks these features by their perceived importance. The research was conducted in two phases. In Phase 1, online interviews were used to identify the most salient features professors reported using to detect AI-generated texts. In Phase 2, an online survey asked professors to rate the extent to which each identified feature contributes to the successful detection of AI-generated text. The interview data yielded seven features that professors reported using when they suspected a text was AI-generated. Survey ratings varied across features, with hallucinated facts or explanations, nonexistent sources, and the absence of language errors receiving the highest mean ratings in this sample. The use of difficult words received the lowest mean rating. These results have important pedagogical implications, as they can inform the development of more effective detection tools and guide the design of academic integrity policies and instructional strategies to address the challenges posed by AI-generated content. Full article