Based on theoretical analysis, practical innovation and empirical evidence, I will talk about what future education may look like and how to address the concerns and challenges we are facing in education.
Biography
Dr. ZHANG Yu is Professor and Chair of the School of Education, Tsinghua University. Her research focuses on learning sciences and future education, and educational economics, integrating interdisciplinary innovations in cognitive neuroscience and artificial intelligence. She has special interests in learning theories, educational paradigm innovation, and methodological advancements in education research. She also serves as the Vice Chairman of the Education Evaluation Professional Committee, Chinese Society of Educational Development Strategy, Executive Council Member of the Learning Sciences Branch, China Association of Higher Education, Deputy Director of the Capital Education High-Quality Development Policy Research Base, and the Program Chair of the “Brain, Neuroscience, and Education” Special Interest Group, American Educational Research Association (AERA).
Innovating AI for Impact: People, Pedagogy, Purpose
ABSTRACT. Artificial intelligence and digital learning are reshaping higher education landscape, while unlocking new pathways for inclusion, engagement, and innovation. In response, universities worldwide have strategically embraced AI-powered Hybrid-Flexible (HyFlex) Learning, championing flexibility, personalisation, and multimodal access articulated by Beatty (2019). Grounded in the three pillars of People, Pedagogy, and Purpose, this symposium moves beyond theoretical promise to examine practical, human-centred AI implementation in real academic environments. Participants will engage with evidence-based approaches to AI-enhanced instruction, explore frameworks supporting diverse learners, and critically align AI with sound pedagogical principles. This symposium calls educators and academics to become active architects of AI in education, ensuring innovation serves people first, guided by strong pedagogy and purposeful outcomes for all learners.
Students’ cognitive and metacognitive interaction patterns with generative AI for scientific inquiry
ABSTRACT. This study investigates students’ interaction patterns with a customised chatbot using a conceptual–epistemic–social (CES) framework, distinguishing between the object-level cognition and metacognition. Data were collected from two secondary science classrooms involving two teachers and their students. Students interacted with the chatbot across two inquiry designs: one emphasises exploring relationships among variables, the other focuses on constructing and refining scientific models. Student chatlogs were coded and analysed using a descriptive analysis and epistemic network analysis (ENA) to examine both the prevalence and structural relationships of cognitive and metacognitive processes. We found that metacognition was more dominant than the object-level cognition overall, and epistemic metacognition occurred most frequently and showed strong across both inquiry approaches, widespread connections with other components, particularly epistemic cognition and social metacognition. This suggests that students progressed epistemic practices by orchestrating both cognitive and metacognitive interactions with the GenAI chatbot, while also metacognitively regulating the interactions.
Problem Statement Auditing in Engineering Design: An LLM-Based Socratic Scaffold
ABSTRACT. Solutioneering, or the premature focus on solution generation prior to problem exploration, is a persistent challenge in novice design practice, carrying significant pedagogical risk in project-based learning. While So-cratic dialogue can redirect misinformed problem framing by eliciting and examining assumptions, its application in large-class design courses is constrained by time and cognitive demands on instructors. This paper reports on the deployment of Laura, a pre-prompted LLM-based chatbot functioning as a Socratic problem-auditing tool in a second-year engi-neering design and innovation course. Drawing on 164 student–chatbot sessions and an inductively developed coding framework, we examine student engagement during assumption surfacing, problem-statement reformulation, and the conditions under which the chatbot maintained or drifted from its role. Findings suggest that pre-prompted LLM chatbots hold promise as scalable Socratic scaffolds in design education, while also revealing methodological and pedagogical boundary conditions, particularly around student agency, authorship, and the reliable evalua-tion of problem-statement quality.
AI as Scaffold, Not Shortcut: Preserving Productive Struggle in Inquiry-Based Science Learning
ABSTRACT. This design-based case study presents an AI-supported, gamified inquiry lesson on paper chromatography in Singapore’s Student Learning Space. Framed by a pedagogy-first approach, students act as forensic investigators using chromatography evidence to identify a suspect. AI is integrated through narrative hooks, game stories, simulations, chatbot-supported exploration, hints, feedback assistants, and learning analytics. These tools scaffold inquiry, preserve productive struggle, and support scientific communication rather than provide answers. Students remain responsible for observing, comparing evidence, interpreting chromatograms, and justifying conclusions, while AI provides timely support and feedback. The case study argues that effective AI-supported science education requires teacher-led orchestration, where technology strengthens motivation, scaffolding, formative assessment, and collaboration without replacing student reasoning or teacher judgment.
AI as a Pedagogical Partner in Integrated STEM Education
ABSTRACT. This paper reports a case study of a science-focused integrated STEM lesson in a Singapore secondary school, examining how a teacher purposefully incorporated multiple AI tools to scaffold scientific inquiry into local biodiversity, cultivate ethical reasoning about AI outputs, and develop cross-disciplinary thinking. The case is analysed through Tan's (2024) four dimensions — learning from, with, about, and beyond AI — to illustrate what meaningful, pedagogically informed AI integration looks like in practice and the conditions that enable it.
Co-constructing knowledge through human-AI collaboration in physics learning
ABSTRACT. This study examines how generative AI reshapes knowledge construction in a collaborative learning setting. Adopting a case study design in a high school physics classroom, the study analyses small-group interactions with AI, focusing on prompting, interaction processes, and their temporal development. Collaboration is conceptualized as a triadic system (student-student, student-AI, group-AI) with distributed knowledge construction. Video and screen recordings were analyzed using theory-driven coding and temporal timelines. The findings show distinct interaction patterns in how groups position, prompt, and critically engage with AI. Deeper epistemic processes, though less frequent, were linked to more substantive collaboration. The study highlights the value of process-oriented analysis and pedagogical designs that support shared prompting and critical engagement with AI.
Does Spatially Anchored AI Feedback Improve Learners’ Ability to Locate, Interpret, and Apply Feedback on Handwritten Work?
ABSTRACT. Feedback improves learning only when learners can locate, interpret, and act on it. This is a challenge in handwritten assessment, where AI-generated feedback is typically delivered as a separate text list, decoupled from the work it describes. This study examines whether spatially anchoring feedback to handwritten work improves the learner experience. Using a within-subjects design, ten secondary school students evaluated identical AI feedback in two formats: a text-list condition and a spatially-anchored condition, in which comments were overlaid directly onto the scanned work. Across a 12-item questionnaire measuring cognitive load, split attention, and UI/UX, the anchored format scored higher on every construct, suggesting that how AI feedback is presented may matter as much as its content.
From Content Delivery to Cognitive Orchestration: Reinterpreting AI-Empowered Python Instruction Through the MTA-TPACK Framework
ABSTRACT. This paper reinterprets a vocational engineering teacher's iterative use of generative AI (GenAI) for Python instruction as a principled instantiation of the MTA-TPACK Dynamic Collaboration Spiral. Using one instructional unit on Python container data structures as an illustrative case, we trace how Meta-Task Awareness (MTA) progressively activated static TPACK resources through a four-stage Chain of Learning Design and Evaluation (CoLDE). The culminating artifact—an interactive Python Playground—is interpreted as Visible Pedagogical Thinking (VPT). We argue that this process enables five paradigm shifts: from content delivery to epistemic orchestration, from scope-agnostic to professionally contextualized materials, from passive reception to active engagement, from summative to embedded diagnostic assessment, and from post-hoc to sandbox-embedded feedback—offering a transferable framework for AI-empowered technical and vocational education.
AI-Enhanced Morphophonemic Feedback for Tamil Writing: A Design-Based Study of Low-Resource Language Learning
ABSTRACT. Tamil writing pedagogy requires learners to coordinate orthography, grammar, and morphophonemic rules that are often difficult to practise through conventional correction alone. This design-based paper proposes an AI-enhanced feedback framework for Tamil morphophonemic writing, with particular attention to low-resource language constraints, pedagogical explainability, and teacher-controllable feedback. The proposed system combines linguistically grounded rule tests, synthetic learner-error generation, and large language model-assisted explanation drafting. Rather than positioning AI as an automatic grader, the framework treats feedback as a scaffold for noticing, revision, and metalinguistic awareness. The paper contributes design principles, an evaluation protocol, and implementation considerations for inclusive AI-supported Tamil language learning.
Making AI Feedback Meaningful: Teacher and Student Experiences in AI Augmented Concept Mapping
ABSTRACT. AI‑generated feedback is increasingly used to support formative assessment, yet how it is interpreted and enacted in classroom practice remains underexplored. This extended summary reports a pilot classroom study of AI‑augmented concept mapping in UK secondary Computer Science education, focusing on how teachers and students engage with automated feedback during learning activities. We find that AI feedback is often perceived by students as ambiguous or difficult to act upon, requiring teacher intervention to become pedagogically meaningful. Teachers play a critical role in interpreting, filtering, and contextualising feedback in relation to lesson goals and student understanding. These findings highlight that the value of AI feedback lies not in its generation alone, but in how it is mediated within classroom interaction, further reaffirming Teacher–AI complementarity.
Training Higher Education Students to Co-Create with Generative AI in Creative Problem Solving: The Role of Professional Identity and Motivation
ABSTRACT. Current discourse on generative AI (GAI) in higher education is dominated by concerns over academic integrity, with limited attention to its creative and pedagogical potential. As GAI becomes embedded in professional practice, collaborative creative problem solving with GAI is emerging as a relevant capability for graduate outcomes. Using a qualitative design, this study trained undergraduate engineering and education students to co-create with GAI using a structured creativity framework to scaffold intentional interaction. It examines how training influences co-creative behaviour and how professional identity and motivation shape these processes. Findings indicate limited meaningful engagement with GAI. Although students identified as creative problem-solvers, this identity did not translate into authentic collaboration with GAI. Instead, compliance-oriented task framing and cognitive fatigue reduced engagement quality, resulting in surface-level, copy-and-paste use. Overall, the findings suggest that training alone is insufficient to foster meaningful human-AI co-creation without addressing identity enactment, motivational and task conditions.
What Are We Trying to Protect? A Five-Stage Design Model for Whole-School AI Implementation
ABSTRACT. This paper presents a human-centred, five-stage design model for whole-school AI implementation developed at an IB boarding and day school in Southeast Asia over seven months (September 2025–April 2026). The model anchors AI adoption as a means of protecting and developing five core human competencies: Critical Thinking, Creativity, Collaboration, Compassion, and Communication. A 238-stakeholder baseline established community state before any intervention was designed; five sequenced stages followed. A whole-staff PD session (n=48 entry, n=49 exit) produced a +0.76 confidence shift and a +0.50 mean fluency shift (one full level) within 60 minutes; low-confidence staff fell from 40% to 4%. A concurrent Grade 9 unit saw 11 multilingual students build and teach AI-literacy tools across five year groups, rated 4.75 out of 5 at showcase.
Rethinking Technology Acceptance in Generative AI Learning Contexts: Students’ Perspectives from Higher Education
ABSTRACT. Generative artificial intelligence (GenAI) tools such as ChatGPT, Gemini, Claude, and Grammarly are increasingly used by students to support learning, revision, and assessment preparation in higher education. This study re-examined the Technology Acceptance Model (TAM) in GenAI learning contexts, focusing on the relationships among perceived ease of use (PEOU), perceived usefulness (PU), and actual use (AU). A quantitative survey involving 561 university students from multiple disciplinary clusters was conducted. Structural equation modelling demonstrated acceptable model fit and showed that PEOU significantly predicted PU. However, PU negatively predicted AU, whereas PEOU did not. The findings suggested that students recognised the usefulness of GenAI while simultaneously exercising restraint in AU, possibly due to concerns regarding academic integrity, overreliance, and educational authenticity. The study highlighted the need to rethink traditional assumptions about technology acceptance in AI-mediated educational environments, while accounting for ethical, disciplinary, and institutional considerations.
Artificial Intelligence as the New Administrator: Transforming Educational Governance in South Asian Universities
ABSTRACT. Artificial Intelligence (AI) is swiftly transforming higher education globally; however, its influence on university governance in South Asia is insufficiently examined, as institutions grapple with issues of efficiency, accountability, and digital preparedness. This study examined the impact of AI on governance practices in South Asian universities utilizing the Technology–Organization–Environment (TOE) framework, Transformational Leadership Theory, and the Input–Process–Output (IPO) model. A qualitative phenomenological methodology was used, incorporating semi-structured interviews with seven senior administrators from both public and private universities. Thematic analysis indicated that AI improves decision intelligence, administrative efficiency, transparency, accountability, and evidence-based leadership. Nonetheless, infrastructural inadequacies, financial constraints, insufficient technical proficiency, and ethical issues related to privacy and algorithmic bias hinder adoption. This study enhances the literature by framing AI as a strategic governance facilitator rather than simply a technological instrument, offering one of the initial context-specific insights into AI-driven university governance in South Asia.
Design-Based Study of a Multimodal Learning Analytics Phonics Prototype
ABSTRACT. Scholarship about AI in education blossomed and gained traction as large language models proved its widespread applications in daily and professional contexts. Empowering primary language teachers to fine-tune culturally responsive small language models for ethical, bilingual and personalised instruction across Singapore’s four official languages is realistic. This positions multi-modal learning analytics for its next iteration as personalised and adaptive learning is boosted by algorithms and compute power. Language teaching is at a pivotal moment. Artificial intelligence can handle the mechanics of language learning such as grammar drills, vocabulary scaffolds, comprehension practice, and diagnostic feedback. Thus, AI-driven scaffolds accelerate language learner progression, freeing primary language teachers to focus on higher-order skills. Privacy-first personalised AI could be a smart tutor, enabling primary language teachers to dedicate their energy to creativity, character building, and authentic communication. At the same time, it could also advance equitable, bilingual, and dyslexia-inclusive education in Singapore and ASEAN.
Learning to Prompt: Evaluating an Experiential AI Workshop Through LLM-Based Prompt Quality Assessment
ABSTRACT. We examined prompt quality as an indicator of students’ ability to interact effectively with generative AI. Using data from a one-day AI workshop grounded in experiential learning theory with first-year university students in Singapore (N = 434), we evaluated student-written prompts using an LLM (GPT-4o-mini) across five repeated runs. Results demonstrated high reliability, significant pre-to-post improvements in prompt quality, and consistent gains regardless of students' prior AI experience. These findings suggest that rubric-guided LLM assessment offers a reliable method for evaluating prompt quality, and that experiential learning–based workshops can support students' prompting skill development.
ABSTRACT. There is a pressing need to adopt AI both ethically and responsibly. The rapid rise in artificial intelligence (AI) has created many opportunities globally, from facilitating healthcare diagnoses to enabling human connections through social media and creating labor efficiencies through automated tasks. However, these rapid changes also raise profound ethical concerns. These arise from the potential AI systems have to embed biases, contribute to climate degradation, threaten human rights and more. In the field of education, the impact of AI is even more profound and long-lasting as it could lead to excessive dependency on AI-based technologies as well as conditioning the young minds with automated responses and texts that can stifle their creativity & imagination and can also shape their behaviors. The present paper aims at developing an ethical framework for integrating AI in education so that AI can have a positive impact on education and the stakeholders it impacts.
Developing Ethical, Transparent, and Human-Centered AI Governance Ecosystems for Quality Assurance and Sustainable Transformation in Higher Education
ABSTRACT. This conceptual article proposes a governance approach that organically links quality assurance and sustainable transformation amidst the expanding use of artificial intelligence (AI) in higher education. The author identifies the disconnect between fragmented ethical principles, technological processes, and institutional policies as a research gap. Based on an argumentative synthesis, a multi-layered "AI governance ecosystem" model is developed, which integrates ethics, transparency, and human-centricity into a unified system. The model connects the normative framework, institutional governance, operational practices, and human oversight through a continuous quality assurance loop. Five propositions and a table of practical indicators are presented. The findings indicate that the responsible implementation of AI must be an integral part of a quality culture.
Ethical AI Literacy Under Pressure: A Reflective Practitioner Analysis of Discipline-Sensitive Assessment Design
ABSTRACT. This paper examines ethical AI literacy as an assessment design challenge in higher education, arguing that current pedagogical responses remain underdeveloped despite rapid institutional uptake of generative AI policies. Drawing on reflective practitioner analysis across two semesters at a technology-focused university, it investigates how ethical AI requirements are differently enacted in social science and STEM-adjacent courses. It finds that practices such as disclosure, verification, and ethical justification are not equally transferable across disciplinary contexts and become especially fragile under deadline pressure. In response, the paper proposes the UJTVS Framework Understanding, Judgment, Transparency, Verification, and Stress Resilience as a discipline-sensitive model for embedding ethical AI engagement within assessment design. Illustrated through two course cases, the framework repositions ethical AI literacy as a structural and pedagogical issue rather than primarily a matter of student compliance. The paper concludes by identifying priorities for future empirical research.
Ethical AI Use in Mathematics Education: Preliminary Survey Evidence from Mathematics Teachers and Mathematics Teacher Educators
ABSTRACT. Generative artificial intelligence (AI) is reshaping mathematics education, raising questions about ethical use, classroom policy, and student reasoning. This exploratory study examines how mathematics teachers and mathematics teacher educators define and use AI in teaching contexts. Open-ended survey responses from 25 mathematics educators, including two mathematics teacher educators, were analyzed thematically with descriptive counts. Findings show that most participants defined ethical AI use as a “tool, not a crutch” that should support rather than replace thinking (18/25), while many reported no formal AI policy in their schools or institutions (19/25). Most participants used AI for lesson preparation (22/25), but over half noted mathematical errors or formatting problems (13/25). The two mathematics teacher educators further framed ethical AI use as a teacher-preparation issue involving professional responsibility, transparency, equitable access, and preservation of student reasoning. The study argues that ethical AI integration requires transparent policies, teacher preparation, and mathematics-specific guidance.
ABSTRACT. This paper presents CoMPA(I)SS, a work-in-progress framework for trustworthy learning analytics and explainable AI in education. Building on Koh et al.’s (2025) SWOT analysis of existing frameworks, it addresses gaps in credibility, interpretability, and socio-technical coverage. Updated literature and policy perspectives extend the framework to include social equity, cognitive atrophy, human oversight, and ecological impact. CoMPA(I)SS positions educational AI as a socio-technical system that should complement, rather than replace, human judgement. It treats transparency as assessable and meaningful for diverse stakeholders and incorporates eco-impact assessment alongside educational value. Overall, CoMPA(I)SS offers a basis for future framework development, validation, and responsible innovation in technology-rich education.
Self-Regulation in the Intelligence Era: Predicting Academic Procrastination in AI-Enhanced Learning Environments
ABSTRACT. In the contemporary intelligence era, higher education faces a digital paradox characterized by expanded autonomous learning opportunities alongside increasing online distractions. This study examined the relationships among self-regulation, cyber-loafing, and academic procrastination among higher education students in India (N = 438) using a quantitative correlational design. Findings revealed that Self-Regulation significantly predicted lower Academic Procrastination (β = .34, p < .001), whereas cyber-loafing did not emerge as a significant predictor. Interestingly, a weak positive correlation between self-regulation and cyber-loafing suggests that highly self-regulated students utilize digital breaks as restorative buffers to manage cognitive fatigue. Significant differences were observed across gender and academic programs, with female students demonstrating stronger regulatory competencies. The findings highlights that internal behavioral regulation the "will and skill" of the learner is more influential than digital exposure itself. Institutions should therefore strengthen self-regulatory competencies and promote hybrid human-AI regulatory strategies to enhance academic engagement in digital ecosystems.
The Guidance Paradox: Trust, Emotional Reliance, and Generative AI Use in Urban Indian Schools
ABSTRACT. Generative AI tools such as ChatGPT and Gemini have entered school life in urban India faster than institutions have adapted to them. This study examines trust in AI, emotional reliance, and institutional guidance through a multi-stakeholder survey of 166 participants — 48 students, 15 teachers, and 103 parents — from urban private schools in India. Findings showed near-universal student AI adoption, with almost half using AI daily. Students increasingly viewed AI as more understandable and less biased than human feedback, while many also reported emotional reliance on AI during stress. A major divergence emerged between teacher-reported and student-perceived guidance, termed the Guidance Paradox: 70% of teachers believed they guided students on responsible AI use, while only 20% of students reported receiving meaningful guidance. Cluster analysis identified an At-Risk Reliers profile characterised by heavy AI use, reduced academic independence, and elevated emotional reliance.
MEDIA-Q: A Parent-Report Measure of Children's genAI and Digital Media Use and Associated Parenting Practices
ABSTRACT. The rapid emergence of genAI has created new challenges for measuring young children's digital media experiences. Existing parent-report instruments assess screen time, content, and parental mediation of traditional digital media but do not adequately capture children's exposure to and interactions with genAI. This study describes the development and pilot testing of the Media Exposure and Digital Interaction with genAI Questionnaire (MEDIA-Q), a parent-report measure assessing children’s and parents’ use of genAI, entertainment, educational, and communication media, alongside parental attitudes and mediation practices. Pilot data were collected from 50 Singaporean parents of preschool-aged children. Results indicated limited direct genAI use among children but substantially higher use among parents. Internal consistency was generally acceptable across subscales. Parents reported significantly less positive attitudes toward genAI and entertainment media than toward educational and communication media. MEDIA-Q provides a novel tool for examining children's digital media ecologies and parenting practices in the age of genAI.
The Association Between Loneliness and Emotional Support Experiences with Generative AI: The Mediating Role of AI Attachment
ABSTRACT. As generative artificial intelligence (GAI) becomes increasingly embedded in adolescents’ daily lives, some middle school students may use it not only for learning support but also for emotional expression and comfort seeking. Grounded in a compensatory surrogacy perspective and attachment theory, this study examines the association between loneliness and emotional support experiences with GAI, with AI attachment as a mediator. A two-wave longitudinal survey design with a three-month interval will be adopted among Chinese middle school students. Measures will include the UCLA Loneliness Scale, the AI Attachment Scale, and a GAI Emotional Support Experience Scale. Expected results indicate that loneliness will be positively associated with AI attachment, AI attachment will be positively associated with emotional support experiences with GAI, and the mediating effect of AI attachment will be significant. The study extends current GAI research from learning support to adolescents’ emotional support experiences.
Emotion-Aware Formative Assessment Dashboard for Teacher Orchestration in Productive Failure Classrooms
ABSTRACT. Productive Failure asks students to solve complex problems before formal instruction, but failed or suboptimal solutions do not automatically lead to deep understanding. The teacher-led consolidation stage is critical because teachers compare, organize, and connect students’ initial solutions with formal concepts. Yet teachers often lack access to students’ emotional trajectories during problem solving, such as confusion, frustration, doubt, or engagement. This extended summary proposes an emotion-aware formative assessment dashboard to support teacher orchestration during consolidation. The dashboard presents class-level emotional trends, connects emotional signals with cognitive patterns, suggests possible consolidation actions, and keeps AI-based inference transparent for teacher calibration. As a conceptual design work-in-progress, this paper connects Productive Failure, teacher orchestration, learning analytics dashboards, and emotion-aware formative assessment.
Multi-Platform Generative AI Integration in Engineering Mechanics Education: A Case Study in Maritime Engineering Pedagogy
ABSTRACT. This paper explores the use of multi-platform generative AI in technical engineering education, focusing on the Engineering Mechanics (MA1118) module at Singapore Polytechnic’s Singapore Maritime Academy. Three AI tools (ChatGPT, Claude.AI, and DeepSeek) were integrated into tutorial and practical sessions to support student learning.
Findings show that when applied in a structured manner, these tools enhance conceptual understanding, improve problem-solving skills, and strengthen collaboration among students. The paper also addresses concerns around academic integrity and proposes a practical framework for responsible AI use in the classroom. By aligning specific AI platforms to different learning tasks, student engagement and learning outcomes can be improved while fostering essential AI literacy for Industry 4.0.
This study contributes to ongoing discussions on technology-enhanced learning and offers practical guidance for educators integrating GenAI into calculation-intensive modules.
Mapping Algebra Concept Connections in Indonesian Mathematics Textbooks: Foundation for Adaptive Learning
ABSTRACT. This study maps mathematical concept connections within algebra problems in Indonesian junior high school mathematics textbooks under the Merdeka Curriculum. Using a mixed-methods exploratory sequential design, 1,056 algebra problems from government-approved Grades 7–9 textbooks are coded for connection type (within-concept, between-concept, or no connection) and directionality (typical, reverse, or bidirectional). Social Network Analysis (SNA) constructs directed concept networks characterised by density and centrality metrics. The resulting knowledge graph serves as an evidence base for curriculum evaluation and a structured data foundation for AI-powered adaptive learning systems that personalise mathematics instruction.
Designing AI for Music Creating in a Human-AI World: Reframing Generative AI as a Cognitive and Epistemic Partner
ABSTRACT. Generative AI in education is often framed as a productivity tool or personalised tutor. In music education, it is also seen as a possible threat to creativity. This presentation proposes a different view: generative AI as a cognitive and epistemic partner in music creating. Rather than simply accelerating production or supplying answers, AI can shape musical thinking by prompting ideas, supporting reflection, surfacing alternatives, and challenging aesthetic and compositional choices. Building on posthuman and sociocultural perspectives, the paper argues for AI-supported music learning environments that foreground co-creation, judgement, and reflexivity. It suggests that when grounded in domain-specific sources and authentic creative processes, AI can function as a provocateur, feedback partner, and reflective interlocutor. It will also address questions about authenticity, curatorial bias, disciplinary socialisation, and whose musical knowledge is amplified or excluded. The presentation contributes a framework for human–AI creative partnership in education.
ABSTRACT. The proliferation of Artificial intelligence (AI) in higher education has widened a critical gap for students, the lack of practical scaffolding on responsible use of AI for their academic work. Current frameworks serve mainly policymakers and developers, not students. This paper, therefore, proposes SMARTE - Scope, Meaning, Accountability, Responsibility, Transparency and Ethics, a student-centric framework grounded in Self-Regulated Learning theory (Zimmerman, 2002) that scaffolds planning, monitoring, and ethical reflection when using AI for academic work. Methodologically, a design-based approach will be adopted and the SMARTE checklist will be piloted for six courses. It is hoped that SMARTE will contribute as a practical instructional tool that moves beyond AI policy toward AI pedagogy.
ABSTRACT. AI based apps that are available today follow the paradigm: Teacher submits a prompt, App provides solution. A better paradigm is to look at an AI app as an assistant and use a paradigm where the teacher and the app repeatedly interact during the creation stage using intermediate forms of the final product. I discuss the workflow of three apps that I have created: Kathakar for creating stories, Prastutkarta for presentations and PrashnaMitra for question sets.
In this talk, Dr. Zhai will introduce a comprehensive framework for discipline-based AI literacy (DAIL), which emphasizes integrating AI competence within the context of specific subject areas such as science and engineering. Rather than treating AI as a standalone topic, DAIL focuses on helping learners meaningfully apply AI concepts to disciplinary practices, fostering deeper conceptual understanding and critical thinking. To provide concrete evidence of DAIL in action, Dr. Zhai will demonstrate a multi-agent system called GenAgent. This system is designed to actively engage students in authentic science and engineering practices, such as modeling, data analysis, and problem-solving. Through interactions with multiple intelligent agents, students collaboratively explore complex phenomena while simultaneously developing AI literacy. The talk will highlight how GenAgent supports both content learning and AI competency, offering a scalable and innovative approach to preparing students for an AI-driven future.
Biography
Xiaoming Zhai is Professor of Science Education and AI, (courtesy) Professor of Computer Science and Statistics, Founding Director of the AI4STEM Education Center, and Director of the National GENIUS Center, University of Georgia. His research pioneers the use of AI to transform science and STEM education, supported by NSF, IES, NIH, and NAEd/Spencer Foundation. He has published extensively in leading journals and premier AI and AI-in-Education venues, including Journal of Research in Science Teaching, Computers & Education, AIED, LAK, AAAI, NeurIPS, KDD, and ICLR. Dr. Zhai chaired the 2022 and 2025 International Conferences on AI in STEM education, leads NARST’s RAISE group, and has edited four books and eight special issues. His research has earned prestigious honors such as the AERA TACTL Early Career Award, Humboldt Research Fellowship, and NAEd/Spencer Award, establishing him as a leading voice in shaping the future of AI in STEM education.
From Institutional Support to Classroom Practice: Advancing AI-Enabled Teaching in Higher Education
ABSTRACT. This symposium brings together scholars and practitioners engaged in AI-enabled teaching to examine how institutional, pedagogical, and evaluative dimensions of AI integration can be aligned within a unified professional development framework. Drawing on Dogan's (2026) Integrated Pathways for Teacher Growth model, the three presentations share common ground in moving beyond fragmented, tool-driven AI adoption toward purposeful, evidence-informed practice. Despite addressing distinct domains: institutional ecosystem development, expert coaching through discipline-specific use cases, and structured learning evaluation, the three presentations are united by their focus on how professional development can be designed as a connected system rather than isolated interventions. Examples include faculty-facing AI literacy resources, custom AI applications for legal writing, AI avatars for personalised feedback, and a survey instrument grounded in Bloom's revised taxonomy and Self-Determination Theory. The role of institutional support structures in enabling and sustaining these practices will be explored.
Pedagogy First: Fostering Critical Thinking Through Responsible AI Integration - Lessons from a Teacher Professional Development Programme in Portugal
ABSTRACT. This paper reports on a structured teacher professional development programme delivered at Colégio Alfacoop, Portugal, during the 2025/2026 academic year, engaging 143 teachers and educators in the responsible, pedagogically grounded integration of Artificial Intelligence (AI). The programme, structured across 25 hours in a blended learning format, was organised around three interconnected priorities: redefining the teacher’s role as a critical mediator of learning; equipping educators to cultivate students’ critical thinking in AI-rich environments; and promoting inclusive and ethical AI use as non-negotiable design conditions. Drawing on a pedagogy-first orientation, findings indicate that structured professional development can shift practice from tool-centred adoption towards approaches in which learner agency, intellectual interrogation and knowledge co-construction remain central. The paper contributes a replicable model and a set of pedagogical propositions for the field.
Consequential Validity of AI-Native Assessment in Professional Hiring: Evidence from Legal Recruitment Interviews
ABSTRACT. Generative AI (GenAI) enables automated assessment of interactional competence (IC) through spoken dialogue systems, yet the validity of the interpretation and use of assessment results in high-stakes professional contexts remains underexamined. This study investigates AI-mediated legal job recruitment interviews as a site of professional IC assessment, focusing on consequential validity. Data comprise 13 video-recorded English interviews between an LLM-driven AI interviewer and early-career lawyers, complemented by expert interviews with legal professionals. Conversation Analysis reveals that expert-informed hiring decisions show partial alignment with AI-generated hiring recommendations. Discrepancies suggest culturally patterned bias, raising fairness concerns for candidates from diverse cultural backgrounds. These findings highlight risks associated with AI-native hiring operating systems and and call into question the validity of extrapolation from AI-mediated interview performance to real-world professional interaction. They further point to the need for more transparent, culturally responsive AI assessment to better support professional evaluation and learning in high-stakes contexts.
From Tools to Dialogues: A Contextualised and Dialogic Approach to GenAI Professional Development
ABSTRACT. Professional development (PD) for educators has its fair share of challenges; however, the sudden emergence of generative AI has created an urgent need to go beyond tool use to support meaningful, contextual integration into teaching, learning and assessment.
This extended summary presents Temasek Polytechnic’s three-module GenAI PD programme as a contextual and dialogic approach to educator learning. Steering away from a one-size-fits-all training, the programme emphasised contextualisation, co-creation and dialogue. Leveraging the contextual knowledge from school-based academic mentors (SAMs), staff developers co-designed and co-facilitated workshops with discipline-relevant use cases of GenAI, fuelling discussions and critique exercises.
Over 97% of staff completed the program, reporting increased confidence in using AI to foster deeper learning, heightened ethical awareness and evidence of assessment redesign to maintain academic rigour.
The initiative suggests that GenAI professional development is engaging when situated within educators’ contexts and shaped through collaborative dialogue and reflection.
Preparing Indonesia’s Teaching Workforce for AI: Rethinking Towards Required Teachers’ Competencies
ABSTRACT. Global data from the 2024 Teaching and Learning International Survey (TALIS) results show that a large portion of teachers use AI for different purposes. However, this survey also reported that three in four teachers lack knowledge in using AI for teaching. Therefore, in this policy-oriented paper, we aim to examine the extent to which Indonesia needs to redefine teachers’ competencies that correspond to required students’ skills in the presence of AI. We analyze current teachers’ AI competencies frameworks from the UNESCO and OECD, and compare other countries’ AI frameworks to guide our recommended teachers’ AI competencies framework for Indonesia context. This study may benefit teacher educators and education policy makers in Indonesia.
Create-Check-Critique-Refine (C3R): Using AI Assessment Evaluator to Develop Preservice Physics Teachers’ Assessment Literacy
ABSTRACT. Student teachers routinely design assessments but rarely test whether their tasks elicit the intended reasoning because they lack actual learner responses. This work-in-progress presentation introduces the Create–Check–Critique–Refine (C3R) model, supported by an AI Assessment Evaluator that simulates differentiated low-, mid-, and high-performing responses to teacher-designed physics assessments. The AI produces response evidence that student teachers interrogate against the Assessment Triangle, linking cognition, observation, and interpretation. Peer critique and instructor modelling support evaluative judgement, while AI responses provide timely feedback on design quality. The model will be piloted with preservice physics teachers in late 2026. We outline the design rationale, the C3R workflow, the planned evaluation, and implications for pedagogically grounded, ethical use of generative AI in teacher education.
Development and Evaluation of Gene: An AI Chatbot for Personalized Learning of Mendelian Inheritance
ABSTRACT. Mendelian inheritance is foundational and perennially difficult, with students at both basic education and pre-service teacher training levels demonstrating persistent conceptual and procedural difficulties. This study developed and evaluated Gene, an AI-powered chatbot for personalized learning of Mendelian inheritance, using a descriptive-developmental design guided by the Design Thinking Framework. The study aimed to design Gene from empirical diagnostic data and evaluate its usability and quality from user and expert perspectives. Achievement tests revealed low to average performance among Grade 8 students and pre-service science teachers, with six recurring error categories confirmed through error analysis. Gene received an Excellent Chatbot Usability Questionnaire score (82.3) and Very Satisfactory expert ratings across all DepEd LRMDS dimensions; visual presentation was identified as the primary area for improvement. Gene demonstrates promising proof-of-concept value as a pedagogically scaffolded, domain-specific AI chatbot. Future research should address its classroom effectiveness, scalability, and alignment with national AI education frameworks.
Integrating AI tools in Work-Integrated Learning Online Micro Lessons to enhance Student engagement.
ABSTRACT. This is a conceptual paper aimed at exploring and understanding how student teachers integrate AI tools into their Work-Integrated Learning Online Micro Lessons to enhance student engagement. This paper is a qualitative study working from an interpretivist perspective with an explorative research design. The data that will be used will be a Google Form link, which will be in the form of discussion forums and reflective journals. This study will be encapsulated in a conceptual framework that will be embedded in four types of engagement, which include: behavioural, cognitive, emotional, and social engagement. This study looks at long-term sustainable ideas to engage students, especially on online platforms like Google Meet, Microsoft Teams, and Zoom.
Interviewing with the Machine? Student Perceptions of Humanlike AI Avatars in Virtual Communication Training
ABSTRACT. This study investigates how Hong Kong secondary school students perceive a humanlike AI avatar used for mock university admission interviews. Following online workshops led by undergraduate mentors, 115 Secondary 5 students completed an AI interview, and 69 participated in face-to-face reflection sessions. Three key themes emerged. First, students approached the AI with anxiety and skepticism, expecting a basic "smart assistant." Second, they questioned the educational value of automated scoring, viewing the feedback as rigidly focused on linguistic clarity while ignoring context, tone, and nonverbal reciprocity. Third, visible recording cues triggered feelings of surveillance, forcing an appearance-focused "performative labor" that distracted from message quality. These insights suggest a hybrid sandbox model and transparent data protocols are needed to shift the student experience of AI from "watched" to "supported."
ABSTRACT. Students at the Institute of Technical Education (ITE) often struggle to learn complex industry software such as Unity game engine. These tools require students to manage programming, interface navigation, and design workflows at the same time, which can be overwhelming. Although tutorials and online resources are available, students often spend too much time troubleshooting errors and searching for solutions. This disrupts their learning and makes it harder to understand key concepts, especially when working independently outside of class. This highlights the need for an AI-based solution that can provide immediate, contextual support during self-directed learning.
Iterative Designs of an AI-Coding Hackathon Program to Cultivate Young Students’ AI Literacy
ABSTRACT. This study reported on an iteration of an AI-coding hackathon program to cultivate young students' AI literacy. In total, 54 Singapore primary school students participated in the program. Based on the analysis of students' reflection notes, post interviews, and the program design log data, the results show that in the consistent improvement of the three iterations, from AI-powered chatbot, to Python learning, to AI-powered vibe coding, there was an increase in students' participation and learning experiences. The results of this study pilot a new curriculum design for AI literacy cultivation and a transferable design prototype for future K-12 AI education.
Students' Agency in Educational vs. General-Purpose GenAI Agents: An Exploratory Study in Entrepreneurship Education
ABSTRACT. GenAI conversational agents (CAs) are becoming more common in higher education, with many institutions developing purpose-built solutions. However, the comparison between educational CAs and general-purpose ones is underexplored, particularly regarding agency, an area of growing importance in AI for Education. In this month-long mixed-methods study, 19 Entrepreneurship master students interacted with Gaia, an educational CA for learning about sustainable business cases, alongside general-purpose CAs of their choice, as part of a course project. Using pre- and post-questionnaires, an interaction diary, and an in-class discussion, we analyzed usage patterns and task diversity, perceived sense of agency, and usage reflections. Findings show that students apply CA orchestration behaviors that involve deciding which tools to use, how to use and move between them, and how to shape the direction of conversations and end them at will. In this, the educational chatbot Gaia was perceived as affording less agency than general-purpose CAs.
When the Machine Mimics the Form but Misses the Meter: Students' Judgment and Verification in LLM-Assisted Classical Poetry Learning
ABSTRACT. This preliminary study examines how undergraduate students construct literary knowledge while co-writing regulated Chinese verse (five-character lüshi) with a large language model (LLM). We paired a stronger model (DeepSeek-V4) with a demanding genre and added a prompt-turn constraint inspired by the "seven-step verse" allusion: one task left turns unlimited, the other capped them at seven. Analyzing two students' interaction logs, turn by turn, we observe that the model satisfies explicit rules (tone, rhyme) yet silently violates unstated ones (parallelism), and, during self-checking, may hallucinate, recasting genuine defects as legitimate "tone-rescue" or "loose parallelism." Students' aesthetic sensitivity often outran their prosodic knowledge. Critically, this gap made the more knowledgeable student more vulnerable to the model's pseudo-technical defenses, not less—partial expertise supplied just enough vocabulary to be persuaded, but not to resist. We discuss implications for fostering critical, prosody-aware human–AI collaboration.
Measuring to Assess Student Agency in AI-mediated Learning Systems
ABSTRACT. As AI‑mediated learning systems increasingly guide, evaluate, and redirect student activity, student agency risks becoming a normative aspiration rather than an operational concern. This paper argues that student agency must be made assessable if it is to be meaningfully protected. Drawing on international policy frameworks and recent conceptual work, we position agency as an interactional and multidimensional outcome shaped through assessment and feedback practices. We identify key agency‑relevant decision spaces and articulate a research direction for examining how AI systems structure, render legible, and distribute evaluative authority across students, teachers, and automated processes. The contribution is a conceptual lens that reframes student agency as a measurable and contestable property of AI‑mediated learning systems, informing the design and governance of educational AI.
Illusions of Control? Sensed and Enacted Student Agency in AI-mediated Learning
ABSTRACT. Students using generative AI in educational settings may feel highly autonomous (e.g., choosing prompts, requesting help, or steering dialogue) while their learning trajectories are increasingly shaped by system steering (suggestions) they do not fully understand or meaningfully influence. We conceptualise this mismatch as a tension between sensed (perceived) agency and enacted (actual) agency in AI‑mediated learning. Drawing on international education policy and learning‑design scholarship, and using children’s interactions with AI as a motivating case, we argue that perceived agency alone is insufficient for student (learner) empowerment. We outline the implications for learning design and classroom practice.
Becoming Critical Human-AI Hybrids: Insights from a pedagogical innovation
ABSTRACT. The emerging field of ‘hybrid intelligent systems’ adopts an augmentation instead of a replacement paradigm and AI is considered a collaborative partner within higher-level intellectual hybrid systems. In this context, we present a practical and research-based pedagogical innovation that transitions students from passive ‘AI-adopters’ to ‘critical human-AI hybrids’. The study is based on extensive fieldwork with 450+ students at University College London in 2025 and 2026 who took part in a 5-week ‘Sprint Project’. Teams collaborated with digital and AI tools to design and develop new solutions for their client organisations. A multi-method data collection approach was adopted including individual learning reflections, weekly questionnaires, and observations of students’ ‘live’ teamwork. We contribute with the Critical Human-AI Hybrid Framework, which shows how hybrid activities take place, including human pushback, agency negotiations, and cognitive augmentations that are all part of the evolutionary process of becoming a critical human-AI hybrid.
From Oracle to Tool: Demythologising AI in Higher Education
ABSTRACT. Contemporary discourse surrounding artificial intelligence in higher education frequently oscillates between utopian promise and existential fear. In both cases, however, AI is increasingly framed not merely as a technological tool, but as a quasi-transcendent force possessing exceptional authority, inevitability, and agency. This paper argues that such discourse mythologises AI by obscuring the situated, contingent, and fundamentally instrumental nature of computational systems. Drawing upon critiques of AI hype and anthropomorphism, alongside Aristotelian conceptions of phronesis, praxis, and poiesis, the paper examines how narratives of technological inevitability and opacity contribute to the erosion of human practical judgment within higher education. Against this tendency, the paper proposes demythologisation as an ethical task for universities. Ethical AI education must move beyond technological adoption and literacy toward cultivating critical discernment concerning the limits, purposes, and appropriate role of AI within educational life and human flourishing.
Bridging the Chasm Between Macro Policy and Material Pedagogy: A Critical Evaluation of Digital Literacy and Proposed Curricular Innovations in State-Approved Modern Literature Textbooks Under the 2022 Revised National Curriculum
ABSTRACT. The global digital transformation (DX) in education—epitomized by South Korea’s 2022 Revised National Curriculum and High Touch High Tech (HTHT) frameworks—faces a critical chasm between macro policy and classroom reality. This study evaluates the modern literature chapters across all nine state-approved 10th-grade Common Korean Language 1 & 2textbooks.Integrating educational psychology, critical pedagogy, and digital humanities, our empirical analysis reveals five structural limitations: cognitive misalignment under Bloom’s Taxonomy and the SAMR model, deficient digital ethics, superficial collaboration within the Zone of Proximal Development (ZPD), instrumentalist media exploitation, and weak interdisciplinary efficacy.To bridge this gap, we propose four pedagogical innovations: fostering learner agency and self-efficacy, fortifying critical-ethical digital citizenship, implementing constructivist Project-Based Learning (PBL) via digital humanities tools, and driving socio-political literary practice through interdisciplinary data convergence. This research provides an actionable blueprint for global curriculum reconstruction and teacher development in the AI era.
Building a modified Kantian triangle for cyber trust
ABSTRACT. In this paper, we discussed the proposed normative model for “good” AI in education. It continues the literature review done for my thesis on “Building Cyber Trust: Market Value of Cybersecurity Breach Disclosures” and industry insights from tech risk leaders. Personalised AI learning collects a lot of data - keystrokes, engagement patterns, even emotional cues and this is necessary to power better support and feedback. Yet, where is the line separating support from surveillance? What happens when databases are breached and student data is exfiltrated? An updated Kantian triangle - pedagogical innovation (incentive dimension), ethical governance (content dimension), risk management (interaction dimension) - is mutually reinforcing. When balanced, this dynamically stable tripod creates a virtuous cycle where technology serves the primary goal of strengthening both the individual and society.
Neutral, Discriminatory and Inclusive AI for Education
ABSTRACT. Many scholars have argued against the idea that AI technologies are neutral and objective. Instead, a range of evidence has been presented suggesting that AI technologies are prone to value preferences and social-cultural biases (Jackson 2025, Tan et al 2026). Left unaddressed, these biases risk promoting increased social discrimination and disrupting social harmony both in the present and in the future. This paper explores the arguments against the idea that AI technologies are neutral and objective. It then explores the potential discriminatory implications that such systems may have for the education sector in Singapore. Finally, it concludes with several proposals for how the education system can embrace much of what is good in AI technologies while mitigating the potential biases and discriminatory implications that come with it.
Understanding Cognitive Distribution Patterns Between University Students and Generative AI
ABSTRACT. As generative AI becomes increasingly integrated into higher education, understanding how cognitive work is distributed between students and AI has emerged as a critical educational concern. This study explored how university students distribute cognitive tasks to generative AI and what factors influence variations in these patterns. Semi-structured interviews were conducted with 11 undergraduate and graduate students, and the data were analyzed using reflexive thematic analysis. The findings identified four patterns of human–AI cognitive task distribution: assistance, augmentation, co-construction, and delegation. Students flexibly transitioned across depending on educational context and norms, orientation toward learning and AI literacy. The findings extend distributed cognition theory and provide implications for designing educational practices that support productive human–AI collaboration while maintaining students’ cognitive agency.
Beyond Résumés and Career Maps: How AI Literacy Shapes Motivation, Engagement, and Higher-Order Thinking in Workplace Education
ABSTRACT. Generative artificial intelligence (AI) is increasingly used for feedback, revision, and reflective learning in higher education. This study examines an AI-enhanced workplace communication and training course implemented in a vocational-track university in Taiwan. The content of this course integrated career reflection, résumé writing, career mapping, design thinking projects, peer feedback, AI-supported feedback, and instructor feedback. Data from 100 senior business students were analyzed to examine a research framework illustrating the relationships among AI literacy, learning motivation, learning engagement, and higher-order thinking tendencies. Our results showed that AI literacy is positively associated with learning motivation, including both intrinsic and extrinsic goal orientations. In addition, learning engagement mediates the relationships between intrinsic and extrinsic goal orientation and higher-order thinking tendencies. Based on the findings in this study, the implications for AI-enhanced feedback design and workplace communication education are discussed.
An Interpretable AI Workflow for Metacognitive Reflection Feedback in Digital Learning
ABSTRACT. As digital learning becomes more interactive, learners are increasingly asked to explain what they understood, where they struggled, and how they approached a task. These open-ended responses contain valuable traces of metacognition. Recent advances in large language models (LLMs) and agentic artificial intelligence (AI) now make it possible to analyze such responses and generate support at scale. Yet current systems still struggle to turn reflective responses into valid, interpretable assessments of metacognition and into personalized scaffolds that learners can understand and use. We introduce a metacognitive AI engine that combines probabilistic multidimensional scoring, weak-dimension detection, interpretable diagnostic tagging, and scaffold selection with learner-centered personalization. Here, we report on the design of the system and preliminary validation from simulated stress-test cases, with authentic workshop data collection underway. Together, this work supports personalized reflection feedback within a single human-AI learning loop.
Know Your AI Persona: A Reflective Framework for University Students
ABSTRACT. Students are increasingly engaging with artificial intelligence (AI) tools in different ways for academic tasks and this creates a critical pedagogical challenge as institutions lack the means to identify and address these differences meaningfully. This paper proposes a behavioural framework to examine student AI usage and promote self-regulated, academically productive AI engagement. Using an empirical design, students will complete a diagnostic questionnaire reflecting on their usage such as frequency, purpose, and manner. Responses are analysed through clustering techniques to derive distinct behavioural personas. Each persona is paired with personalised, “Start, Stop, and Continue”, accompanied by practical guidance to support reflective behavioural change. The framework draws on self-regulated learning (SRL) theory to ground the reflective process, positioning students as active agents in developing better AI usage habits to aid in their learning. This study contributes a scalable, replicable instrument for institutions to move beyond uniform AI policies toward personalised, behaviour-informed interventions.
A Comparative Analysis of Institutional Generative AI Guidelines in Japanese and American Universities
ABSTRACT. This paper compares the official generative AI guidelines of 30 matched universities in Japan and the United States, alongside Japan's Ministry of Education (MEXT) frameworks. Both systems share one broad area of agreement: they permit AI use conditionally, leaving decisions to individual instructors. In implementation, however, the two systems differ considerably. Japanese guidelines emphasize moral principles but typically lack clear ownership, are rarely updated, and provide no institutional access to AI tools. American guidelines, by contrast, tend to be practical, enforceable, and linked to university technology infrastructure, though they are often fragmented and focus more on legal compliance than educational goals. Importantly, neither system does enough to train teachers and students about these policies. To address these gaps, we recommend that universities establish clear policy ownership, state explicit default rules, arrange shared AI tool access to address equity concerns, and introduce mandatory policy training—with particular urgency in Japan.
The Pedagogical Anatomy of GenAI Chatbots in Teacher Education
ABSTRACT. This study examines how teacher educators’ professional knowledge can be translated into the design of pedagogical generative AI (GenAI) chatbots. Adopting a design-based case approach, the study is rooted in a synthesis of theoretical knowledge applied across iterative cycles of problem analysis and design solutions, resulting in the development of a pedagogical GenAI chatbot. TPACK is used to conceptualise the key dimensions of teachers’ professional knowledge within the prompt. The findings show how pedagogical, content, and technological knowledge are integrated into the prompt, forming what the study terms the pedagogical anatomy of a GenAI chatbot. Across these dimensions, the chatbot was designed to guide pre-service teachers’ conceptual physics learning and scaffold collaborative learning processes. These dimensions also translated pedagogical intentions into task structures, scaffolding strategies, and rule-based interaction logic. The study contributes to emerging research by positioning prompt design as a central mechanism for shaping AI-enhanced learning environments.
Version 62: Student-Led AI Vibe Coding and Peer Digital Citizenship Education
ABSTRACT. This paper documents a 20-week MYP Digital Design unit in which Grade 9 students (n=11) used AI tools — including vibe coding and AI chatbot builders — to design, build, and deploy digital citizenship education for Grades 5–9 at an IB World School in Southeast Asia. Students identified learning gaps through peer surveys, built three deliverables per team (scenario-based AI chatbot, vibe-coded educational game, portfolio website), taught live audiences across the school, and measured outcomes. The unit produced a ~25% increase in self-identified confidence among taught cohorts and a 4.75/5 showcase rating. Students with no prior coding experience built functional, iteratively improved software through natural-language AI prompting — one student reached version 62 before final deployment. The case offers a model for positioning AI as an instrument of student-led creation and peer education.
Digital Twins and Embodied AI: From Simulation to Real-World Learning
Synopsis
This talk is about how digital twins and embodied AI connect virtual environments with real world experiences. It traces the evolution of digital twin technology, from digital replicas to interactive simulations that support experimentation, decisions and learning. Through practical examples from everyday life, education and training, the talk explores how these technologies help people practise skills, understand complex systems and explore situations safely. It also introduces how AI agents learn by interacting with simulated environments and discusses the opportunities and challenges of applying that learning in the physical world, making these ideas accessible to audiences without a specialist technical background.
Biography
Dr. Meg Rajendran is a Solutions Architect at the NVIDIA AI Technology Center. Her research focuses on synthetic data generation and generative AI, and she works on bridging virtual and real-world environments using digital twin technologies. She investigates methods to address data scarcity by turning it into an optimization opportunity, and she explores applications of digital twin approaches in both academic and industry research. Dr. Rajendran has published at conferences like SIGGRAPH Asia, and co-organized events at conferences such as AAAI Summer Symposium, SIGGRAPH Asia.