At the Frontiers symposium “Artificial Intelligence (AI) at the crossroads of impact, sustainability, and responsibility,” hosted by the Royal Academy of Engineering (RAEng) in Tokyo from 4 to 6 February 2026, Tetsunari Inamura, the principal investigator of this project, served as a session chair of Session 3 “Education in the age of AI” and gave an invited talk on 5 February entitled “Designing AI for Learning When Values Conflict: From Cost-Performance to Self-Efficacy and Agency.”
Frontiers is an invitation-only, cross-disciplinary symposium series that the RAEng runs around the world. This edition was held in Japan in partnership with the Japan Science and Technology Agency (JST). The programme was organised around three sub-themes — “Responsible AI,” “Efficient and Sustainable AI,” and “Education in the age of AI” — and brought together researchers and practitioners from a wide range of disciplines and sectors worldwide, not only from engineering. The symposium was chaired by Professor Maja Pantić (Imperial College London / NatWest Group) and Professor Tomohiro Shibata (Kyushu Institute of Technology).
Session 3, “Education in the age of AI,” was co-chaired by Ms Ojoma Ochai (Co-creation Hub (CcHUB) Africa, Nigeria), Professor Tim Coughlan (Institute of Educational Technology, The Open University, UK) and Inamura.
Abstract of the Keynote
AI is rapidly entering education, yet “quality” is often reduced to short-term performance metrics such as accuracy, test scores, or time saved. In practice, however, the quality of an AI-enabled learning system is a value judgement made by multiple stakeholders—learners, teachers, system developers, institutions, and society—and these values can legitimately conflict. What a student experiences as “good” support (e.g., autonomy, confidence, motivation) may not align with what a teacher prioritizes (e.g., fairness, accountability, workload), or what an institution optimizes (e.g., scalability, cost-effectiveness, measurable outcomes). These tensions become sharper when we also require AI to be responsible (ethics, inclusivity) and sustainable (energy and resource constraints) across the AI value chain.
The talk proposed a practical “perspective-aware evaluation” template: (1) map stakeholders and the values they use to judge educational quality, (2) separate short-term efficiency from long-term developmental outcomes such as self-efficacy and agency, and (3) make trade-offs explicit as design decisions rather than hidden assumptions. Inamura connected this lens to JST CREST Self-Mirroring Twins, framing value alignment as a core design goal for AI that adapts to learners while preserving autonomy and supporting durable motivation.
Session Format
After the three chairs’ presentations, participants were divided across ten tables, each taking on a different perspective: a student in higher education, a student in secondary school, a secondary school teacher, a university lecturer, a lifelong learner in a job with minimal digital tools, a lifelong learner working with digital tools, an educational technology company, an employer in industry, a primary school head, and a university administrator. Each group discussed what “good AI” looks like for its role and what education should look like for it. Inamura chaired the closing plenary discussion and the wrap-up, drawing together how these differing value judgements can be reconciled.
After the symposium, seed funding of £20,000 per project is available for new international collaborations formed among participants. We expect the connections across disciplines and regions created through this symposium to contribute to the interdisciplinary research community that SeMT aims to build.