Title: Reimagining Collaboration for Learning in an Era of Hybrid Human–AI Intelligence: Agency, Judgment, and Responsibility in AI-Rich Education

Abstract:Artificial intelligence is rapidly becoming woven into the everyday infrastructure of education, from teaching and assessment to feedback, planning, inquiry, and collaborative knowledge work. Yet much of the current discourse still treats AI mainly in terms of capability, efficiency, and adoption. In this keynote, I argue that a learning sciences perspective is needed to shift attention from what AI can do to how it reshapes the conditions of learning: how tasks are framed, how thinking is scaffolded, how judgment is exercised, and how agency and responsibility are distributed. Drawing on recent research on student–AI dialogue, teacher–AI collaboration, and AI-supported feedback, I examine how AI mediation can both support and weaken learning. AI may help learners generate ideas, receive feedback, and extend inquiry; but it may also encourage cognitive offloading, premature closure, over-trust, and passive acceptance of plausible but unexamined outputs. Central to this argument is meta-task awareness: the capacity of teachers and learners to monitor the evolving task, judge the quality of AI contributions, decide when to rely on AI, when to question it, and when to reassert human responsibility. I conclude by proposing a learning sciences agenda for hybrid human–AI intelligence that prioritizes visible thinking, productive friction, contingent scaffolding, feedback literacy, developmental readiness, responsible teacher judgment, and accountable human–AI partnership. The central issue is not simply whether AI should be integrated into education, but how education can design human–AI collaboration so that AI expands, rather than diminishes, learners’ capacity to inquire, judge, create, and participate responsibly in knowledge-building communities.