Bodong Chen’s “Beyond Tools: Generative AI as Epistemic Infrastructure in Education” argues that AI’s educational significance lies not merely in output accuracy, bias or efficiency, but in its restructuring of epistemic agency: the human capacity to form, evaluate and validate knowledge. Rejecting the neutral “tool” metaphor, Chen frames generative AI as epistemic infrastructure, a mediating system that changes what teachers can notice, question, judge and habitually practise. The article develops three analytical conditions for evaluating such infrastructures: whether they afford skilled epistemic action, sustain epistemic sensitivity, and cultivate virtuous habits rather than dependence. Its case-study synthesis examines MagicSchool AI’s lesson-planning interface and Brisk’s automated feedback system, illustrated respectively on pages 9 and 10. Both promise time-saving support, yet Chen argues that they risk epistemic substitution: AI performs the very cognitive operations—pedagogical reasoning, interpretive judgement, contextual feedback—through which teaching expertise is developed. The teacher is thereby repositioned from primary epistemic agent to reviewer of machine-produced outputs. Chen’s conclusion is neither technophobic nor naively instrumentalist: AI should be designed to preserve expertise formation, make reasoning transparent, reinforce epistemic virtues and leave room for situated professional knowledge. Educational AI becomes defensible only when it sustains, rather than supplants, the judgement through which teaching remains an intellectual practice.
