When does AI support thinking, and when does it replace it?...
Chan builds a typology of nine ways secondary students describe using AI to learn, and organizes them around a single boundary: the same interaction either extends thinking or substitutes for it, depending on how the learner positions the system. Working from Vygotsky, distributed cognition, and the extended mind rather than from LIS, she arrives at a claim adjacent to Information Shock: AI is not a tool delivering outputs but a form of cognitive mediation operating inside the act of thinking, and learning has relocated from the object to the moment-to-moment exchange. Where she parts company is the remedy. Having placed cognition inside the exchange, Chan turns to reflective prompts, metacognitive checklists, and better learner positioning, the literacy frame Information Shock argues cannot reach the second mode. Notably, no one in her model is accountable for what the learner leaves believing; the burden falls entirely on the learner's own regulation. A preprint, self-report only from a single cohort fresh off an AI literacy course, so suggestive rather than conclusive, but a rare case of the learning sciences naming the mediation shift in their own vocabulary.
