Reconfiguring thinking through ChatGPT ‐mediated learning: Metacognition, engagement and cognitive agency in a youth apprenticeship programme

Abstract Generative artificial intelligence (AI) is increasingly promoted as a means of supporting learning, metacognition and self‐regulated learning. However, less is known about how thinking is reconfigured through generative AI‐mediated learning and how these processes unfold within particular pedagogical and social conditions. This qualitative study examines how students' thinking practices, learning processes and cognitive agency are reconfigured through ChatGPT‐mediated learning within a youth apprenticeship programme in the United States. Guided by distributed cognition and learning ecology, the study employed an interpretive multi‐perspectival design drawing on semi‐structured interviews, approximately 300 h of classroom observations, ethnographic conversations and student‐generated artefacts. Reflexive thematic analysis indicated that thinking emerged through iterative interactions among learners, AI, peers and instructional contexts. Students' engagement varied according to the pedagogical and social conditions surrounding AI use, while cognitive agency was enacted through continual judgements about credibility, authorship and responsibility for AI‐generated knowledge. These findings suggest that generative AI participates in reorganizing the conditions under which cognitive work is accomplished. The study extends distributed cognition and learning ecology by conceptualizing AI‐mediated thinking as relational, dialogic and contextually situated. Practically, the findings suggest that AI‐mediated learning should be designed around opportunities for explanation, verification, dialogue and learner accountability.