Recursive self-regulation in genAI-assisted experiential learning: A task-contingent model

Existing literature on generative AI (genAI) in business education has examined students’ attitudes, motivation, and the ethics of genAI use, alongside ongoing debate about whether genAI over-reliance may diminish self-regulated learning (SRL). However, how students self-regulate their learning while working with genAI during authentic tasks remains under-explored. This study addresses this gap by examining (RQ1) how students' self-regulatory strategies manifest during genAI-assisted learning and (RQ2) how these strategies change through genAI interaction . A thematic analysis was conducted of 34 metacognitive reflections generated across two scaffolded, four-stage experiential learning activities in a Lean Startup unit. Findings inform a three-stage AI-mediated recursive loop model, comprising monitoring, strategic adoption, and limitation identification, that extends cyclical models of self-regulated learning by showing how regulation may be initiated within the task by genAI-generated suggestions. Within this structured, reflective learning context, students indicated evaluative judgement and metacognitive regulation, providing preliminary evidence that scaffolded genAI activities can elicit active rather than passive engagement. GenAI-supported cognitive deepening was task-contingent, emerging most strongly during applied refinement tasks. The model offers management educators a pedagogical process model for designing scaffolded genAI-assisted activities that elicit monitoring, selective adoption, and critical reflection.