Selective self-regulated learning changes in an AI-supported inquiry environment: evidence from an authentic secondary physics module

Generative artificial intelligence (GenAI) is increasingly utilized to enhance students’ self-regulated learning (SRL), yet it remains unclear what roles such tools play during authentic subject-specific inquiry-based learning. This mixed-methods field study examined SRL in an AI-supported inquiry environment that combined a simulation, a GenAI chatbot, and task-embedded SRL prompts in a Grade 9 physics unit. Quantitative data were used to examine pre-post SRL change and predictors of post-intervention SRL, while interview data explained students’ uptake of the supports. Results showed small and selective SRL change. Motivational SRL improved modestly, especially self-efficacy, whereas strategic SRL did not show significant improvement. Post-intervention SRL was explained mainly by baseline SRL and students’ task engagement, not by raw counts of chatbot questions or simulation actions. Interviews suggested that the environment supported regulation most clearly during task execution: students used the chatbot to seek help, verify ideas, and continue inquiry, whereas worksheet-based planning prompts were weakly enacted as genuine forethought. The study contributes classroom-based evidence that AI-supported inquiry may provide situated regulatory support without necessarily producing broad SRL development. It highlights the need for phase-sensitive scaffolds, dialogic chatbot design, and process measures that capture the quality, timing, and uptake of interaction.