Exploring the Impact of AI ‐Based Learning Environments on Student Self‐Regulation and Adaptive STEM Learning

Background Although AI‐based STEM environments provide adaptive feedback and personalised learning pathways, empirical evidence explaining how these systems shape students' self‐regulation and adaptive problem‐solving processes remains limited. Prior research has largely focused on perceptions of AI use or tool validation rather than modelling the cognitive mechanisms underlying adaptive STEM learning. Objective This study examines how AI‐supported STEM environments influence secondary school students' self‐regulated learning processes and identifies which metacognitive components most strongly predict adaptive STEM performance. Methods A total of 649 secondary students engaged in structured AI‐enhanced STEM learning activities. Interaction‐derived indicators of metacognitive monitoring, strategic adjustment, cognitive flexibility and transfer were analysed. Network analysis was employed to model the structural interrelations among cognitive and metacognitive processes, and multiple regression models were conducted to determine their predictive contribution to adaptive STEM learning outcomes. Results Network modelling revealed a densely connected structure among metacognitive and adaptive learning processes, with cognitive transfer and adaptability (CTA) and AI‐enhanced self‐regulated learning (ASRL) emerging as central nodes. Regression analyses indicated that metacognitive and adaptive components significantly predicted adaptive STEM performance, explaining 68% of the variance in learning outcomes ( R 2 = 0.68, F [4, 55] = 31.74, p < 0.001). AI‐based metacognitive awareness ( β = 0.38, p < 0.001) and cognitive transfer and adaptability ( β = 0.29, p < 0.001) were the strongest predictors of adaptive learning performance. Conclusions AI‐supported STEM environments foster adaptive learning primarily through strengthening students' metacognitive awareness and cognitive transfer processes. These findings clarify the cognitive mechanisms through which AI feedback enhances learning and provide empirical guidance for designing intelligent systems that promote transferable and self‐regulated STEM competencies.