Effects of Integrating an AI Learning Companion into a Self-Regulated Learning 4L Framework on Elementary Mathematics Learning

Abstract

This study investigates whether integrating TALPer, an AI learning companion, into a self-regulated learning four lesson framework (SRL 4L framework) enhances fifth-grade students’ mathematics learning. Using a quasi-experimental design, instruction supported by TALPer was compared with SRL 4L framework–based instruction without AI support. Epistemic Network Analysis (ENA) and self-regulated learning coding systems analyzed cognitive structures and learning strategies. Results showed that students supported by TALPer achieved significantly higher learning outcomes. Process analyses further revealed that high-ability students demonstrated deep and multidirectional self-regulation with robust cognitive networks, whereas lowability students exhibited more surface-level and fragmented patterns. Overall, the findings suggest that the effectiveness of TALPer is closely related to learners’ regulatory awareness, highlighting the importance of strategy guidance and reflective scaffolding, particularly for low-ability students.