How could roles shape teachers’ cognitive and interactional processes in collaborative learning? The design of GenAI-assisted collaborative role-based strategy

Online collaborative learning plays an important role in in-service teachers’ professional development, yet unclear role definition often leads to insufficient cognitive engagement and knowledge construction. Although role-based strategies can address these issues, fixed role scripts may over-structure interaction. To overcome this limitation, this study designed a GenAI-Assisted Collaborative Role-Based Strategy (GenAI-CRS) consisting of four roles: moderator, initiator, facilitator, and summarizer. A quasi-experimental design was employed to examine its effects. Results show that the experimental group significantly outperformed the control group on academic performance. The epistemic network analysis and behavioral sequence analysis further reveal that, the experimental group were more likely to engage in high-level cognitive processing and task-oriented metacognitive activities. In contrast, the control group’s interaction centered primarily around cognitive conflicts and emotional expressions, indicating a context-driven self-organized collaborative pattern. Additionally, the moderator and the initiator are active in metacognitive and emotional dimensions; the facilitator primarily promotes group members to engage in high-level cognitive processing, while the summarizer more frequently stays at low-level cognition and irrelevant communication. In summary, the study reveals the mechanism of GenAI-CRS, which may provide insights for AI-enhanced teacher professional development and collaborative learning design.