Student-AI interaction patterns in collaborative problem solving using ordered network analysis

This study examined the dynamics of student-AI interaction (SAI) patterns in collaborative problem-solving, focusing on a curriculum design task facilitated by the design thinking (DT) process. We retrieved 597 graduate students and AI agents' conversation threads and annotated them with the coding schemes of the collaborative problem-solving process. Then, we conducted an order network analysis to identify meaningful SAI patterns in each phase of the DT process and visualize the sequential networks of each phase. Findings revealed that three behavior paths: (1) students frequently request the AI to respond to viewpoints or suggestions and the AI provides task-relevant information or resources to support; (2) students ask AI to respond to a viewpoint or suggestions to clarify a statement, (3) Students request AI to respond to viewpoints, and AI elaborates on its suggestion in detail, followed by various types of mutual cognitive interaction between students and AI in each phase of DT. The findings suggest that instead of AI being a passive tool, it becomes a dynamic learning partner, creating a collaborative learning peer and hybrid intelligence environment. This study offers theoretical, methodological, and practical implications for fostering educationally meaningful SAI.