Generative AI as 'Epistemic Arbiter' boosts student collaboration, study finds

Researchers Jiaxin Zou, Xiaoming Zhai, and Chunlei Gao conducted a six-week quasi-experimental study with 201 fifth-grade students to examine how generative AI (GAI) influences collaborative problem-solving (CPS). The team split students into two conditions—one with GAI support, one without—and used chi-square and lag sequential analysis to map interaction patterns. Results showed that GAI-supported groups exhibited significantly more social behaviors, particularly engagement and conflict management, but less frequent cognitive behaviors like task planning and solution reasoning. This suggests AI changes the dynamics of group work, prioritizing social cohesion over traditional planning steps.

The study's lag sequential analysis revealed a distinct interaction pathway in the GAI group: task planning led to conflict management, which then transitioned to solution reasoning—whereas the control group followed a conventional listening-to-planning trajectory. Thematic analysis of AI interaction logs showed students used GAI as an 'epistemic arbiter,' leveraging AI-generated facts and visualizations to resolve disagreements constructively. This finding highlights that AI doesn't just retrieve information; it mediates socio-cognitive processes, helping students move past social friction into collaborative reasoning. For educators and AI system designers, this signals a new role for AI in classrooms—not as a content deliverer, but as an active facilitator of group dynamics and critical thinking, with potential trade-offs in cognitive effort that merit further investigation.