Simulating Children’s Emergent Roles in Collaborative Learning through Multi-Agent Systems

• In Event: AERA Research in Progress Roundtable Session Three
  In Invited Roundtable: Research-in-Progress RT076 (Friday)

Abstract

This study explores the feasibility of using multi-agent systems (MAS) to simulate children’s emergent roles in collaborative learning. We coded classroom dialogues from a fourth-grade creative design course with a predefined role framework to extract behavioral features. These features, combined with academic and peer data, informed the design of an AutoGen-based MAS. Agents were assigned prompts that directed them to enact roles and interact within task settings, enabling the MAS to model classroom collaboration and generate dialogue. Simulated interactions are compared with authentic ones using statistical analyses to evaluate consistency in discourse patterns, role distribution, and content. The study integrates empirical coding with computational simulation and provides teachers and pre-service educators with a tool to understand group dynamics.

Author

• Xi Chen, Shanghai Jiao Tong University