Entropy-Based Analysis of Teacher Adaptability in Generative AI Agents in 3D Teacher Simulations

• In Event: AERA Poster Session Wednesday 3:45pm
  In Poster Session: Modeling, Measurement, and Design in Technology-Enhanced Learning Environments

Abstract

Educational simulations are pivotal in teacher preparation, enabling novice educators to hone pedagogical strategies in risk-free settings. Conventional platforms employ scripted, rule-based agents with predictable responses, which may limit teachers' adaptive skills. The integration of large language models (LLMs) revolutionizes this by deploying generative AI agents that produce diverse, context-sensitive interactions, potentially enhancing teacher flexibility. Analyzing 184 sessions encompassing 1,282 interactions via entropy metrics—(1) Interaction Entropy, (2) Adaptation Rate, and (3) Reactive Diversity—we found GenAI agents yield higher teacher adaptation rates, though not statistically significant, suggesting a trend toward greater strategic shifts in dynamic contexts.

Authors

• Jewoong Moon, University of Alabama

• Jieun Lim, Daegu National University of Education

• Yunseo Lee, University of Wisconsin - Madison

• Gyuri Byun, Seoul National University

• Yeil Jeong, Seoul National University

• Unggi Lee, Korea University