Which Readabilities Influence Grade Level of GAI-Generated Math Stories? Insights for Effective Materials (Poster 8)

• In Event: AERA Poster Session Thursday 1:45 p.m.
  In Poster Session: Advanced Technologies for Learning Poster Session 1

Abstract

Mathematical stories enhance students' interest in mathematics and improve academic performance. This study uses generative artificial intelligence (GAI) to create mathematical stories for grades 3, 4, and 5. We developed a framework to automatically assess the quality and suitability of these stories for different grade levels. By analyzing text features and using machine learning models, we identified factors influencing story grades. Our findings show that higher-grade stories are more concise but use more complex grammar and syntax, suggesting a need for improvement in GAI-generated content. The semantic correlation between student parameters and generated stories decreases with higher grades, indicating further refinement of GAI algorithms is needed. This approach provides insights into leveraging GAI for designing effective mathematical education material.

Authors

• Wanli Xing, University of Florida

• Hai Li, University of Florida