Automated Error Classification and Analysis of Fraction Computation Leveraging Large Language Model (Stage 2, 11:50 AM)

• In Event: e-Lightning Ed-Talk Session 22 - Stage 2

Abstract

This study supports the development of an AI-driven platform that diagnoses student misconceptions in fraction computation by analyzing final answers. We trained a machine learning model to generate and rank plausible error pathways, aiming to reduce teachers’ diagnostic burden and enable more personalized instruction. By modeling common misconceptions across operations and fraction types, the study explores how accurately the model can detect errors, its consistency across varying complexities, and optimal representations for student thinking. Our findings demonstrate the potential of integrating error analysis with AI to identify persistent gaps in understanding and offer targeted instructional strategies—paving the way for more efficient, responsive, and equitable support for students struggling with foundational math concepts.

Authors

• Isha Patil, California State University Los Angeles

• Lexi Hwang, California State University - Los Angeles

• Ken-ichi Nomura, University of Southern California

• Kyle Chasen Hay, University of California - Los Angeles