AI Chatbot Support During Prospective Teachers’ Proof Revision Activities

Source abstract

This study examines the use of a Large Language Model (LLM)-powered chatbot to support prospective teachers (PSTs) in learning about proof and proving. For this purpose, we developed MathCLAIM, a tool based on the GPT-4o model and informed by research-based teaching practices. We then designed a learning module integrating MathCLAIM to enhance PSTs’ understanding of proof and proving, and we implemented it in a mathematics content course for K-6 PSTs. In this study, we focused on five PSTs to explore how MathCLAIM facilitated improvements in the quality of their self-constructed arguments across two proof tasks. Arguments produced by PSTs with and without MathCLAIM, as well as their written reflections and interview responses, were analyzed to investigate how PSTs evaluated and revised their arguments with and without MathCLAIM’s support. The findings indicate that MathCLAIM provided various forms of support, guiding PSTs to refine specific components of their arguments by correcting reasoning errors, adding justifications, or reducing reliance on examples for generalizations. The paper discusses the implications of these findings for enhancing support for PSTs in learning proof and proving through LLM-powered tools.