Characterizing Teachers' Mathematical Conversations with Llms Through the Lens of Mathematically Responsive Teaching

Abstract

This exploratory study proposes a novel approach for integrating communication with large language models (LLMs) into teacher education, and a theoretical framework for analyzing the resulting interactions between prospective teachers (PTs) and artificial intelligence (AI) agents. The study involved 20 PTs who were tasked with identifying a mathematical problem from the curriculum where the LLM generates an erroneous solution, and subsequently guiding it to identify and correct its error. To investigate PT–AI interactions, particularly teachers’ mathematical responsiveness (or lack thereof), we employed methods from Systemic Functional Linguistics (SFL). We found substantial variation in responsiveness, shaped by whether PTs framed the interaction as a discussion with AI or adopted a teaching–learning stance.

Keywords

Preservice teacher educationLarge Language Models (LLMs)Mathematical responsivenessTeacher-AI interaction

Recommended citation

← All PME 49 Research Reports