ChatGPT Chatbot in the Teaching and Learning of Mathematics: The Role of Reasoning Processes

Abstract

The release of ChatGPT in November 2022 represented a significant advancement in creativity and innovation across various fields, particularly in the teaching and learning of mathematics. This AI-driven chatbot conversationally engages users, utilizing human-like language to respond to inquiries and generate content. It follows prompts and previous interactions, adjusting and refining its answers based on the sequence of inputs and outputs. Our study examines ChatGPT's modeling of mathematical reasoning by analyzing its solutions and proofs across a spectrum of problems with varying levels of complexity. The dataset consists of 12 activities categorized into four types of mathematical reasoning. This research investigates whether this emerging technology, now accessible to students, can improve the relationship between chatbots and students, particularly in terms of reasoning processes. We employed a qualitative descriptive approach to evaluate the reasoning performance of ChatGPT, focusing on its latest versions, GPT-4o and GPT-4o Mini. This evaluation is grounded in a classification of reasoning performance characteristics, emphasizing the structural steps of each reasoning type (Jeannotte, 2015; Nhiry et al., 2023). Our analysis revealed that ChatGPT demonstrates a satisfactory ability to maintain the structure of mathematical reasoning; however, several shortcomings were identified in terms of clarity, rigor, and generalization. Enhancing its formalization, pedagogical explanations, and result validation would strengthen and enrich its approach.