Exploring the Use of Conversational AI Tools in Mathematical Problem-Solving among Higher Education Students

Noor Latiffah Adam; Shaharuddin Cik Soh; Iskandar Zulqarnain Shaharuddin

Authors

Noor Latiffah Adam Multidiciplinary Information Retrieval (MuDIR), Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Shah Alam, Malaysia
Shaharuddin Cik Soh Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Shah Alam, Selangor, Malaysia
Iskandar Zulqarnain Shaharuddin Kulliyah of Sustainable Tourism and Contemporary Languages, International Islamic University, Muar, Malaysia

Keywords:

Conversational AI, Large Language Model, educational technology, mathematics education, problem-solving

Abstract

This preliminary study examines how higher education students utilize large language models (LLMs), such as ChatGPT, Claude, and Gemini, for solving math problems. As these conversational AI interfaces become more common, understanding their role in mathematical learning is essential. This research examines key aspects of LLM interaction in mathematics, including problem types, prompt engineering, solution presentation and interpretation, solution verification, learning impact, and ethical considerations. Using a structured survey of 34 students across disciplines, we assessed their awareness and usage of LLMs for mathematical tasks. Preliminary findings indicate high awareness and diverse approaches to prompting and solution verification. Students primarily use AI for step-by-step explanations (78.8%) and solution verification (78.8%), with 93.9% reporting increased understanding. However, 84.8% experienced incorrect solutions, highlighting verification challenges. These findings offer valuable insights into student behavior and provide a foundation for developing pedagogical strategies that leverage conversational AI to enhance mathematical learning while addressing its limitations.