Large Language Models in Tertiary Mathematics Education: A Systematic Literature Review

Abstract

Large language models (LLMs) have quickly become a focal point, sparking both excitement and questions within higher education, particularly concerning mathematics instruction. Our systematic literature review (SLR) explored peer-reviewed research published from 2020 through 2025 to understand how LLMs, including tools like GPT, are being used in tertiary mathematics education. The findings reveal a range of applications: serving as digital tutors, providing learner support, automating assessments, assisting with content creation, and aiding curriculum planning. These models show significant potential to enhance teaching and learning. Looking at how they function, LLMs can deliver detailed step-by-step explanations, create practice problems and materials, and offer personalized support to students. They are also valuable for instructors, assisting with tasks like feedback and grading. Studies point towards effective LLM use potentially leading to better student engagement, motivation, and problem-solving skills. Furthermore, educators are starting to adopt these tools, finding them helpful for streamlining. However, challenges persist. LLMs may produce errors, foster student over-reliance, or raise academic integrity issues. Ethical concerns, such as bias and responsible use, underscore the need for clear institutional policies and thoughtful integration. This review identifies key trends and gaps, including the lack of longitudinal classroom research and professional development. With proper oversight, LLMs offer significant potential to support personalized, innovative mathematics education without replacing the critical role of human educators.

Keywords: large language models, mathematics education, tertiary mathematics education, systematic literature review, LLMs, ChatGPT, higher education

DOI: 10.7176/JEP/16-5-10

Publication date: May 30^th 2025