ChatGPT and other AIs: Personal relief and limitations among mathematics-oriented learners

Abstract

Mathematics utilizes symbolic language, logics, relationships, and numerical connections that made it challenging for students to learn and develop their computational skills. Recently, artificial intelligence emerged as a supplement tool for education and learning because of its ability to detect relationships and logics. Academic institutions are looking on how to implement AI-assisted learning specifically for mathematics to aid in increasing the quantitative competence of students. This study analyzed the relief and limitations of AI chat models in learning among mathematics-oriented students in higher education. Fifteen students participated in this study from mathematics-oriented courses e.g., engineering, statistics, and education. Interviews were conducted on how students used AI chat models to assist in learning mathematical concepts and methods. Narratives indicated that AI chat models like ChatGPT and Bard were capable of accurately responding to chat prompts in problem solving, proving, and explanations. It was prominent that these AI models understood the mathematical language and use of symbols for integration, derivatives, limit, fractions, exponentials, and intervals. At some instances, AI models could give inaccurate results or incorrect methods for solving; they also sometimes give correct answers on second run of chat prompts after these mistakes. These results had promising implications in education as these accessible AI models could reinforce the firsthand learning of mathematical foundations. This preliminary study offered important usability of AI models in mathematics in assisting students and monitoring their learning progress.