ChatGPT’s Capability in Solving Mathematical Model ling Problems
Nemanja Vučićević1[0000-0002-4903-7280], Aleksandar Milenković1* [0000-0001-6699-8772], and Marina Svičević1[0000-0003-2791-3849]
1 University of Kragujevac, Faculty of Science, Radoja Domanovića 12, 34000 Kragujevac, Serbia
nemanja.vucicevic@pmf.kg.ac.rs,
aleksandar.milenkovic@pmf.kg.ac.rs,
marina.svicevic@pmf.kg.ac.rs
DOI: 10.46793/eLearning2025.184V
Abstract. In this paper, we examine the capability of ChatGPT in solving problems based on mathematical modelling. In line with the research objective, we assigned the free version of ChatGPT (GPT-4o) six problems that fourth-year mathematics students, majoring in Applied Mathematics, had solved as part of their exam. We analyzed whether the following steps were properly carried out: constructing a mathematical model, the problem-solving process in the mathematical context, interpreting the solution in relation to the problem, and finally, validation. The results were then compared with the work of the aforementioned students. The findings indicate that, in all parts of the solution and for all tasks, the free version of ChatGPT was either completely or partially successful. Nevertheless, it achieved results that were weaker than those of the students. Based on these findings, we can say that solutions to problems addressed through mathematical modelling – which, in the second step, involve solving first-order differential equations, game theory problems, and linear programming tasks – can be used to some extent as student support. However, it is equally necessary to work on the specialization of LLMs that would solve problems of this level of complexity in accordance with the mathematical modelling process.
Keywords: Mathematical Modelling, ChatGPT, Student Performance Comparison.
Acknowledgments. This work was supported by the Ministry of Science, Technological Development and Innovation of the Republic of Serbia through the Agreement No. 451-03-65/2024-03/200122.
References
[1] Blum, W., & Leiss, D. (2007). How do students and teachers deal with modelling problems. Mathematical modelling (ICTMA 12): Education, engineering and economics, 222-231.
[2] Schukajlow, S., Kaiser, G., & Stillman, G. (2018). Empirical research on teaching and learning of mathematical modelling: A survey on the current state-of-the-art. ZDM, 50(1), 5-18.
[3] Sen Zeytun, A., Cetinkaya, B., & Erbas, A. K. (2023). Why do prospective teachers have difficulties in mathematical modelling? Insights from their perspectives. International Journal of Mathematical Education in Science and Technology, 55(1), 94–117. https://doi.org/10.1080/0020739X.2023.2171922
[4] Gilson, A., Safranek, C. W., Huang, T., Socrates, V., Chi, L., Taylor, R. A., et al. (2023). How does ChatGPT perform on the United States medical licensing examination? The implications of large language models for medical education and knowledge assessment. JMIR Med. Educ. 9, doi: 10.2196/45312
[5] Svičević, M., Milenković, A., Vučićević, N. & Stanković, M. (2025) Evaluating the Success of AI Tools in Supporting Student Performance in Mathematical Kangaroo Competition, Computer Applications in Engineering Education 33(4), https://doi.org/10.1002/cae.70063
[6] Milenković, A., Vučićević, N., & Svičević, M. (2025) Evaluating open ai tools o1 and o3 mini in solving high school problems from Serbian national mathematics competition, The teaching of mathematics, XXVIII (1), 14 – 29, DOI:10.57016/TM-EYRQ1676
[7] Helfrich-Schkarbanenko, A. (2023). Mathematik und ChatGPT: ein Rendezvous am Fuße der technologischen Singularität. Springer Spektrum.
[8] Koceska, N., Koceski, S., Lazarova, L. K., Miteva, M., & Zlatanovska, B. (2023). Can ChatGPT be used for solving ordinary differential equations. Balkan Journal of Applied Mathematics and Informatics, 6(2), 103-114.
[9] Zhu, T., Li, X., & Zhang, W. (2023). Applying chatgpt-powered game theory in ransomware negotiations. Authorea Preprints.
[10] Huang, X., Shen, Q., Hu, Y., Gao, A., & Wang, B. (2024). Llms for mathematical modeling: Towards bridging the gap between natural and mathematical languages. arXiv preprint arXiv:2405.13144.
[11] Li, Q., Zhang, L., & Mak-Hau, V. (2023). Synthesizing mixed-integer linear programming models from natural language descriptions. arXiv preprint arXiv:2311.15271.
Izvor: Proceedings of the 16th International Conference on e-Learning (ELEARNING2025)