Evaluation of Artificial Intelligence Performance in the Mathematical Problem-Solving Process

Akti Aslan, Seda, Miser, Meltem and Allison, Jordan ORCID: https://orcid.org/0000-0001-8513-4646 (2026) Evaluation of Artificial Intelligence Performance in the Mathematical Problem-Solving Process. Sage Open, 16 (3). pp. 1-21. doi:10.1177/21582440261486168

Preview

Text 16529 Aslan, S. A. et al. (2026) Evaluation of Artificial Intelligence Performance in the Mathematical Problem-Solving Process.pdf - Published Version Available under License Creative Commons Attribution Non-commercial 4.0. Download (1MB) | Preview

Abstract

Plain Language Summary This study looked at whether popular artificial intelligence (AI) tools, ChatGPT and Gemini, can help middle school students with algebra problems. We gave each tool ten math problems written in Turkish and then evaluated their answers step by step. We checked if the tools understood the problem, chose the right method, carried out the solution correctly, and checked their final answer. To make the evaluation fair, we used a scoring guide developed and reviewed by mathematics education experts. In addition, ten 8th-grade students were asked if the explanations were clear and helpful. The results showed that ChatGPT performed better overall, with a score of 83 out of 100, compared to Gemini’s 71.5. Both tools did well on straightforward problems but had trouble with harder questions that required deeper reasoning or multiple steps. Students reported that the AI explanations were often useful and helped them see different ways to solve a problem. However, some found the solutions too complicated or lacking enough explanation. This means that while AI tools can support learning, teachers still play an important role in guiding students and checking the accuracy of the responses. Overall, the findings suggest that AI chatbots can be a helpful resource in the classroom, especially for practicing math and exploring alternative solution methods. However, they cannot fully replace teacher support, particularly for more complex or challenging problems. Future research should look at other grade levels, more types of math, and different AI tools to better understand how these technologies can be used effectively in education.

Item Type:	Article
Article Type:	Article
Uncontrolled Keywords:	Artificial intelligence; ChatGPT; Gemini; Mathematics education; Problem-solving
Subjects:	Q Science > Q Science (General) > Q336 Artificial intelligence Q Science > QA Mathematics
Divisions:	Schools and Research Institutes > School of Business, Computing and Social Sciences
Depositing User:	Kamila Niekoraniec
Date Deposited:	21 Sep 2026 11:59
Last Modified:	21 Sep 2026 13:00
URI:	https://eprints.glos.ac.uk/id/eprint/16529

University Staff: Request a correction | Repository Editors: Update this record