ASSESSMENT OF CHATGPT’S MATH SOLVING ABILITY: THE CASE OF EXERCISES IN TWELFTH GRADE MATH TEXTBOOKS IN VIETNAM

Tóm tắt

In the rapidly evolving landscape of Artificial Intelligence (AI), particularly its advanced data processing capabilities, AI applications in learning support are opening innovative approaches in mathematics education. The 2018 General Education Mathematics Curriculum emphasizes integrating digital technologies, including AI’s data processing potential, to develop essential student mathematical competencies. The Ministry of Education and Training specifically identified the ability to use mathematical learning tools’ as a critical educational component. Focusing on this practical context, the study evaluated the mathematical problem-solving capabilities of ChatGPT o1-mini within Vietnam’s 12th-grade curriculum. Researchers analyzed 675 problems across Algebra, Geometry, and Statistics, carefully refined to match ChatGPT's processing capabilities by excluding visually-dependent tasks. Using targeted prompts to guide the AI to solve problems like a mathematics teacher, the research required solutions in LaTeX format. The results revealed that ChatGPT o1-mini correctly solved 53.9% of problem sets, with highest effectiveness in Statistics. However, significant limitations emerged: 16% of solutions contained errors, and 30.1% were problematic due to methods or terminology beyond the curriculum’s scope. This research transcends a mere assessment of ChatGPT’s current capabilities. It provides critical insights into artificial intelligence’s potential in mathematics education, offering educators and students a foundational understanding for appropriately and effectively leveraging AI tools in learning environments.