ASSESSMENT OF CHATGPT’S MATH SOLVING ABILITY: THE CASE OF EXERCISES IN GRADE 12 MATH TEXTBOOKS IN VIETNAM

Abstract

Artificial intelligence (AI) is transforming how we do things. In school, teachers use it to improve lessons or help students. Many math teachers pay attention to the ChatGPT version o1-mini since it can solve math problems well. This study examines how accurately ChatGPT o1-mini can solve math problems for 12th-grade students in Vietnam. We looked at 675 math problems in three 12th-grade math textbooks in Vietnam (Canh dieu, Chan troi sang tao, and Ket noi tri thuc voi cuoc song). The results showed that ChatGPT o1-mini could solve 53.9% of the exercises, and it did the best in Statistics. However, 16% of the responses provided by ChatGPT o1-mini had mistakes, and 30.1% of the responses required further explanation as ChatGPT o1-mini used methods or terms that did not belong in the curriculum. The study also found out which contents have mistakes or wrong ideas, in the answers given by ChatGPT o1-mini. These results give math teachers a clear view of how well AI performs and how correct its answers are. This will help them figure out how to use AI, specifically ChatGPT o1-mini, in their teaching.