人工智能能否超越优等生解决高中化学问题？：基于能力测评的证据 Can Artificial Intelligence Surpass Top Students in Solving High School Chemistry Problems? Evidence from Competency Assessment

Abstract

生成式人工智能的快速发展为教育带来了新的机遇与挑战。为了评估生成式人工智能在高中化学教育中的应用潜力，研究选取部分化学试题对ChatGPT、讯飞星火两大生成式人工智能的解题能力进行了测试，并与重点中学选考化学的高三学生的解题能力进行了比较。结果显示两大模型准确率仅高于小部分学生。进一步分析错误类型后发现，生成式人工智能在解题过程中存在概念理解错误、信息提取不足等问题。基于此，研究为生成式人工智能辅助化学教育及其自身未来的优化与发展提供了启示与建议。 The rapid development of artificial intelligence generated content (AIGC) presents new opportunities and challenges for education. this study tested the problem - solving abilities of ChatGPT and iFlytek Spark on chemistry questions, comparing them with those of senior high - school students taking chemistry in key schools. Results show that the accuracy rates of the two models are only higher than those of a small number of students, with students' performance significantly better. Further analysis of error types reveals problems like conceptual misunderstandings and insufficient information extraction in the models' problem - solving. Based on this, the study provides insights and suggestions for the use of generative AI in chemistry education and its future optimization.