生成式人工智能对大学生学业表现的影响

——基于广义随机森林模型的异质性处理效应

打开文本图片集

中图分类号：G640 文献标志码：A文章编号：1673-3851（2025）12-0766-13

Abstract：The application of generative artificial inteligence （generative AI） in education is becoming increasingly widespread， but its effectiveness in higher education remains highly contested within the academic community. This study conducts a questionnaire survey using a four-stage mixed sampling method across several universities in Hangzhou and employs a generalized random forest model to analyze the average and heterogeneous impact of generative AI on collge students'academic performance. The results show that generative AI has a high adoption rate among college students and significantly improves their academic performance and learning eficiency. Further analysis reveals that the duration of AI usage positively affects academic performance，while usage frequency has no significant effect. The analysis of individual treatment efects indicates substantial heterogeneity：students with higher learning adaptability are more likely to experience improvements in academic performance； students with lower initial grades benefit more in terms of performance gains； those with better initial performance are more likely to see improvements in learning efficiency；and those with stronger anti-interference abilities experience greater gains in learning eficiency. These findings not only provide empirical evidence for scientifically understanding the value of generative AI in higher education，but also offer practical reference for enhancing the learning effectiveness of college students and promoting the implementation of the "AI + higher education" strategy.

Key words： generative artificial intelligence；college students； academic performance； generalizec random forest model；heterogeneous treatment effect

在全球数字化转型加速的背景下，2024年中国政府工作报告提出“人工智能 + "行动，推进以人工智能为引擎的新质生产力发展。（剩余17569字）