AI Boosts Homework Scores But May Hinder Test Performance, Study Finds
Research indicates a significant disconnect between AI-assisted coursework and actual knowledge retention for exams.
Students who leverage artificial intelligence for their homework see improvements in assignment scores and complete tasks more quickly, but their performance on exams may suffer compared to peers who do not use AI, according to a large-scale study.
The research, conducted over six months, tracked approximately 27,000 students aged 12 to 18 in China. It found that students utilizing AI tools for homework assignments experienced an 18% increase in their scores across various subjects. Additionally, the time spent on assignments decreased from an average of 64 minutes to 45 minutes. However, the study also revealed that these same students scored 20% lower on exams than those who completed their work without AI assistance.
Experts suggest that the widespread adoption of AI in education may be altering the traditional relationship between homework grades and the retention of knowledge for summative assessments. In China, the study noted that 80% of students reported using AI models like Doubao and DeepSeek. This trend mirrors adoption rates in Western countries, where surveys indicate that 80-90% of U.S. undergraduates, 94% of British students, and 93% of German students use AI in their studies.
However, not all research presents AI as detrimental to learning. In a separate study conducted at Middlebury College, students were introduced to an unfamiliar topic with and without the aid of an AI chatbot. Those who used the AI chatbot achieved higher scores on tests and retained this advantage when the task was repeated a week later. This suggests that AI can enhance learning productivity, but its effectiveness is contingent on intelligent usage.
The Economist reported that AI can boost learning productivity "only for those who use the technology intelligently." The implication is that students must be guided to use AI as a tool for understanding and exploration rather than as a simple answer generator. The temptation for students to rely on AI-generated responses without engaging in their own critical thinking processes is a key concern highlighted by researchers.