An economics professor at Brown University observed a dramatic drop in student performance when switching from a take-home exam to a proctored in-person final without AI assistance, according to The Decoder. The professor suspects most of the 86 students used AI tools on the take-home exam, which averaged a high score of 96 percent.
When the final exam was administered in person, 18 students dropped the course, nine failed to appear, and the average score plunged to 48.6 percent. These results highlight the potential overreliance on AI for academic work.
Supporting this, two large studies from China and UC Berkeley also found that students who depend on AI for homework tend to score significantly lower on supervised exams. For Japanese investors and educators, these findings underscore the growing impact of AI on learning outcomes and the importance of monitoring integrity in assessments amid expanding AI adoption.