A Brown University economics professor reported a dramatic drop in exam scores after moving from a take-home format to an in-person, proctored test, suggesting widespread AI cheating. The class average fell from 96 percent to 48.6 percent, the lowest ever recorded for the course. Nearly all students who scored near-perfect on the take-home midterm failed the proctored final, according to Inside Higher Ed. The professor, Roberto Serrano, said he noticed the unusually high scores on take-home exams and suspected AI use. He then administered an in-person test, which revealed a sharp decline in performance, with 19 students failing outright and 18 dropping the course.
Two additional studies support Serrano's findings, showing a consistent pattern of inflated homework grades and poor exam performance when AI is used. One study tracked over 26,000 students in China and found that homework scores rose by 18 percent after AI adoption, but exam scores fell by 20 percent. The long-term impact was significant, with top students losing up to 24 percent of their performance. Another study from a Texas university found a 13 percentage point increase in A grades after ChatGPT was introduced, primarily in courses with heavy homework loads.
Serrano criticized the university's response, calling it 'meek,' and urged stronger action against AI cheating. He argued that widespread use of AI in education risks normalizing cheating and harming academic integrity. 'We cannot afford to have a society in which a significant fraction of our best young minds think that cheating is OK,' he said. The issue continues to spark debate, with further discussions ongoing on how to address AI's role in education.
Source: thedecoder
