Grades dropped from 96 to 48 percent when a Brown professor made students take the exam without AI

What happened

An economics professor at Brown University suspected widespread AI assistance on a take-home exam after students averaged 96 percent. To test his theory, he made the final exam an in-person, no-AI proctored test. The results plunged: the average score dropped to 48.6 percent. During the switch, 18 of the 86 students dropped the course, and nine did not show up for the final. Two large independent studies from China and UC Berkeley confirmed this pattern: students who rely heavily on AI for homework see their scores fall sharply under supervised exam conditions.

Why it matters

This case exposes how heavily some students lean on AI to complete take-home assessments, inflating their grades artificially. For educational institutions, the incident pressures testing formats to ensure academic integrity and accurate skill measurement. It also raises questions about the fairness and validity of remotely administered or AI-augmented exams. More broadly, this episode signals that as AI tools become ubiquitous, traditional assessment methods may no longer reflect a student’s true capabilities, forcing schools to reconsider testing policies and proctoring.

What to watch next

Expect more educational institutions to intensify AI policies and monitor for academic dishonesty on take-home assignments. Proctored exams or other safeguards may become standard, potentially increasing administrative burdens and student dropouts in courses where AI use is suspected. Investors and product builders should watch for education technology solutions that balance AI integration with trustworthy evaluation methods. Regulators might also get involved in defining acceptable AI use in learning environments.

AI Quick Briefs Editorial Desk