Professor suspected AI-powered cheating on take-home midterms, makes finals in-person — only two students scored within 10% of their midterm score
TL;DR
At Brown University, economics professor Roberto Serrano moved the final exam back in person after unusually high take-home midterm results raised AI-cheating concerns. The class had 86 enrolled students instead of the usual roughly 30. The midterm average was 96 percent, while the course historically averaged 65 to 80 percent.
Nauti's Take
This is less a story about a few lazy students than a stress test for universities. When a midterm jumps to 96 percent and the in-person final drops to 48.6 percent, the assessment design has failed, regardless of how many individual cases can later be proven.
AI detectors will not fix that. Exams need formats where reasoning, context and personal work become visible again.
Briefingshow
The case shows how quickly take-home exams can lose assessment value when generative AI can produce plausible, technically correct answers at scale. At the same time, it is not clean proof of individual misconduct: the university still has to balance fairness, evidence and due process against an obvious assessment breakdown.
Sources