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
Brown University professor Roberto Serrano moved the final exam in person after an unusually strong take-home midterm. The class averaged 96 percent, while past midterm averages were reportedly between 65 and 80 percent. Of 86 enrolled students, 18 dropped the course after the in-person final was announced and nine skipped the final. That left 59 students taking the exam.
Nauti's Take
This is not a win for AI detectors; it is a warning about assessment design. Serrano’s numbers look brutal, but the practical lesson is simple: old take-home exams in an AI world often test access, risk appetite, and prompting more than understanding.
Universities need less panic and more resilient formats.
Briefingshow
The case shows how quickly traditional assessment breaks when powerful AI tools are available without clear rules. This is not just about cheating, but about measurement: if take-home exams cannot distinguish skill, prompting, and outsourcing, universities need to redesign tasks, grading, and supervision.
Sources