Brown University Professor Proved Mass AI Cheating on Exams

Economist Roberto Serrano, a professor at Brown University, has published figures that, according to him, prove mass use of AI by students on exams: the average score on a take-home midterm was 96 out of 100, and after switching the final exam to an in-person format without access to neural networks, it dropped to 48 out of 100.

What the Professor Noticed

Serrano taught an economics course and, like many instructors, gave students a take-home midterm exam that could be completed outside the classroom. The average class score hovered around 96 points out of 100 — an almost perfect result across the entire class at once, which itself looked suspiciously uniform for a complex economics course.

Why the In-Person Exam Changed Everything

To test his hypothesis, Serrano switched the final exam to an in-person format — students took it in the classroom without computers and without access to chatbots. The result was dramatic: the average score plummeted to 48 out of 100, or approximately half of the take-home midterm score.

• Take-home midterm — average score 96 out of 100
• In-person final exam without AI — average score 48 out of 100
• Difference — nearly twofold
• Author of the observation — Roberto Serrano, economics professor at Brown University

Serrano made this story public, presenting the score gap as quantitative proof that a significant portion of students used AI tools when completing homework and exams that are not monitored in person.

"We cannot afford to become dumber," warns

Serrano, commenting on the consequences of students losing the ability to solve problems independently.

Why This Is Difficult to Prove Directly

Accusing an individual student of using AI on homework is almost impossible to prove formally — there is no camera, no proctoring, no technical way to capture the moment of cheating. This is why Serrano's approach seems more convincing than many other teachers' complaints about AI cheating: instead of accusing specific students, he built a controlled comparison of results from the same group of students under two different conditions — with access to tools and without. The difference in scores speaks for itself without needing to prove anyone's guilt personally.

What This Means

Professor Serrano's story is one of the most vivid public examples of how the gap between results on remote and in-person assessments reveals the scale of AI use in universities. For instructors, it is another argument in favor of returning to in-person exams and rethinking the evaluation format in the age of accessible chatbots: as long as assignments can be completed at home without supervision, the results cease to reliably reflect the knowledge of the student themselves.