Brown University Professor Moves Final Exam In Person After AI Cheating Concerns
The rapid rise of artificial intelligence is forcing universities to rethink how they evaluate students, especially as concerns grow that AI tools may be weakening learning, critical thinking, and academic honesty. One recent case at Brown University has drawn attention after an economics professor changed his final exam format following unusually high midterm results.
Robert Serrano, an economics professor at Brown University, became suspicious after students in his ECON 1170 course achieved exceptionally strong scores on a take-home midterm exam in March. According to Serrano, the course has traditionally been small and academically demanding, usually attracting fewer than 30 students and sometimes as few as eight. This semester, however, enrollment jumped to 86 students.
The midterm results were far above anything Serrano had seen in previous years. The class average reached 96 out of 100, and 40 students earned perfect scores. Serrano noted that past midterm averages for the course typically ranged from 65% to 80%. He also said this year’s exam was more difficult than usual because take-home exams allow students more time and give instructors room to ask more challenging questions.
The sharp increase in perfect scores raised questions about whether some students had relied on AI tools to complete the exam. In response, Serrano announced that the final exam would be held in person instead of as a take-home assessment.
After the announcement, 18 students dropped the course. Another nine students stayed enrolled but did not show up for the final exam. Of those 27 students, 22 had received perfect scores on the take-home midterm.
The results of the in-person final were dramatically different. Among students who took the exam, the average score fell from the midterm’s 96 to just 48.
The situation highlights a growing challenge in higher education: how to handle AI in academic work. Universities around the world are moving away from simply trying to ban AI entirely and are instead developing clearer rules for when it can and cannot be used.
Many institutions now separate AI use into different categories. Acceptable uses may include brainstorming ideas, checking grammar, summarizing readings, or creating practice questions. Restricted uses often involve students relying on AI to produce assignments without proper disclosure. Prohibited uses generally include using AI during exams or on assignments where instructors have clearly banned it.
Serrano has taken a firm position on the issue, warning that normalizing cheating could have serious consequences beyond the classroom. He argued that society cannot afford for large numbers of talented young people to believe academic dishonesty is acceptable, saying such a mindset could lead to broader decline.
The Brown University case reflects a larger debate taking place across colleges and universities: AI can be a powerful learning tool, but it also creates new risks when students use it to avoid doing the work themselves. As AI becomes more advanced and easier to access, educators are likely to continue shifting toward in-person exams, oral assessments, handwritten work, and stricter academic integrity policies.
For students, the message is becoming clearer: AI may help support learning, but it cannot replace genuine understanding. For universities, the challenge is finding a balance between embracing new technology and protecting the value of education.
