Roberto Serrano has taught economics at Brown University for 34 years. This past spring, for the first time in nearly tw
Roberto Serrano has taught economics at Brown University for 34 years. This past spring, for the first time in nearly two decades, he gave his ECON 1170 students a take home midterm. The reason was humane: a mass shooting on Brown's campus in December had left several students anxious about sitting in classrooms. When the exams came back, 40 out of 86 students had scored a perfect 100. The class average was 96. In the three decades Serrano had been teaching the course, the midterm average had historically ranged between 65 and 80. He and his graders ran the exam through ChatGPT. The AI produced answers that, as Serrano told the Chronicle of Higher Education, were "kind of correct, but very off and with a very convoluted style." They matched what his students had submitted. Serrano moved the final exam in person. Twenty seven students dropped the course, twenty two of whom had scored 100 on the midterm. Among those who remained, the average score fell from 96 to 48. The gap between those two numbers is not a measurement error. It is the distance between what students know and what a large language model can produce on their behalf. What makes Serrano's case more than an anecdote is what happened next, or rather what did not. He filed a complaint through Brown's formal academic integrity process. The provost did not respond. The committee did not acknowledge receiving the case. After weeks of silence, Serrano went public in the Spanish newspaper El País in late June. Brown has since said the matter is under review. The institutional paralysis is not unique to Brown. A 2025 survey of 501 Princeton seniors found that 30 percent admitted to cheating at least once. Twenty eight percent reported using ChatGPT on an assignment that specifically prohibited it, more than double the rate from the prior year. The numbers were bad enough that Princeton's faculty voted in May, with near unanimity, to station proctors in every exam room starting July 1, a change now in effect. The decision ended a tradition dating to 1893, built on the principle that students could be trusted to police themselves. What is notable about Princeton's vote is that it was not imposed from above. Students themselves pushed for the change. The honor code had become a liability for the honest ones. Reporting a classmate for cheating meant risking doxing, teasing, and ostracism. When the cost of integrity exceeds the cost of silence, the system does not enforce norms. It punishes the people who follow them. The universities caught in this bind are facing a problem that has no clean technical solution. AI detection tools are unreliable and produce false positives that disproportionately flag non native English speakers. Banning laptops from exams treats the symptom, not the cause. Redesigning every course around in person assessment is logistically brutal for institutions that have spent a decade building hybrid and online infrastructure. And the fundamental asymmetry is only growing: the tools that generate plausible academic work are improving faster than any institution's ability to detect them. The deeper issue is what a degree means when the work behind it cannot be verified. An Ivy League diploma has always been a signal, to employers, to graduate schools, to the world, that the person holding it demonstrated a certain level of intellectual capability under conditions of genuine rigor. If 40 out of 86 students in an economics course at Brown can score 100 on an exam they did not meaningfully complete, the signal is degraded. Not for that one course or that one semester, but for every transcript, every recommendation letter, every hiring decision that relies on the assumption that the grades reflect the student. Serrano, in an interview with IBTimes, put it simply: "We cannot choose to become idiots." The question facing every university in the country is whether they have already made that choice by looking the other way.
