Brown University students average 96 on advanced mathematical economics midterm exam

An advanced mathematical economics exam at Brown University produced a seismic shift in results last semester, with the average score soaring to 96 - a number that left the professor stunned. Past averages for the same midterm had ranged from the 60s to the 80s, and this year nearly half the class achieved a perfect 100. Roberto Serrano, who teaches the course, saw the anomaly immediately. The scores represented a departure from every previous semester's distribution, raising immediate questions about what had changed. No curriculum overhaul or pedagogical shift preceded the test. A class transformed overnight The jump in performance was not marginal. In prior years, the exam had reliably separated students across a normal curve, with a minority cracking 90. This semester, the curve collapsed. Students who had previously posted mid-range results suddenly joined the top tier. The results landed in a semester when AI tools capable of solving complex mathematical proofs have become widely available and easy to use. Universities across the country have scrambled to update honor codes and assessment methods in response to tools like ChatGPT and specialized math solvers. Brown's own academic integrity policy explicitly prohibits unauthorized assistance, including AI, unless the instructor permits it. What the data doesn't say No public evidence has linked the score surge to AI use. Serrano has not released a statement on the cause, and the university has not disclosed whether an investigation is underway. The numbers alone cannot explain why the scores changed. Still, the timing and scale of the anomaly have put a spotlight on the vulnerability of traditional assessments. For educators, the incident is a data point in a growing pattern: when online proctoring is absent and AI tools are unrestricted, exam results can become unreliable signals of student learning. Why this matters for educators The Brown exam case is a reminder that assessment design must evolve alongside technology. If a test's integrity can be undermined by a tool that students already carry in their pockets, the test itself needs rethinking. Educators can explore strategies like in-class writing, oral exams, and problem sets that require students to show their reasoning step by step. Resources such as the AI Learning Path for Teachers offer practical frameworks for redesigning coursework in AI-fluent classrooms. For those tracking the broader impact of AI on academic work, AI for Education provides ongoing coverage of how schools and universities are adapting their policies and practices. The Brown results may be an outlier, but they underscore a challenge that every institution now faces.