A Brown University economics professor has raised concerns regarding the integrity of academic assessments after observing widespread use of generative artificial intelligence among his students. Roberto Serrano, who teaches Welfare Economics and Social Choice Theory, reported that a significant portion of his class likely utilized AI tools to achieve perfect or near-perfect scores on a take-home midterm exam conducted this past spring. The incident has prompted a broader discussion within higher education regarding how institutions should adapt to the proliferation of AI-enabled cheating.
According to reports, the decision to allow a take-home midterm was made in response to student anxiety following a violent incident on the Brown University campus in December, which resulted in the deaths of two individuals and injuries to nine others. Serrano noted that he felt it was an appropriate accommodation at the time, but he later expressed regret as he evaluated the unexpected results. The discrepancy between the high performance on the take-home assessment and student performance in subsequent, proctored environments led the professor to question the validity of the initial grades.
The Shift to In-Person Assessment and Academic Consequences
In response to the suspected use of AI, Serrano modified the format for the final examination, requiring students to complete it in an in-person setting. This policy change resulted in immediate attrition; more than a dozen students dropped the course, and others failed to meet the requirements for passing. Serrano provided students with an opportunity to validate their earlier performance, stating that if the distribution of final exam scores aligned with the midterm results, he would accept the original grades. However, he indicated that if the results were significantly lower, he would declare the midterm void and reweigh the final exam accordingly.
The subsequent data supported the professor’s suspicions. The final exam saw an average score of 48.6 percent—a historic low for the course, which had previously maintained an average final exam score above 65 percent. Additionally, 18 students dropped the class following the announcement of the final exam policy, and nine students who remained enrolled failed to sit for the final. Only a small fraction of the class demonstrated a level of consistency between their midterm and final exam scores.
Institutional Responses to AI-Enabled Cheating
The Future of Take-Home Exams
As of the end of the term, the professor maintained that the midterm would be reweighted based on the final exam performance, a decision that directly impacted the final grades of those who remained in the course.