AI-Assisted Cheating Allegations Surface at Brown University
An economics professor at Brown University believes that a take-home exam, intended to support students affected by a campus tragedy, has unveiled what he considers to be one of the major AI-assisted cheating scandals in Ivy League history.
Professor Roberto Serrano grew concerned about the rising use of AI after observing unexpectedly high scores on take-home midterms in his Welfare Economics and Social Choice Theory class during the spring semester. This shift followed a mass shooting incident in December that took the lives of two students, prompting Serrano to convert the exam to a take-home format to alleviate stress for those grappling with the aftermath.
Notably, out of 86 students, 40 achieved perfect or near-perfect scores on the midterm, leading to an average of 96 percent, in stark contrast to previous average scores of 65 to 80 percent in earlier iterations of the course. Serrano had designed this exam to be more challenging, allowing unlimited time for completion.
After noticing similarities between some students’ answers and ChatGPT responses, Serrano suspected that many students had leaned on AI instead of working independently. Instead of dismissing the interim results, he decided to hold final exams in-person in May.
Serrano communicated to his students that if the final scores mirrored the midterm distribution, those midterm results would be considered valid. However, he anticipated a discrepancy, leading him to state he would invalidate the midterms and reassess the finals accordingly.
This announcement reportedly led some students to withdraw from the course. For those who remained, the average score plummeted to approximately 48 percent, with 19 failing the course.
Out of the 59 students who took the final exam, only two had midterm scores aligned with their final results. One student scored 95.5 on the midterm and 95 on the final, while another was the only one to improve, going from 55 points on the midterm to 59 on the final.
After analyzing the outcomes, Serrano submitted the findings to Brown’s Standing Committee on Academic Regulations. However, he later went public, urging university officials to take formal academic integrity actions against students who allegedly breached university policy.
Expressing frustration over the university’s handling of the issue, Serrano felt it was imperative for the institution to confront the realities posed by generative AI. His assertions come as Brown University’s Committee on Generative AI in Teaching and Learning is assessing the growing influence of technology on campus. A recent report suggested avoiding overly complex assignments and emphasized reducing punitive measures.
The report stated that universities should refrain from implementing strict rules on GenAI use, as definitive detection is challenging, and standards are likely to evolve over time.
Upon returning to class, Serrano posed a thought-provoking question to students: if their intention was merely to cheat, why were they in college? He questioned the purpose of their education if they were unwilling to exert effort in developing critical thinking skills.
He further stressed the importance of addressing AI-related misconduct in higher education as generative artificial intelligence becomes more ubiquitous. “We can’t have a society where many talented young people regard cheating as acceptable,” he remarked. “We cannot afford to be naive.”
