When Professor Roberto Serrano of Brown University altered his midterm exam format last spring, his primary concern was the mental well-being of his students rather than preventing cheating. This decision was influenced by the tragic shooting incident that had affected the campus, resulting in two students, including Ella Cook—who had recently expressed her desire for Serrano to be her academic advisor—among the casualties.
Serrano recounted their last conversation, noting, "She was a wonderful young woman, full of energy, full of ideas." The emotional toll of this tragedy led him to implement a take-home, closed-book midterm for his ECON 1170 course, which focuses on advanced mathematical economics. His aim was to alleviate the pressure of a traditional classroom exam at a time when many were still grappling with trauma. Notably, two of his students had sustained serious injuries in the shooting but ultimately survived.
However, instead of gratitude, Serrano found himself at the center of one of the Ivy League's largest cheating scandals involving artificial intelligence, as reported by El Pais.
The outcome of the exam was alarming. Out of the 86 students who took it on March 5, 40 achieved perfect scores, while the class average soared to 96, starkly contrasting with the typical averages of 65 to 80, despite the exam being designed to be more challenging. Serrano remarked, “The beauty of take-home exams used to be that we could push students to a higher level. The distribution of scores indicated something was amiss.”
Serrano discovered that some exam responses contained peculiar phrases that matched results from queries made to ChatGPT. His grader team tested the exam questions against the AI, uncovering that the AI provided unnecessarily complex arguments instead of simpler, more straightforward solutions. “This distribution made it clear that something was fundamentally wrong,” he said, deeming the situation "absolutely ridiculous."
Opting for a measured approach, Serrano did not invalidate the midterm. He informed students that the final exam would be in person, emphasizing that if the final grades did not reflect the midterm's distribution, only their performance on the final would count. When he revealed his findings to the class, he challenged the students, questioning their commitment to their education if they relied on AI instead of striving for personal growth and critical thinking.
The immediate aftermath was telling—27 students withdrew from the course, 22 of whom had scored perfectly on the midterm. When the final exam occurred, only 59 students participated, resulting in 19 failing, with the average score plummeting to an unprecedented 48. Serrano described the evidence of academic dishonesty as overwhelming, noting, “The distributions of the two exams confirmed my suspicions.”
Serrano reported the incident to Brown's dean and provost, initially receiving no response. After escalating the situation to the Academic Code Committee, he received a note acknowledging the incident as a “wake-up call,” yet the provost has remained silent.
Serrano, who holds a distinguished position as the Harrison S. Kravis University Professor of Economics and has published extensively in the field, expressed his frustration with the rapid advancements in AI that have left educational institutions scrambling to adapt. “Silence on this issue is the worst response we could have,” he lamented.
He emphasized that this issue extends beyond Brown University, referencing a New York Times article that highlighted an alarming trend of AI-assisted cheating at elite institutions like Stanford, where students appeared more interested in obtaining credentials than engaging in genuine learning. Serrano argued, “If institutions continually produce students who refuse to learn, the market will eventually recognize that the value of their degrees is diminished.”
As a broader trend unfolds, Serrano highlighted Princeton's recent decision to modify its long-standing honor code, mandating proctoring during exams—signifying a shift in academic policy that reflects growing concerns about integrity. Reports indicate that 57% of U.S. college students are using AI tools in their studies, raising alarms about declining independent reasoning skills, while Harvard students also reported a high incidence of cheating.
Looking forward, Serrano plans to implement significant changes in his teaching approach. For the upcoming academic year, coursework will shift to eliminate weekly assignments that could be easily completed using AI, and take-home exams will be discontinued. “The concept of a take-home exam has become obsolete,” he remarked.
While he recognizes the potential benefits of AI for student learning, Serrano stressed the importance of maintaining academic integrity. “We need to set guardrails to ensure that educational values aren't compromised,” he stated. Ultimately, he believes the issue transcends academia and calls into question the credibility and moral compass of higher education institutions. “If we fail to uphold truth and integrity,” Serrano warned, “what credibility can we as academics claim?”
