When Brown University Professor Roberto Serrano changed the format of his midterm exam last spring, he was thinking about his students’ mental health, not academic fraud. Two of them had been shot, including Ella Cook, a young woman who had sat in his office just days before the December 13 massacre at Brown University and asked him to be her academic advisor.
“We had a very nice conversation,” Serrano recalled in an interview with Fortune. “She was a wonderful young woman, full of energy, full of ideas. Imagine my shock when a few days after that conversation took place, they released the names of the two mortal victims, and I saw that one of them was her.”
In that grief, Serrano made a decision he had never made in his 34-year career at Brown and gave his ECON 1170 class—an advanced undergraduate course in mathematical economics—a take-home, closed-book midterm. He wanted to remove the stress of sitting in a classroom on a campus where, he says, quite a few students were still too traumatized to set foot. Two of his students had been among the nine wounded in the attack; they fought for their lives for weeks, and both survived.
What Serrano got instead of gratitude was the largest known AI-assisted cheating scandal in the Ivy League, as previously reported by El Pais.
Cheating on a mass scale
Of the 86 students who took the March 5 exam, 40 scored a perfect 100. The class average was 96 whereas in previous years, the average had ranged between 65 and 80—and this exam, by design, was harder than usual. “The beauty of take-home exams used to be that we professors were able to challenge students a little more, just to push them to a higher level,” Serrano said. “The fact that this was a harder exam and this distribution made it absolutely clear that something very unusual had happened.”
Serrano got tipped off by something that was just too smart, he said. “Some answers contained unusual passages that coincided with results obtained after running the questions through ChatGPT,” Serrano said. His graders ran the exam questions through ChatGPT and made a telling discovery: the AI had generated a convoluted argument for a problem that has a much simpler, more elegant proof, and that same convoluted reasoning appeared across dozens of student exams. “This distribution made it clear that something seriously wrong had happened,” he said, calling it “absolutely ridiculous.”
But Seranno said he decided to give his students the benefit of the doubt: He wasn’t going to void the midterm, but told them the final exam would be in person. If the grade distribution didn’t roughly mirror the midterm’s, only the final would count.
When Serrano returned to class after grading, he told his students exactly what he’d found. “If you did this, if you just press a button to ask an AI agent to do this for you, you’re showing to be completely irrelevant. So my question to you is, why are you here? Why are you at a university if you refuse to learn, you refuse to work hard, if you refuse to put in the necessary effort to develop critical thinking?”
“If all you’re doing is just pressing a button to do to have this machine do the work for you, then you think you need a Brown degree for that?”
When asked about the initial reaction from his students, Serrano answered with just one word: “silence.” He suspected the cheaters weren’t even there: “I think most of the cheaters were not in class, frankly.” He closed class that day by reminding the students of the honor code. “You all signed this, right? Sadly, that’s the value of your signature.”
Following his speech, 27 students dropped the course; 22 of those had scored 100 on the take-home.
When the final came around, only 59 showed up for the in-person exam and 19 failed. The class average collapsed to 48 out of 100: by far the lowest final exam average in the course’s history. “The empirical evidence of fraud is overwhelming,” Serrano said. “When you put together all this information and the distributions of the two exams, it’s absolutely clear.”
After assembling his evidence, Serrano sent it to Brown’s dean of the college and provost. Neither responded initially. After he escalated the case to the university’s Academic Code Committee, he received a note calling the incident “a wake-up call.” The provost, he said, has maintained complete silence to this day.
The man who wrote the book on game theory explains game theory
Serrano holds a named chair, the Harrison S. Kravis University Professorship in Economics, among the most prestigious appointments a university bestows. He serves as an editor at Games and Economic Behavior, the leading journal in a field that covers the economics of risk, uncertainty and information, often known as “game theory,” exactly what’s at play when, say, cheating on an exam.
Serrano has over 6,100 citations on Google Scholar and is the author of two widely used textbooks, including the one Brown’s own economics department uses. He’s a fellow of prestigious academic societies and even got the King of Spain Prize for Economics in 2024.
The game theory expert looks at the current situation and despairs. “I’m very frustrated,” Serrano told Fortune. “I believe the arrival of AI has been like a tsunami for all of us. It’s caught everybody unprepared. But in my humble opinion, silence is the worst treatment for this problem.”
Serrano, who has been blind since age 17, earned his PhD at Harvard, and has spent more than three decades at Brown, acknowledged that AI has moved so quickly that institutions haven’t known how to respond.
According to a representative of Brown University, this is something the university is grappling with. “The rapid proliferation of generative AI has raised many questions for universities, from how new technologies can advance educational innovation to how to define acceptable and unacceptable uses for teaching and learning,” Brown’s Vice President for News and Strategic Campus Communications Brian E. Clark said, before explaining how the university is integrating AI into classroom instruction and hired a new associate provost for AI.
“Among many areas of focus, this includes examining Brown’s academic codes in the context of generative AI. As that work progresses, Brown’s Standing Committee on the Academic Code addresses charges of academic code violations, as has been the case for decades. The charge of the committee, which is an elected, faculty-led body, is to investigate reported violations of Brown’s academic code to determine whether or not a violation occurred and decide on any penalty for those found in violation.”
But it’s not just Brown dealing with AI, Serrano said. He pointed to a recent New York Times essay that described a pervasive culture of AI cheating among Stanford peers: students who were at elite universities not to learn but to collect the credential. “What they miss in that very naive analysis,” Serrano said, “is that the Brown label is Brown for a while. But if Brown continues to produce mediocre students who refuse to learn, sooner or later the market is going to find out that the Brown label is not what it used to be.”
The broader trajectory, he warned, points somewhere darker. “If workers are just going to press a button to ask an AI agent to do the work for them, that’s inscribing a world in which humanity has chosen to become idiots,” he said. “We stop thinking.”
Brown is far from alone. Princeton’s faculty voted in May to end its 133-year-old honor code tradition of unsupervised exams, mandating proctors in every room starting July 1, the most significant change to the policy since students first petitioned for it in 1893. As Fortune reported in May, 57% of U.S. college students now report using AI tools in their coursework weekly. A separate Fortune analysis found that AI is causing measurable cognitive atrophy among students, with educators warning of a “great unwiring” of the ability to reason independently. And just last week, 47% of surveyed Harvard seniors admitted to cheating.
Serrano has already made changes for the coming academic year. Weekly homework assignments will carry zero weight toward final grades, since those can be completed with AI. Take-home exams are gone, permanently. “Unfortunately, the idea of a take-home exam is a thing of the past,” he said. “It’s too easy for students to succumb to temptation.”
“I’m sure there are appropriate uses of AI: it has the potential to be something very useful for students that will contribute to learning,” he said. “But we have to be absolutely clear about the risks it poses to academic integrity, which is a value we cannot drop.”
The final word, for Serrano, is not about exams or grade distributions. It’s about what kind of people universities are producing. “We need to establish the necessary guardrails — and if they fail, be prepared to implement consequences,” he said. “But this is bigger than academia.” “If we no longer defend truth and decency and honesty,” Serrano said, “then what kind of credibility are we going to have as academics?”
This article has been updated with a statement from Brown University.
