Brown professor exposes mass AI cheating in economics exam

A prominent economist at Brown University has exposed what he calls an 'overwhelming' case of mass AI-assisted cheating, rattling the Ivy League and reigniting a crucial debate about the role of artificial intelligence in higher education. Professor Roberto Serrano, a leading game theorist, announced that at least 50 students in his advanced mathematical economics course (ECON 1170) used ChatGPT on a midterm exam, making it the largest known academic integrity scandal at the institution.

The Anatomy of the Fraud

Serrano, who has taught at Brown for 34 years, offered a take-home, closed-book midterm exam in March to accommodate students still traumatized by a campus shooting that occurred the previous December. The exam was designed to be challenging, asking students to prove mathematical statements under novel assumptions. The results, however, were suspiciously perfect: the average score was 96 out of 100, with 40 students earning a perfect score.

Teaching assistants flagged the exams after noticing 'unusual passages' that matched output from ChatGPT. Serrano then administered an in-person final exam, which counted for 50% of the grade. The average score plummeted to 48 out of 100. Crucially, 27 students who scored 100 on the midterm did not even show up for the final exam, providing what Serrano calls 'overwhelming empirical evidence' of fraud.

A Cold Institutional Response

When Serrano reported the incident to Brown's administration, he received what he describes as 'absolute silence' from the president and a lukewarm acknowledgment from the dean, who called it a 'wake-up call.' Frustrated, Serrano took the case to the Academic Code Committee. 'Academic integrity is a value worth defending,' he told EL PAÍS. 'The faculty cannot be left on its own in a battle that is decisive if we want to preserve the future of higher education.'

Serrano suspects that institutional reluctance to act stems from the university's reliance on donations from wealthy families, whose children often receive the 'benefit of the doubt.' He is now overhauling his course: weekly exercises will no longer count toward the final grade, and take-home exams are permanently scrapped.

AI: The New Academic Temptation

The Brown scandal is part of a broader crisis. Princeton recently ended a 133-year-old Honor Code tradition that allowed professors to leave the room during exams, citing the ease with which AI enables cheating. Theo Baker, a recent Stanford graduate, wrote in The New York Times that he 'doesn't know a single person who hasn't used A.I. to get through some assignment.'

In Asia, the problem is even more acute. A recent experiment at the Hong Kong University of Science and Technology (HKUST) demonstrated that commercially available AI glasses can transmit exam questions to large language models and display answers on the lenses, effectively bypassing proctors. 'Every teacher feels that [education] is finding it hard to keep up,' said HKUST professor Zhang Jun.

The Case for Integration

Not all educators see AI as a purely destructive force. Liam Mayes, a Rice University professor and former honor board member, argues in the Houston Chronicle that universities should 'stop treating students' use of AI as cheating.' He contends that the real challenge is not preventing AI use but redesigning assessments to ensure genuine learning occurs. 'The question isn't how to keep AI out of education, but how to make good learning happen when AI is only ever a few clicks away,' Mayes wrote.

This view is echoed by Dr. Charlie Hannagin of USC's AI for Business Program and Kong Siu Cheung of the Education University of Hong Kong, who warns against 'outsourcing your thinking capability.' The consensus among these experts is that blanket bans are ineffective; instead, institutions must develop clear, nuanced policies that define acceptable AI use on a per-assignment basis.

A Personal Crisis for a Blind Professor

For Serrano, who lost his sight at age 17, the cheating scandal is deeply personal. He views his blindness as 'one more restriction' to optimize around, and he relies on technology himself. 'It hurts,' he said, 'that the one time in 34 years that I decided to offer a take-home exam, for highly justified reasons, the response was wide-scale fraud.' The shooting that prompted his leniency—a former PhD student opened fire on campus, killing two students and injuring nine—had already left him in a 'really bad place mentally.'

Serrano's story underscores the complexity of the AI-in-education debate. While tools like ChatGPT can democratize learning and assist students with disabilities, they also create unprecedented opportunities for dishonesty. As Serrano put it, 'If we no longer defend truth and decency and honesty, then what kind of credibility are we going to have as academics?'

What Comes Next

The Brown incident is a watershed moment. It proves that even elite institutions with rigorous honor systems are vulnerable to AI-enabled cheating on a massive scale. The response from universities will likely shape the future of assessment: expect a shift back to in-person, proctored exams and a move away from take-home assignments that can be easily gamed by LLMs.

However, as the HKUST experiment shows, even in-person exams are not safe; AI glasses and other wearable tech can provide real-time assistance. The long-term solution, many argue, is not technological countermeasures but a fundamental rethinking of what we test and why. 'We should use technology. We should use AI. We should not just say avoid it, stop using it,' said Kong. 'The bottom line is: don't outsource your thinking capability.'