AI Cheating Allegations: Brown University Professor Uncovers the Truth in Final Exam

A Brown University economics professor, Roberto Serrano, has drawn attention for uncovering widespread AI-assisted cheating in his class. According to a report by Solidot, Serrano first instructed his students to take their midterm exam at home in December 2025, following a school shooting incident. When the results showed that most students achieved perfect or near-perfect scores, Serrano became suspicious that AI had been used extensively for cheating.

In response, Professor Serrano decided to conduct the final exam in person. Instead of immediately invalidating the midterm results, he opted to first verify the scores from the final exam. He stipulated that if the score distribution from the final exam closely matched that of the midterm, the midterm scores would be included in the final grades.

The outcome demonstrated a clear disparity. Eighteen students dropped the class, nine students skipped the final exam altogether, and three students received a score of zero. The average score for the final exam was just 48.6%, marking the lowest in Serrano’s teaching career; previous finals had never dropped below an average of 65%. Only a handful of students managed to achieve scores on par with their midterm performance in the face-to-face final exam.

How AI Has Changed the Premise of Exams

This case highlights the fundamental challenges posed by generative AI to higher education. Traditional online exams have relied on the assumption that students would answer questions using their own knowledge. However, the advent of large language models (LLMs) like ChatGPT has shattered this assumption.

When students take exams at home, physically preventing access to AI tools becomes nearly impossible. While measures such as browser monitoring software and screen-sharing requirements exist, they cannot entirely prevent students from using smartphones or other devices to access AI platforms.

Professor Serrano’s approach was simple yet effective: comparing the score distributions of the midterm and final exams to visualize discrepancies caused by AI reliance. The fact that many students who achieved near-perfect scores on the midterm performed poorly in the face-to-face final exam strongly suggests a heavy dependence on AI during the home-based test.

The Severity of Class Dropouts and Score

Distribution

What stands out in this case is that 18 students dropped the class, and nine others chose not to attend the final exam. In total, 27 students avoided the in-person exam, likely because they felt they couldn’t perform at the same level without the aid of AI.

The fact that three students scored zero is also noteworthy. This could indicate that they were entirely unprepared for the exam or chose to submit blank answers for unknown reasons. The average score of 48.6% was the lowest in Serrano’s career, a stark contrast to the historical average of over 65%.

The few students who maintained comparable scores between the midterm and final exams likely relied on their own knowledge and avoided AI assistance during the midterm. This disparity in performance between students who depended on AI and those who didn’t highlights a clear inequity in educational evaluation.

Rethinking Educational Assessment

The issue of AI-based cheating is not unique to Brown University; similar instances have been reported nationwide, leaving educational institutions scrambling for solutions. While some universities are reintroducing face-to-face exams and implementing policies explicitly banning AI use, these measures have not yet led to a comprehensive solution.

The fundamental question revolves around the nature of educational assessment in the digital age. With AI tools readily accessible, tests based on rote memorization or knowledge recall are losing their relevance. There is an urgent need to develop new evaluation criteria, focusing on critical thinking, problem-solving abilities, and collaborative work using AI tools.

Professor Serrano’s case serves as a symbolic example of this ongoing transformation. Educational institutions must move beyond simply banning AI and redefine the purposes of education itself, while overhauling assessment methods to align with the realities of an AI-driven world.

Editorial Opinion

In the short term, this case is likely to prompt universities across the country to accelerate the reintroduction of face-to-face exams and the adoption of AI detection tools. Many educational institutions will likely enforce stricter online exam protocols starting in the fall semester of 2026. However, there are inherent limitations to technological countermeasures, as the rapid evolution of AI may outpace the development of detection tools.

From a long-term perspective, it seems inevitable that educational assessment methods will undergo significant transformation. Traditional exams focused on knowledge recall may give way to project-based evaluations, portfolio reviews, and frameworks that assess students’ ability to collaborate with AI for problem-solving. These changes could ultimately extend to college entrance exams and professional certification processes.

The coexistence of education and AI is inescapable. Our editorial team believes that instead of solely framing AI as a tool for academic dishonesty, more discussions should focus on leveraging AI to enhance the quality of education. Alongside the reform of evaluation methods, there is an urgent need for curricula that teach ethical and effective use of AI.

References

• “布朗大学经济学教授怀疑班级里多数学生使用 AI 作弊”, by (author unknown) — Solidot, 2026-07-11T16:40:28.000Z (ARR)
• Source URL: https://www.solidot.org/story?sid=84806