Roberto Serrano’s class did strangely well on their take-home midterm exam. Suspecting that AI cheating was widespread, students took the final exam face-to-face, and their grades plummeted.
The Brown University professor teaches welfare economics and social choice theory. Midterm exams were held from home after two students were killed in a mass shooting in December.
“The problem with this technology is that the cost of fraud has essentially become zero,” he told Business Insider. “It’s very easy for students to succumb to temptation.”
When he told students that the final exam would be held in person, many of the students who had previously received high scores dropped out. Athletes who scored in the upper 90s on the midterm exam also scored in the 50s in the final.
A graph of the data, first published by Inside Higher Ed, shows each student’s performance.
Brian E. Clark, Brown University’s vice president for news and strategic campus communications, wrote in a letter to Business Insider that Serrano shared details with the university’s Standing Committee on Academic Regulation on July 8, and that the committee is “proceeding according to procedure.”
“Brown takes all allegations of academic integrity with the utmost seriousness,” Clark wrote.
The scandal has sparked interest across the internet, especially among people working in the technology industry. Y Combinator co-founder Paul Graham posted about it on X. Two Google DeepMind staff members also shared their thoughts.
Serrano wasn’t shocked that the score change garnered attention, but was surprised by its scale. “I’m a little overwhelmed,” he said.
He said he received “hundreds of emails,” many from Brown University alumni. Colleagues who are off campus for the summer have also emailed him about this.
Some online commenters praised the student for consistently scoring high scores, first at 95.5 and then at 95. Serrano said the student was a “good student” who was “very familiar.”
Others yelled at students who consistently underperformed, first with 55 points and then with 59 points. “I respect that man,” Serrano said.
Glass AI founder Tom Henke explains the score table. X.com/@TomGlassAI
There is further debate about the benefits of this generation entering the workforce. Can students who use AI to cheat on exams be trusted with their efforts? Some argued that consistent scorers would be the best workers.
Serrano agreed. “I’m a champion of integrity, so I hire that person,” he said.
University professors continue to debate how to deny AI fraud. Last year, teachers told Business Insider that chatbots were creating tasks that were more difficult to complete.
Of course, Serrano’s grade distribution is not a perfect study. There are other reasons for the disparity in test scores, such as final exams being more difficult than midterm exams, and although universities are conducting investigations, mass cheating by AI has not been conclusively proven.
But this incident shows how concerns about AI cheating are becoming a headache for instructors. Serrano herself said she would never administer take-home exams again and would eliminate the homework portion of students’ grades.
He advised other educators to think critically about their own AI policies.
“This is definitely a wake-up call for professors,” he said. This requires caution.