Professor Roberto Serrano of Brown University identified that students used artificial intelligence during a remotely administered exam. This suspicion arose after a sharp decline in grades when a second assessment was conducted in person, without permission to use electronic devices.
Serrano teaches the course on welfare economics and social choice theory. Although considered challenging, this year's class was large, possibly because he announced that the midterm and final exams could be taken at home. This flexibility occurred due to a shooting attack on the Brown University campus in December 2025, which resulted in two deaths, allowing the course to continue after the tragedy.
During the midterm exam, students achieved remarkable results, with an average score of 96 out of 100. This performance was significantly higher than the historical average for the discipline, which usually fluctuated between 65 and 80 points. The professor began to become suspicious, especially since the exam had been designed to be more difficult than previous ones.
Serrano's suspicion was reinforced by observing that many answers presented a 'very elaborate style.' When testing the exam questions in ChatGPT, he noticed a similarity between the responses generated by the artificial intelligence tool and those provided by the students themselves.
To confirm his suspicions, Roberto Serrano opted to administer the final exam in person, prohibiting access to cell phones and computers. He informed the students that the grade from the midterm exam would be recalculated based on the performance in the final exam. In response to this notice, 18 students dropped the course, and nine others did not attend the assessment.
Among the students who took the final exam, the average score plummeted from 96 to 48 points, representing a fifty percent reduction. This confirmed the professor's suspicion that AI was used in the remote midterm exam. The topic has sparked intense debates, with Serrano reporting being overwhelmed by hundreds of emails from former students, professors, and other interested parties.
A central point of the debate is the ability to trust the new generation of students in tasks requiring logical reasoning, given the access they have to 'technological shortcuts' to solve complex problems. However, Serrano emphasizes that generalization should not occur; he cited a student who scored 95 on the first test and 95.5 on the second, indicating personal effort. Furthermore, ethics were addressed by mentioning another student who scored 55 on the first test and 59 on the second, demonstrating that he did not use AI on the first assessment, which led the professor to express admiration for this student.
