A professor at Brown has demonstrated evidence of widespread cheating using artificial intelligence.

      A professor of economics at Brown University, Roberto Serrano, suspects that most of his students cheated using AI tools and has data to support this claim. He noted that the average score for his take-home midterm was an impressive 96 out of 100, but when he switched to an in-person format for the final exam, the average plummeted to 48. Serrano has made this issue public, speaking to El País and Inside Higher Ed, and is determined to pursue it further.

      The decision to have a take-home format stemmed from a tragic event in December when a gunman killed two students on campus. Many felt uncomfortable taking exams in crowded settings, which prompted Serrano to offer take-home midterms and finals to alleviate their anxiety. The irony is that during the one semester he loosened the rules, a significant portion of the class ended up cheating.

      The data revealed the issue. Serrano’s course, ECON 1170, is an advanced undergraduate economics class typically attended by a small, capable group of students. He had never had more than 30 students in a class before, and at one point had just eight. This semester, however, enrollment spiked to 86, likely due to the new take-home format.

      The midterm results were, in Serrano's words, astonishing. The class averaged 96, with 40 students achieving a perfect score of 100. Historically, the average for this course ranges from 65 to 80, and this year's exam was more challenging than usual. Serrano thought that having unlimited time with the take-home format would encourage deeper engagement.

      However, the style of many correct answers seemed off, characterized by a "very convoluted style." When Serrano and his graders tested the questions with ChatGPT, they received similarly phrased answers.

      To investigate further, he devised a plan. He informed the class that the final exam would be in person and that he would compare the score distributions of the two assessments. If they matched, he would keep the midterm results; if they didn't, he would invalidate the midterm and adjust the weighting of the final.

      The outcome was telling. Eighteen students dropped the course, and nine more did not take the final exam. Of those 27 students, 22 had scored a perfect 100 on the midterm, according to El País. Among the students who did take the final, the average dropped dramatically from 96 to 48. Serrano believes that at least 50 students cheated on the midterm, considering the evidence to be compelling.

      Brown University is not the only institution facing this issue. A survey conducted at Princeton found that 29.9 percent of students admitted to cheating on at least one exam or assignment, primarily using AI tools. Schools have spent the last two years focusing on detection methods and reevaluating their testing practices.

      Students are feeling the impact as well. A report from Brown, led by the provost, indicated that most undergraduates use generative AI weekly or daily, while a significant majority express concerns about its effect on their learning and cognitive abilities.

      This concern is part of a larger trend, as AI begins to change hiring practices and the way people think and write. Serrano articulates the issue starkly, stating, “We cannot afford to have a society in which a significant fraction of our best young minds think that cheating is okay. That leads to a declining society, to a failed society. We cannot choose to become idiots.”

      His experiment may involve only one class and one term, but it clearly highlights a pressing concern. Remove the influence of AI, and half of the perceived knowledge vanishes. This revelation is something universities now need to confront.