A Brown University professor's evidence of widespread cheating using mass AI.

      An economics professor at Brown University suspects that the majority of his students cheated using AI, and he has the data to support his claim. Roberto Serrano observed that the average score for his take-home midterm was 96 out of 100. However, when he switched to an in-person final exam, the average score plummeted to 48. He has taken his findings public, speaking to El País and Inside Higher Ed, asserting that he will pursue the matter further.

      The take-home exam format was implemented following a tragic incident. After a gunman killed two students on campus last December, many expressed feelings of anxiety when taking exams in a crowded room. To alleviate this, Serrano offered take-home midterm and final papers. The irony is palpable: the one time in decades that he relaxed the rules, a significant portion of the class resorted to cheating.

      The revealing data

      Serrano’s class, ECON 1170, is an advanced undergraduate economics course that typically attracts a small, capable group of students. He had never taught more than 30 students at a time, and once had as few as eight. However, this term saw 86 enrollments, likely due to the new take-home format.

      The midterm scores were, in his own words, extraordinary. The class averaged 96, with 40 students achieving a perfect score of 100. Historically, the average for the course ranges from 65 to 80, and this particular exam was harder than usual. Serrano believed that the take-home format provided an opportunity to challenge the class further, given the unlimited time allowed.

      The quality of the answers was also concerning. Many correct responses exhibited a "very convoluted style." When Serrano and his grading team input the questions into ChatGPT, they received similarly styled answers in return.

      The test he crafted to confirm his suspicions

      To confirm his suspicions, he devised a strategy. He informed the class that the final exam would be in person and that he would compare the distributions of scores. If they were similar, he would accept the midterm scores; if not, he would invalidate them and adjust the final's weighting.

      The results were telling. Eighteen students withdrew from the course, and nine others skipped the final exam. Notably, among these 27 students, 22 had scored a perfect 100 on the midterm, according to El País. Among those who did take the final, the average score dropped from 96 to 48. By Serrano’s estimate, at least 50 students engaged in cheating on the midterm, and he asserts that the evidence is compelling.

      A broader issue

      Brown is not the only institution facing this dilemma. A recent survey conducted among Princeton students revealed that 29.9 percent admitted to cheating on at least one exam or assignment, predominantly employing AI to do so. Over the past two years, schools have been working on developing detection tools and reconsidering their assessment methods.

      Students are feeling the pressure as well. A report led by Brown’s provost found that most undergraduates utilize generative AI on a weekly or daily basis. However, a significant majority also express concerns about its impact on their learning and worry about what it might do to their "cognitive capacity."

      This apprehension coexists with a larger shift as AI alters hiring practices and even influences how individuals think and write. Serrano expresses this concern in the clearest terms.

      The significance of the findings

      “We cannot afford to have a society in which a substantial portion of our brightest young minds believes that cheating is acceptable,” Serrano told Inside Higher Ed. “That leads to a deteriorating society, to a failed society. We cannot choose to become ignorant.”

      His experiment may be limited to just one class over one term, but it transforms a vague concern into concrete evidence. Remove the AI, and half the apparent knowledge disappears along with it. This is the reality that universities now must confront.