When Brown University economics professor Roberto Serrano switched his class midterm from a traditional in-person exam to a take-home format, he aimed to accommodate students' anxiety after a campus tragedy. Instead, he discovered overwhelming evidence that most students exploited AI tools to cheat, revealed by a dramatic drop in scores when the final exam returned to in-person testing.
- Take-home midterm average was 96 out of 100; in-person final average dropped to 48.
- At least half the class likely used AI tools to cheat during the take-home exam.
- The episode sparks broader concerns about AI ethics and academic testing practices.
What happened
Following a tragic incident on campus in December, Professor Roberto Serrano shifted his ECON 1170 advanced economics course midterm exam to a take-home format to ease student anxiety about in-person testing. The move was unprecedented for his class's usually small enrollment and stricter testing standards.
The results conflicted sharply with historical data. The take-home midterm average soared to 96 out of 100, with nearly half the students scoring perfect marks. When Serrano administered the final exam in person, scores plummeted to an average of 48, revealing a stark contrast indicative of widespread use of AI assistance during the take-home exam.
Why it matters
Serrano's findings highlight the mounting challenge universities face in preserving academic integrity amid the rise of generative AI tools. The discrepancy between take-home and in-person exams quantifies a problem too often discussed only anecdotally, showing that AI can significantly distort student performance evaluations.
This phenomenon raises broader ethical concerns about reliance on AI in education and the potential erosion of genuine learning. Students themselves worry about how AI use affects cognitive skills, while educators are forced to reconsider assessment methods to balance fairness and rigor in an AI-pervasive environment.
What to watch next
The fallout from Serrano's public revelations may prompt institutions to revisit exam formats and incorporate stricter controls or new technologies to detect AI-enabled cheating. Universities will also likely expand discussions around AI's role in learning, including crafting policies that balance innovation with integrity.
Meanwhile, the experience at Brown serves as a case study for educators worldwide on the risks and realities of AI in education. Tracking policy responses, student attitudes, and emerging detection tools will be key to understanding how academia adapts to the ongoing AI disruption.