Brown University Uncovers Massive AI-Assisted Cheating; Report Warns of Cognitive Decline
Key Takeaways
- ▸Dramatic score gap between unsupervised (96% average) and proctored (48.6% average) exams revealed systematic AI-assisted cheating among Brown economics students
- ▸Survey shows 56% of Brown undergraduates use AI tools daily or weekly; 25% of students submitting AI-completed assignments with rates increasing sharply each year
- ▸95% of faculty and 88% of students worry AI could damage cognitive development and critical thinking; 75% of faculty reported expectations of increased cheating
Summary
Brown University has issued a stark warning about generative AI's impact on student learning after an economics professor discovered evidence of widespread cheating. Professor Roberto Serrano's take-home midterm revealed a dramatic disparity: students averaged 96% on the unsupervised exam but just 48.6% on a proctored final, suggesting systematic use of AI tools to complete assignments. Serrano subsequently voided the midterm results, raising broader questions about institutional responses to AI-assisted academic misconduct.
Brown's Generative AI in Teaching and Learning (GAITL) committee subsequently published a report documenting the scale of the problem across campus. The study found that 56% of undergraduates use AI tools daily or weekly, with usage rates rising to 85% among master's students. Critically, around 25% of students are submitting assignments completed with AI assistance, a figure that has increased sharply year over year.
Beyond cheating, the committee found widespread concern among both students and faculty about AI's cognitive impacts. Eighty-eight percent of Brown undergraduates expressed worry that AI could harm their thinking abilities, while 95% of faculty feared negative effects on long-term learning. Four in five faculty members expect students' cognitive capabilities to decline as a result of over-reliance on these tools.
- Brown's response—voiding results and adjusting exam weighting—highlights institutional struggle to address widespread AI integration without clear policy frameworks
Editorial Opinion
Brown's findings suggest that banning generative AI is neither practical nor advisable—instead, institutions must fundamentally rethink assessment and pedagogy to address cheating while building genuine critical thinking. The real challenge isn't preventing students from using these tools; it's ensuring they develop deep understanding rather than outsourcing cognition. Educators should focus on designing assignments that require genuine problem-solving and moving beyond easily-gamed assessments, rather than futilely trying to ban tools that will only become more capable and ubiquitous.
