started · updated
US universities rethink assessments as AI cheating rises
United States universities are grappling with the rise of 'homework scams,' where students use generative AI to complete assignments. Educators are increasingly finding that traditional written tasks are vulnerable to chatbot-generated content, leading some professors to implement 'traps.' For example, Timothy Paustian, a biology professor at the University of Wisconsin-Madison, inserted hidden prompts into assignments—such as instructions to mention the color 'orange'—to identify AI usage. In one instance, 60 out of 350 students were caught.
Despite these efforts, many institutions, including Yale and Cornell, are discouraging or banning the use of AI detection software due to its unreliability and high rate of false positives. Experts warn that these tools can be biased against non-native English speakers and are often bypassed by newer 'humanizing' AI tools.
As a result, faculty are rethinking assessment methods. Strategies include combining take-home assignments with oral exams, debates, or video presentations to ensure students are demonstrating genuine critical thinking and learning rather than merely reading AI-generated text.
Entities
Association of American Colleges and Universities · Cornell University · GPTZero · JAMA Pediatrics · Timothy Paustian · University of Wisconsin-Madison · University of Wisconsin–Madison · Yale University