US universities shun AI detectors as cheating grows

US universities are moving away from relying on AI detection software as professors grapple with student use of chatbots to complete assignments.

Educators say chatbots can generate credible answers to written assignments within seconds, but detection tools can produce false positives and are becoming easier to evade.

Timothy Paustian, a biology professor at the University of Wisconsin–Madison, said he once inserted a hidden prompt into an assignment asking AI to mention the color orange. He said 60 of 350 students were caught using AI, but added that newer large language models and more careful students have made such traps less useful.

“I have pretty much thrown in the towel on classic writing assignments for lower-level classes,” Paustian told AFP. He said instructors are being pushed to make assignments more creative but warned that writing remains an important way for students to show they can think.

Four US university professors told AFP they are rethinking assessment, including by pairing take-home written work with oral exams, because they cannot rely on detection tools alone.

Universities including Yale and Cornell have banned or discouraged faculty from using AI detectors as the primary evidence of alleged cheating. University of Wisconsin–Madison faculty guidelines say the tools are imperfect, risk false positives, have shown bias against non-native English speakers and will not stop students from using AI tools.

Cornell University said detection technologies were unlikely to provide a workable solution because accurately identifying AI-generated content can be very difficult.

Researchers also warn that AI “humanizing” tools can rewrite chatbot-generated text to make it seem more human. Misinformation researcher Timothy Caulfield described the situation as a technology arms race between detectors and evasion tools.

A survey by the American Association of Colleges and Universities and Elon University found that 95 percent of college faculty feared students' overreliance on AI tools. A recent MIT report warned that getting answers from chatbots can create an illusion of learning and lead students to rely on AI when work becomes difficult.

Some academics also argue students need AI skills for modern workplaces. Harvard College Dean David Deming has suggested accepting or even encouraging AI in writing-based courses, according to the Harvard Crimson student newspaper.