UC Berkeley math professor stops grading answers, credits homework completion amid AI surge

Richard Borcherds, a mathematics professor at the University of California, Berkeley and 1998 Fields Medal winner, has stopped evaluating homework for correctness, opting instead to give credit for on-time submission because AI-generated answers are increasingly indistinguishable from student work. He argues that grading for accuracy incentivizes cheating and punishes students who do the work themselves. While admitting the new method is imperfect, Borcherds has no ready alternative.

Across U.S. campuses, faculty are turning to oral exams, presentations, and other in-person assessments to gauge true comprehension. A Cornell study of more than 95,000 students found 37 % use generative AI at least monthly, with 9 % admitting to cheating. Experts like Rene Kizilcec warn that AI misuse threatens the credibility of university credentials, prompting institutions such as MIT and the University of Pennsylvania to rethink assessment strategies.