Staff Reports
UC Berkeley Professor Richard Borcherds just did something refreshingly sensible: he quietly moved away from grading homework for correctness because generative AI makes that kind of grading meaningless. Instead, he gives completion credit for assignments and puts real weight on in‑person, device‑free exams. Call it old school, or call it common sense — either way, it’s a lesson other campuses should learn fast.
Fields Medalist goes old‑school to defend real learning
Professor Richard Borcherds, a Fields Medal winner and current UC Berkeley math instructor, told a campus paper that homework has become an unreliable signal of learning. Students can now feed problems into AI and get perfect answers. So Borcherds stopped grading for correctness and shifted to in‑person exams where phones and laptops don’t cheat for you. That’s the specific move making headlines, and it’s one worth talking about.
Why this matters for AI in education and academic integrity
Generative AI is useful, but it’s also a shortcut that makes homework look like learning. Research shows AI can raise scores immediately, but it doesn’t always build lasting skills. When professors grade answers that AI could have produced, they reward outsourcing, not thinking. That ruins the purpose of education and weakens academic integrity. Other teachers have quietly banned devices for quizzes and brought back pencil‑and‑paper tests for the same reason. If schools care about learning, they should follow suit.
Good policy — and a rebuke to campus administrators
Borcherds’ choice highlights a bigger problem: too many colleges treat technology like a silver bullet and forget their job is to teach students to think. Instead of embracing fads, universities should support instructors who adjust assessments to preserve honest work. That means more proctored exams, oral checks, and clear AI rules in syllabi. If administrators prefer shiny policy reports to real classroom results, students lose and credential inflation wins. And yes, professors should be allowed to change grading when AI makes old measures meaningless.
Teach students to think, not to outsource
The real victory in this story is simple: learning should matter more than polished output. Professor Borcherds’ move is a reminder that sometimes the best technology decision is to turn it off for a while. Schools that care about rigorous education will protect in‑person assessment and insist students do the hard work themselves. Otherwise, diplomas will keep getting easier to get and harder to trust — and nobody wants a degree that reads like an AI printout.
