Caltech math professor stops grading homework for accuracy over AI

Yi Ni, a mathematics professor at Caltech, now grades problem sets only for completion because he cannot reliably tell whether students used AI to produce…

November 16, 2026 · Learning

Yi Ni, a mathematics professor at Caltech, now grades problem sets only for completion because he cannot reliably tell whether students used AI to produce the answers, according to a report by Business Insider summarised by Gadget Review. Caltech has no single AI policy and leaves the rules to each course. Its teaching centre points instructors toward checks that are harder to hand to AI: short in-class exercises, explaining work aloud, handing in work in stages, and grading the reasoning behind mistakes. The report notes these take more instructor time.

Why it matters to you: If a top science university can no longer trust take-home answers, every teacher faces the same question. Homework is shifting from something to grade to something to practise, with the real check moving into the classroom.

Try this: Teachers can redesign one assignment so the thinking shows, whether or not students use AI.

Here is a homework assignment I give my [grade or course] students: [paste it]. Redesign it so students must show their own reasoning, for example with a short in-class follow-up, a step they explain aloud, or a reflection on a mistake. Keep it fair for students without paid AI tools.

Source: Yahoo News (Gadget Review)

More from Today in AI

All of Today in AI. Where AI is heading.

Related Power of AI pages

Keep reading with Start here, everyday uses, the tool guide, the writing workshop, and sources and standards.