A professor at the University of California, Berkeley, has stopped grading students’ homework mainly on the correctness of their answers after noticing that artificial intelligence (AI) tools were helping students get high marks. Richard Borcherds, a mathematician who received the Fields Medal in 1998, said the widespread use of AI has made homework grading less useful, reported NYPost .
However, the Daily Californian reported that these tools were not considered reliable enough for standard use across the campus. He explained that giving marks based on correct answers could encourage students to use AI instead of solving questions themselves. Borcherds added that paying people to check answers generated by AI wastes time and money. He now mainly gives students credit for completing their assignments rather than for getting every answer correct. He explained that AI tools sometimes made noticeable mistakes in the past, but they have improved so much that such errors are now rare. The professor also expressed concern about the future of mathematics. He recalled that the subject had appeared less likely to be affected by technological changes just five years ago. he has noticed that average scores are much higher in online examinations Although he cannot identify exactly which students use AI. He also suggested that university mathematics departments might continue working as they do now for several years, even as AI improves. UC Berkeley has issued guidance asking teachers to explain their rules on AI use in their course syllabus. The university has also examined AI-detection tools, including Turnitin and Rumi. He added that he would feel relieved if making homework assessment more difficult turned out to be AI’s worst effect on society.
Because some students use AI to answer questions, he described online exams as a waste of time. Individual teachers can therefore decide their own classroom rules for AI.
Speaking to The California Post, Borcherds said many students now use AI to complete their homework. He also said this could unfairly affect students who complete their work honestly. Borcherds said it has become difficult to tell whether an assignment was written by a student or produced by AI. In an interview with the Daily Californian, UC Berkeley’s student newspaper, he said that if AI continues to improve at its current rate, it could perform every task that human mathematicians can do, and do it better, within a few years. He also said he was unsure whether students should still choose mathematics as a career. Borcherds said he now mainly bases grades on in-person examinations, where students cannot use electronic devices. Anne Marie Richard, the university’s assistant vice provost for undergraduate education and chief academic technology officer, said instructors are expected to outline course requirements, including rules for using AI tools. Borcherds said the difficulty of grading homework might be a small concern compared with the wider effects of AI on society.
