Math professor changes grading method

Richard Borcherds, a mathematician who won the Fields Medal in 1998, has stopped grading students based on the accuracy of their answers. He noticed that AI tools helped students achieve top grades, and he believes grading only on accuracy encourages them to use AI instead of doing the work themselves.

AI use in education becomes widespread

Borcherds said AI use among students is now very common. He explained that grading based on accuracy penalizes honest students while rewarding those who use AI. He now focuses on whether students submitted their work rather than the quality of their answers.

AI detection in exams raises concerns

In July 2026, history professor Jason Gibson used a hidden trick to catch students using AI during a midterm exam. He inserted invisible instructions that prompted AI to add absurd sentences about Madagascar into essays on the industrial revolution. Thirty-two students failed after copying these nonsensical responses without reading them.

Generative AI enters classrooms rapidly

By March 2026, generative AI had become a major force in classrooms, spreading quickly without much time for educators to prepare. Three years after the rise of ChatGPT in schools, teachers and administrators were struggling with the impact of AI on education. The debate had shifted from cheating to the purpose of education itself.

Distinguishing AI from human work is hard

A professor at the University of Berkeley also stopped grading based on accuracy after realizing AI helped students get high marks. He said it's hard to tell AI-generated work from that of students. AI has improved so much that it rarely makes obvious mistakes, making it even harder to detect.

AI challenges the future of mathematics

Borcherds expressed concerns about the future of mathematics if AI continues to advance. He said AI may soon be able to do everything human mathematicians can do. He questioned whether future mathematicians should even pursue the field, noting that mathematics had once seemed safe from technological disruption.

Universities may lag in adapting to AI

Borcherds noted that universities might not adapt quickly to AI's rapid progress. He compared the situation to outdated practices like printing paper journals despite technological advances. He predicted that mathematics departments might continue functioning as they do today for years, even as AI improves.

Grading shifts to in-person exams

Borcherds now bases grades mainly on in-person exams, where students can't use electronic devices. He said online exams are ineffective because some students use AI to answer questions. He noticed that grades during online exams rose significantly, even though he couldn’t identify which students used AI.

Universities struggle with AI policies

UC Berkeley has created guidelines for faculty to address AI use in courses. The university asked instructors to include rules about AI in their syllabi but found that AI detection tools like Turnitin and Rumi were unreliable. As a result, instructors still decide AI policies for their own classes.

AI's impact on education is debated

For Borcherds, the challenges of grading may seem small compared to AI’s broader effects. He said if the worst AI impact is making grading harder, he would be relieved. However, studies show many teachers believe AI negatively affects learning, especially in countries like France, where traditional methods are seen as more effective for building basic skills.