UC Berkeley Math Professor Drops Graded Homework and Online Tests as AI Use Spikes

The rapid advancement and widespread accessibility of generative artificial intelligence have triggered a profound shift in higher education, forcing prominent institutions and educators to rethink fundamental methods of instruction and evaluation. At the University of California, Berkeley, one renowned mathematics professor has taken the drastic step of eliminating graded homework and online tests entirely, concluding that the pervasive use of artificial intelligence has rendered traditional take-home assignments virtually pointless.

According to reports from the New York Post, distinguished mathematician Richard Borcherds decided to abandon online testing and graded homework assignments after observing that it has become nearly impossible to distinguish between genuine student work and content generated by sophisticated AI models. In response to this technological reality, Borcherds has shifted his evaluation strategy, relying instead on traditional, in-person examinations where electronic devices are strictly prohibited to accurately gauge how well his students are mastering the course material.

The challenges posed by generative artificial intelligence in academic settings have been a growing concern for educators across disciplines, but the mathematical sciences present unique hurdles given the step-by-step nature of problem-solving and the ability of modern algorithms to instantly solve complex equations and provide detailed proofs. Borcherds, a Fields Medalist known for his profound contributions to mathematics, pointed out the clear futility of continuing with remote or take-home assessments in the current technological landscape.

"Online exams are now a waste of time as several students just use AI to answer the questions," Borcherds told the publication, highlighting a frustration shared by a growing number of university instructors worldwide. He noted that while he cannot always definitively prove which individual students are relying on unauthorized technological assistance, the behavioral patterns are unmistakable. Specifically, he pointed out that the average scores on remote assessments increase dramatically, pointing to widespread reliance on external algorithmic help rather than individual mastery of the curriculum.

Furthermore, the professor highlighted the paradoxical nature of grading homework in an era where AI tools can complete assignments in seconds. Continuing to grade these submissions not only consumes valuable institutional and instructional resources but also creates an unfair environment that disadvantages honest students. "Grading for correctness just adds a huge incentive for them to use AI for homework instead of doing it themselves and penalizes students for honesty," Borcherds explained. He added that allocating financial resources and teaching assistant hours to review and grade artificial intelligence-generated answers represents a fundamental misallocation of time and money.

The decision by the UC Berkeley professor reflects a broader, ongoing debate within academia regarding how universities should adapt to generative tools like large language models and advanced problem-solving platforms. When these technologies first burst into mainstream consciousness, many educational institutions rushed to issue blanket bans or establish strict honor codes prohibiting their use. However, as AI tools have become more sophisticated, ubiquitous, and deeply integrated into everyday software, enforcement has proven increasingly difficult.

Many professors find themselves caught in a difficult dilemma. On one hand, take-home assignments and regular homework sets have historically been viewed as essential components of the learning process, allowing students to practice complex concepts at their own pace outside of high-stress testing environments. On the other hand, when algorithms can effortlessly generate correct responses to advanced calculus, linear algebra, and theoretical mathematics problems, take-home work risks transforming from an exercise in critical thinking into an empty ritual of copying and pasting.

By returning to traditional, proctored, in-person exams without the aid of computers, tablets, or smartphones, Borcherds is effectively stripping away the digital layer to refocus on direct human assessment. This back-to-basics approach ensures that the grades students receive reflect their own cognitive abilities and understanding of mathematical theory rather than the computational power of the software they can access behind a screen.

As universities continue to grapple with the long-term implications of generative artificial intelligence, the actions taken by faculty members at institutions like UC Berkeley serve as a significant indicator of changing academic norms. While some departments are experimenting with incorporating AI literacy into their curricula and designing assignments that explicitly require critical engagement with AI outputs, others are choosing to circle back to foundational evaluation methods to preserve academic integrity. For Borcherds and his students, the classroom has returned to a strictly offline environment where pencil, paper, and human intellect remain the ultimate standards of achievement.