Top UC Berkeley Professor Ends Grading Homework as AI Makes Student Work Impossible to Verify

9 Sources

Share

Fields Medal-winning mathematician Richard Borcherds at UC Berkeley has stopped grading homework, citing widespread AI use that makes it nearly impossible to distinguish student work from AI-generated assignments. The decision signals a fundamental shift in how universities approach academic integrity and educational assessment methods.

Fields Medal Winner Abandons Traditional Homework Grading

Richard Borcherds, a Fields Medal-winning mathematician at UC Berkeley, has stopped grading homework for correctness, marking a significant shift in how AI in education is forcing professors to rethink traditional assessment methods

2

. The renowned professor now awards full credit to anyone who simply submits homework on time, acknowledging that distinguishing between human- and AI-generated assignments has become nearly impossible

4

. "Online exams are now a waste of time as several students just use AI to answer the questions. I cannot tell which students are using AI, but I notice the average score increases a lot in online exams," Borcherds stated

2

. His decision reflects growing concerns about academic integrity as AI cheating becomes more sophisticated and harder to detect.

Source: New York Post

Source: New York Post

Universities Struggle With AI Detection and Policy Implementation

UC Berkeley evaluated AI detection software but found it unreliable, prompting the university to give professors autonomy over their own AI policies as long as they outline rules in course syllabi

2

. This approach mirrors broader challenges facing higher education institutions across California. USC pays roughly $3.1 million annually for a ChatGPT Edu subscription covering 80,000 users, while simultaneously leaving AI policies to individual instructors

5

. Morgan Polikoff, a professor of education at USC Rossier, captured the sentiment shared by many educators: "The technology has been shoved down all our throats and we're all just dealing with the fallout"

5

. The impact of AI on education has created what one Arizona State University professor described as "a hypervigilant hunt for the hallmarks of AI-generated text"

1

, with educators suspecting that at least 90% of online students use AI for all or part of their work.

Source: CNET

Source: CNET

In-Person Exams Emerge as Primary Solution

Across California universities, professors are pivoting to in-person exams with no electronic devices as the primary method for grading homework and assessing student knowledge. Borcherds now bases grades mainly on these traditional exams

5

. UCLA statistics lecturer Thomas Maierhofer has adopted a hybrid approach, conducting all exams in-person on paper without phones or laptops, while grading outside work like homework and labs on completion only

5

. Research supports this shift: a Brown University professor saw students average 96% on take-home midterms—a dramatic jump from the historical average of 65% to 80%—but when final exams were administered in person, only two students scored within 10% of their midterm grades

2

. These findings demonstrate that unmonitored online exams no longer reflect what students have actually learned.

Source: TechRadar

Source: TechRadar

The Fundamental Challenge to Learning and Academic Integrity

The proliferation of AI-generated assignments since OpenAI released ChatGPT in November 2022 has created ethical dilemmas that extend beyond simple cheating. As one educator noted, "Education is about the journey, not the outcome"

1

. Students who outsource their work to AI risk undermining their own student learning and credibility. Borcherds explained that "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. And paying to grade AI generated answers is a waste of time and money"

2

. The challenge has become so acute that professors report confronting students who cannot explain their own creative output when questioned about AI-generated presentations

1

.

Mixed Perspectives on AI's Role in Educational Assessment Methods

Not all educators view AI as purely detrimental. Stanford Law School professor Julian Nyarko sees potential benefits, stating, "I think AI has tremendous potential. I think many humans are not great teachers"

5

. In a study Nyarko led, 16 law professors blindly rated answers to common student questions and preferred responses from Google's AI over their colleagues 75% of the time

5

. UCLA's Maierhofer is developing ALISSTAIR, a custom chatbot designed to support student learning, arguing that "as long as the exams are in-person and on-paper students are only hurting themselves if they use AI to cheat instead of using it to enhance their own learning"

5

. However, Nyarko warned about irresponsible use: "There's a way to use AI that will harm students, which is basically, just write this paper for me or solve this question for me"

5

.

Uncertain Future for Mathematics and Higher Education

Borcherds' decision coincides with broader questions about AI's impact on entire academic disciplines. The professor, who five years ago believed mathematics was among the fields least likely to be disrupted by AI, now admits uncertainty when students ask about pursuing math careers. "I have no idea what to advise," Borcherds said. "I even asked ChatGPT this exact question, and even ChatGPT said it didn't have any idea what was going to happen"

4

. This uncertainty extends across higher education as institutions grapple with how to maintain academic integrity while AI capabilities continue advancing. As Nyarko observed, "We're basically moving from a world where we kind of thought we knew how education works reasonably well, to a world where we don't know how to best educate students"

5

. Watch for universities to continue experimenting with educational assessment methods, potentially leading to standardized AI-proof measures for both college admissions and enrolled students.

© 2026 TheOutpost.AI All rights reserved