10
Loading video...
By Amanda Greenwood and Hollie Silverman
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
A University of California, Berkeley professor has stopped grading students' homework for accuracy over concerns that AI is making it harder for educators and teachers to use traditional written assignments to assess students' abilities.
Richard Borcherds, a mathematics professor at UC Berkeley who won the Fields Medal in 1998, now gives students credit for completing their homework on time, rather than assessing whether their answers are correct.
Borcherds believes there’s no point in marking papers that have been generated with AI, telling Newsweek in an email Monday: "AI can now solve many undergraduate mathematics problems, often better than the students themselves. This makes it increasingly difficult to tell whether a homework solution represents the student's own understanding or was generated by AI.
Read More on News
"Of course, students have always been able to get help from other sources, but AI makes this much easier," he noted.
His move comes as universities across the country grapple with the rapid advancement of AI and its impact on education, with some professors turning to oral examinations, in-person assessments and other methods designed to establish what students actually understand and can do.
...
Borcherds told Newsweek students' use of AI for homework is now widespread.
"I hope that removing grades for correctness will reduce the incentive to just ask AI for answers and increase the chance that they do the homework themselves," Borcherds explained, saying, "Students still need to understand the material to do well on examinations."
Borcherds has consequently moved to grading homework based on completion, meaning students receive credit for just handing in the work rather than for getting the answers right.
"Assessing students in online courses is now hard because of AI, and I do not know of a good solution," Borcherds said.
He did note that "in-person exams still work fine for assessing students."
Eric Klopfer, a professor at MIT who co-chaired the university’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training, told Newsweek that homework has traditionally served two purposes: helping students learn and allowing educators to evaluate that learning.
But generative AI has complicated the latter. "At this point, at least for some students, homework is no longer a valid evaluation of their learning," Klopfer said.
He cautioned that simply reducing homework's importance in grading can create another problem. If homework counts for less, students may put less effort into it, even though the assignments can still provide a valuable learning experience.
Klopfer said addressing this would require what MIT's report described as a new "social contract" in education, with students taking greater responsibility for their own learning and faculty placing greater trust in them.
Research by Cornell University in May found that, among more than 95,000 students across 20 U.S. public research universities, 37 percent reported using generative AI at least monthly, while 9 percent reported using it to cheat.
"Assessment reform is necessary and urgent," study co-author Rene Kizilcec, an associate professor at Cornell, said. He argued that misuse of generative AI creates a problem for the validity of assessments and, in turn, "the credibility of university credentials."
How Is AI Impacting Mathematics?
Borcherds has said the rapid development of AI has changed his view of the future of mathematics as a profession. Five years ago, he believed mathematics was among the fields relatively protected from advances in AI. Now, he has said he doesn’t know what advice to give students considering a career in the subject.
"If AI continues to improve at the present rate," Borcherds told The Inc., "Then, in a few years, there'll be nothing human mathematicians can do that AI can't do better."
His concerns come as AI systems become increasingly capable of tackling the kinds of assignments traditionally used to teach and assess mathematics.
Related Story
Newsweek Webinar: How the Politics of AI Could Shape Midterms
A 2026 report from MIT's Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training found that AI was forcing the university to reconsider fundamental aspects of education, including how students are assessed. The committee called on instructors to change current teaching and assessment processes for an "AI-aware world," while emphasizing that AI should supplement human learning rather than automate it.
The issue is particularly relevant to mathematics and other technical subjects because AI can help students solve complex problems, making it harder for instructors to assess their real understanding and reasoning.
The challenge for educators, therefore, is not simply whether students should be allowed to use AI, but how to make sure they continue developing the underlying mathematical reasoning and problem-solving skills that traditional coursework is intended to teach.
Teachers Turn to Oral and In-Person Assessments
...
Some educators are responding not by abandoning assessments, but by changing their format.
Klopfer said educators should first consider their learning objectives and then determine which assessment method best establishes whether students have met them.
That could mean supervised writing sessions where students work without AI, oral presentations or exams, or semester-long projects with students required to demonstrate their progress at different milestones.
"What I caution against is designing assessments solely to protect against AI," Klopfer said. "Blue book exams might test memory, but not thought. In subjects where thinking is the goal, they don’t validly assess what matters."
Cornell University's Center for Teaching Innovation says oral assessments can be paired with written assignments, projects or problem sets to "discourage unauthorized or uncritical use of AI." Options include oral exams, question-and-answer sessions about a student's paper or project, presentations, role-plays, debates, and demonstrations.
At the University of Pennsylvania, Associate Professor Emily Hammer has paired oral examinations with written papers in her seminar classes. She has argued that the concern goes beyond preventing cheating, saying: "We're doing this because students are actually losing skills, losing cognitive capacity and creativity."
Penn has also highlighted the experience of mathematics professor Phil Gressman, who replaced a final written exam with individual oral exams in 2020 and has continued using the format. Gressman said the conversations allowed him to determine whether students genuinely understood concepts by asking follow-up questions, telling the university: "I felt like I got a better sense of what's going on."
Universities Rethink Assessment in the AI Era
The issue is prompting changes at an institutional level, too.
MIT President Sally Kornbluth said generative AI's opportunities and risks represented a "watershed moment" for MIT and higher education more broadly. She said practical changes would include "reevaluating how we assess what students have learned" and placing renewed emphasis on hands-on learning.
MIT has also recommended that every class clearly state its policy on generative AI, including what uses are permitted and why. The university has stopped short of imposing a single institution-wide classroom rule, instead providing instructors with a menu of approaches that can be adapted to individual courses.
Klopfer said the answer should not be viewed simply as finding a balance between using AI and protecting learning. Instead, he argues that the two can be complementary when students understand how the technology affects their learning and know when to use it.
"Students need experience to understand and regulate their AI behaviors," Klopfer said. "They need to learn how it can augment, rather than replace, their learning, and decide how to use it."
He also warned of a potential cycle in which students turn to AI when they are short on time, become less confident in their own work as a result and then increasingly rely on the technology.
"We need to give them ways out of that spiral," he said.
Together, the approaches show how the arrival of increasingly sophisticated AI is shifting the question for educators from simply how to stop students using AI to how to design assessments that still demonstrate what students themselves know and can do.
Newsweek contacted Cornell University for comment.
Newsweek’s reporters and editors used Martyn, our AI assistant, to produce this story. Learn more about Martyn here. Contact Newsweek editors for this story: Robert Greenall, John Fitzpatrick, Gray R. Thomas.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines