The End of the Take-Home Era: How AI is Forcing a Radical Shift in Higher Education

In a striking development that signals a growing crisis of academic integrity in the age of generative artificial intelligence, a prominent mathematics professor at the University of California, Berkeley, has officially abandoned the practice of assigning graded homework. Richard Borcherds, a Fields Medalist and renowned mathematician, has concluded that the proliferation of sophisticated AI tools has rendered traditional out-of-class assessments functionally obsolete, effectively ending an era of pedagogical reliance on independent, off-campus work.

This decision marks a pivotal moment in the ongoing battle between educators and emerging technology. As AI models become increasingly capable of solving complex multivariable calculus problems and articulating nuanced proofs in seconds, the foundational premise of homework—that it serves as a reliable measure of student comprehension—has collapsed. Borcherds’ move is not merely a logistical change; it is an admission that the digital landscape has permanently altered the nature of scholastic evaluation.

The Chronology of a Classroom Crisis

The path to this decision was paved by a steady, observable decline in the efficacy of remote assessment. Over the past several academic terms, Borcherds noted a suspicious trajectory in student performance data. While the material remained as challenging as ever, the discrepancy between online exam scores and in-person performance began to widen to an unsustainable degree.

The shift began in earnest with the public rollout of advanced Large Language Models (LLMs). Initially, educators viewed these tools as potential academic aids, but the reality quickly shifted toward mass misuse. Borcherds, observing the trends in his own coursework, noticed that while homework completion rates remained high, the actual depth of student participation in conceptual discussion during office hours began to stagnate.

By the fall of 2026, the situation reached a breaking point. Borcherds reported that it had become nearly impossible to distinguish between the work of a dedicated student and the output of a sophisticated AI. The "homework-as-learning" model, which relies on the assumption of student honesty, was systematically undermined by the sheer convenience of automated solutions. Consequently, the professor made the decisive call to shift his grading architecture entirely toward high-stakes, in-person, proctored environments where electronic devices are strictly prohibited.

Supporting Data: The "Average Score" Anomaly

The evidence cited by Borcherds is rooted in quantitative patterns that are currently plaguing mathematics and computer science departments across the United States. According to data and testimonials provided to the New York Post, the professor identified a recurring phenomenon: a significant, inexplicable spike in average scores on online assignments that did not correlate with performance during proctored testing.

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

The data suggests a twofold problem:

1. The Inflation Effect: When students utilize AI for homework, the aggregate class average becomes an unreliable metric for the instructor to gauge the pace of the curriculum. If the professor assumes the class understands a concept based on homework performance, but that performance is artificial, the instructor risks moving too quickly, leaving honest students behind.
2. The "Honesty Tax": Borcherds highlighted a moral hazard within the current system. By assigning grades to homework that can be easily outsourced, the university inadvertently penalizes students who choose to do the work themselves. These students spend hours grappling with a problem, only to have their hard work mathematically diluted by a class average inflated by AI users.

Official Responses and the Academic Community

The academic community is currently divided on how to respond to the "AI-in-the-classroom" dilemma. While some departments are doubling down on AI literacy—teaching students how to use the tools as "co-pilots"—professors like Borcherds represent a traditionalist camp that argues for a return to "analog" verification.

UC Berkeley, like many peer institutions, has yet to issue a blanket policy on the use of AI in grading, leaving the autonomy largely in the hands of individual professors. However, the university administration has expressed ongoing concern regarding the maintenance of degree integrity.

"Grading for correctness just adds a huge incentive for them to use AI for homework instead of doing it themselves," Borcherds noted, emphasizing the futility of current assessment models. "And paying to grade AI-generated answers is a waste of time and money."

Outside of Berkeley, other institutions are experimenting with oral exams (vivas), handwritten journals, and "in-class only" problem sets. However, the scale of public universities often makes these labor-intensive methods difficult to implement, creating a significant tension between the desire for scalable education and the need for accurate student assessment.

The Broader Implications for Higher Education

The decision made by Professor Borcherds carries profound implications for the future of university curricula. If homework no longer serves as an effective learning tool, the entire structure of the "flipped classroom"—where students study at home and practice in class—may be on the brink of extinction.

1. The Resurgence of In-Person Proctoring

We are likely to see a sharp increase in the reliance on blue-book exams and pen-and-paper assessments. This shift represents a rejection of the digital transformation of the last decade, suggesting that for high-level technical subjects, the only way to verify cognitive progress is through physical observation.

2. The Devaluation of "Take-Home" Credentials

If homework and online assessments are increasingly viewed as "AI-contaminated," employers and graduate school admissions committees may begin to devalue grades earned in courses that rely heavily on remote work. This could lead to a two-tiered system where "verified" in-person degrees carry significantly more weight than those earned through hybrid or online-heavy programs.

3. The Pedagogical Pivot

Educators may need to rethink the purpose of homework entirely. If the goal of a homework assignment is no longer to "get the right answer," but rather to document the process of thought, then the way we grade must change. Future assignments may require students to submit a "trail of thought"—a digital or physical log of their errors, their research process, and their revisions—rather than just the final result.

4. Economic Costs

Borcherds’ mention of "wasting time and money" highlights an often-overlooked aspect of this crisis: the administrative burden. Teaching assistants and professors are currently spending thousands of hours per semester grading AI-generated text. As the cost of education continues to rise, the inefficient use of faculty time on checking AI-written assignments represents a significant financial drain on the university system.

Conclusion: A Return to the Analog?

The stance taken by the UC Berkeley mathematics department is a bellwether for the rest of academia. We are witnessing the end of a period of blind faith in remote digital education. As AI continues to evolve, the distinction between "human-authored" and "machine-authored" will become increasingly blurred, potentially rendering the concept of "take-home work" a relic of the pre-AI era.

While technology has provided students with unprecedented access to information, it has also created a barrier to the fundamental cognitive development that comes from struggling with a difficult problem. For Professor Borcherds, the solution is clear: if you cannot ensure the integrity of the process, you must return to the only environment where you can: the classroom, under the watchful eye of the educator.

As other professors weigh their options, they must decide whether to adapt their curriculum to incorporate AI or to build a digital-free sanctuary for critical thinking. One thing remains certain: the days of assigning homework with the assumption of student honesty are effectively over. The challenge for the next decade of higher education will be to find a way to assess human potential in a world where the machine is always ready to answer for us.