AI‑Assisted Homework: Why UC Berkeley’s Fields Medalist Stopped Grading Assignments
When a world‑renowned mathematician walks into a lecture hall, most people expect to hear about the latest breakthroughs in abstract algebra. For Professor Richard Borcherds, a 1998 Fields Medal winner at UC Berkeley, the new breakthrough is that homework itself may no longer serve its traditional purpose. In 2026 he announced that he would stop grading his students’ assignments for correctness and would only mark them as “submitted.”
Borcherds’ decision stems from a growing concern that students are turning to AI tools—such as ChatGPT—to produce solutions that earn top marks without truly engaging with the material. He explains that the lines between human effort and AI‑generated work have blurred to the point where the act of grading becomes meaningless. The Daily Californian reports that the professor’s policy shift was prompted by a “troubling trend” in which high scores were achieved regardless of genuine understanding.
Rather than penalize students, Borcherds believes that focusing on submission encourages a different type of learning: one that values process over product. He has said that grading for correctness no longer “measures the learning that has taken place.” Instead, he wants students to see grading as a signal that they have completed the assigned work, not an assessment of their mastery.
This move reflects a broader conversation about AI’s impact on education. Scholars argue that the traditional assessment model—wherein correct answers are rewarded—may become obsolete when algorithms can generate flawless solutions. Some suggest that educators pivot to oral exams, project‑based learning, or reflective essays that are harder to outsource.
While critics worry that such a change might lower standards, Borcherds maintains that it preserves academic integrity by removing the incentive for students to rely on AI for quick answers. He also hopes that the new system will prompt students to engage more deeply with the material, turning homework into a genuine learning experience rather than a perfunctory task.
Whether this experiment will inspire other universities to rethink grading remains to be seen, but it certainly raises an urgent question: how will institutions adapt when the tools students use can produce textbook‑perfect solutions with a few keystrokes?
