Overview
- Richard Borcherds, a Fields Medalist and UC Berkeley professor, this week said he will no longer grade homework for correctness and will award credit based on completion.
- He says students routinely use generative AI on assignments, so he now bases most course grades on supervised, in-person exams where electronic devices are banned and calls unmonitored online tests a waste of time.
- UC Berkeley instructs faculty to state course expectations on AI use but leaves specific rules to individual instructors after campus reviews found detection tools such as Turnitin and Rumi unreliable.
- Across California campuses some instructors are adopting similar tactics — on-paper, proctored exams, completion grading for take-home work, sanctioned AI tools or custom chatbots — producing a patchwork of local approaches.
- Faculty warn the shift raises deeper questions about pedagogy and careers in expert fields, and could push colleges to redesign assessments, admissions signals, and how employers judge student skills.