AI use leads UC Berkeley math professor to see 'no point' in grading homework
A UC Berkeley math professor’s claim that AI has made grading homework pointless sparks a wider re‑thinking of how — and whether — coursework should count toward university grades. Many argue homework should become ungraded practice, with learning assessed through proctored in‑person exams, oral defenses, or frequent quizzes that are harder to outsource to AI. Beneath this lies a deeper question about the purpose of higher education itself: is it still primarily about learning, or has it become a credentialing and social-filtering system that AI now threatens to upend?
Homework vs. Exams in the Age of AI
- Many argue there’s now little point in grading take‑home work because AI and solution banks make cheating trivial and undetectable.
- Others note this isn’t entirely new: even pre‑AI, students shared answers and reused problem sets; AI just removes the friction and deniability.
- A common proposed shift: homework for practice only, with grades based almost entirely on proctored in‑person exams.
Role and Value of Homework
- Supporters of ungraded homework say it is essential for practice, early diagnosis of misunderstandings, and building discipline.
- Critics say graded homework mostly measures conscientiousness and gaming skills, inflates grades, and was always a weak signal of mastery.
- Several report systems where homework counted minimally or not at all, and exams alone determined the grade, often successfully.
Alternative Assessment Models
- Proposals include:
- Frequent in‑class written or oral quizzes, whiteboard problems, and vivas.
- Final exams that dominate the grade, with make‑ups or re‑takes to reduce “bad day” luck.
- Flipped classrooms, project‑based work, or portfolio evaluation.
- No-grades or pass/fail models emphasizing feedback over ranking.
- Some advocate random oral questioning on homework to verify understanding.
Cheating, Motivation, and Student Responsibility
- One camp: adults in college should self‑motivate; if they skip homework and fail the exam, that’s on them.
- Another camp: many capable but underprepared students need structured incentives and early feedback; otherwise they crash at the final and disengage.
- Views differ on whether reliance on “sticks” means a student is “not college material” or just mismatched with current teaching practices.
AI as Threat and Tool
- Concerns: AI undermines traditional assessment, could devalue degrees, and may eventually reduce demand for higher education.
- Optimists see a “golden age” for autodidacts: AI can tutor, grade proofs, and provide detailed feedback at scale.
- There is unease that students increasingly prefer AI over human office hours and study groups, with mixed learning outcomes.
Broader Higher‑Ed and Policy Themes
- Debate over whether universities should primarily teach job skills, do research, credential social elites, or foster intrinsic learning.
- Some foresee consolidation or collapse of weaker institutions and more vocational or apprenticeship-style models.
- Long side‑threads argue about state vs. private provision, neoliberal policy, and whether public systems are inherently inefficient or not—opinions sharply diverge.