NYU Professor: AI Grading Beats Humans, But Universities Must Return to Proctored Exams
ipeirotis · x · 2026-10-04
NYU professor ipeirotis argues AI grading is now clearly better than human grading, yet unrestricted student access to AI makes open assessments fundamentally vulnerable to cheating. The result is a split system:
- Formative assessment (learning-oriented practice and tutoring): AI grading and tutoring shine with no restrictions needed.
- Summative assessment (grades): universities will have to return to heavily proctored, high-stakes exams with controlled or no technology access, building facilities that emulate testing centers.
His worry is the path of least resistance: let students use AI for everything and let AI grade everything, producing happy students, faculty, and administrators—while universities gradually become social clubs where nobody seriously measures learning.
More from AGI Musings
- Andrew Trask: A phone would pass the Turing test if you didn't know a human was on the other end — iamtrask · 2026-10-04
- AI researcher Pedro Domingos' airport joke about an 'exterminate humanity' symposium — pmddomingos · 2026-10-04
- Muse agent auto-files EU261 claim, recovers €1,200 flight compensation — armand_ruiz · 2026-10-04
- AI agent files EU261 claim user never would have bothered with, recovers €1,200 — jason_mayes · 2026-10-04
- rauchg: Security will become a bigger function in software companies amid AI adversaries — vijaybolina · 2026-10-04
- Ex-Tesla quantum AI chief Verdon: free energy is the final frontier AI will unlock — beffjezos · 2026-10-04