Michigan study finds no post-ChatGPT grade inflation across 156,135 students
i_dg23 · x · 2026-07-26
A University of Michigan study finds no evidence that ChatGPT caused artificial grade inflation across a large university, even in courses that should be more vulnerable to AI-assisted cheating.
The researchers analyzed records from:
- 156,135 students
- 87,936 course offerings
- Academic years 2015–2025
They tested the “GenAI substitution hypothesis”: if students are outsourcing assignments and take-home work to AI, then courses with more AI-susceptible assessments should show unusually large grade increases after ChatGPT’s release.
What they found:
- No significant change in grades overall after ChatGPT.
- No meaningful grade inflation even in courses with more take-home exams, essays, or homework.
- No drop in failure or withdrawal rates.
- The same holds for students who previously performed worse.
- Self-reported understanding and interest were also largely unchanged.
The authors say this does not mean students aren’t using AI for schoolwork; it only means the feared macro-level grade inflation signal did not appear in the data.
Related event: Study: ChatGPT Shows No Measurable Impact on Student Grades(2 posts)→
More from Infra
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11
- Can a 7900 XTX 24GB run Qwen locally? Reddit seeks ROCm tok/s benchmarks — thenomadexplorerlife · 2026-09-11