What's this about?
A large study of college records finds that once COVID's disruption is modeled separately, ChatGPT's arrival had no detectable effect on grades — and course evaluations for subject understanding, interest, and workload didn't move either. Neither the mass-cheating panic nor the learning-revolution hype shows up in the data.
What exactly does "modeled separately" mean?
ChatGPT launched while colleges were still snapping back from pandemic-era grading and remote learning, so raw before/after comparisons mix two shocks together. The authors statistically separate the COVID rebound from the ChatGPT introduction — and once you do, the ChatGPT line goes flat.
Why does it matter?
It's rare, careful evidence in a debate dominated by anecdotes: either students' AI use substitutes for learning in ways grades don't capture, or higher education absorbed the technology with far less change than either camp claims — both readings are uncomfortable for someone.