Is ChatGPT Killing Peer Reviews Research?
You know that ChatGPT likes certain words? Apparently, it has a thing for adjectives like "commendable", "meticulous", and "intricate".
New research from Stanford University has uncovered, this after ChatGPT hit the scene in November 2022, there was a sudden 9.8x increase in reviews calling papers "commendable", a whopping 34.7x jump in "meticulous" assessments, and papers were deemed 11.2x more "intricate" than before.
Over all 10.6% of ICLR 2024 reviews and 16.9% of EMNLP 2023 reviews are likely AI-generated.
Commendable? I think not!
But wait, there's more! The study found some juicy correlations:
- More AI-written reviews submitted close to deadlines.
- AI-generated reviews are less likely to cite references.
- Reviewers using AI reply less to author rebuttals.
- AI reviews are more generic & similar to each other.
If you thought that academic abuse of ChatGPT was just a problem for school teachers, then think again.
A link to the study - https://arxiv.org/pdf/2403.07183.pdf