Chapters of Concern
Authors/Creators
Description
I detected reference-integrity issues in 384 book chapters/conference proceedings with the reference-checking tool Slopcheck and reported the issues on PubPeer.
With the help of the AI-text detector Pangram, run on the abstracts of these chapters and on 500 randomly sampled control abstracts, the probability of LLM involvement is estimated here. 68.3% of chapter abstracts with reference-integrity issues were flagged as AI-generated (Pangram score > 0.5), versus 1.3% of controls where the reference issues were not detected.