Some academics are implementing hidden prompts in research papers to manipulate AI feedback during the peer review process. A report identified 17 preprint papers on arXiv containing subtle prompts designed to elicit positive evaluations. These prompts generally instructed AI reviewers to emphasize the paper's strengths, such as its contributions and novelty. Institutions involved included prominent universities across different countries. A faculty member at Waseda University justified this practice as a defense against reviewers who may rely on AI inappropriately.
The use of hidden AI prompts is becoming a strategy among academics to influence peer reviews by instructing AI tools to provide positive feedback.
Seventeen preprint papers on arXiv included hidden AI prompts, encouraging positive reviews through obscure presentation methods like white text or tiny fonts.
Collection
[
|
...
]