Researchers seek to influence peer review with hidden AI prompts

In a striking development within academic circles, researchers are reportedly utilizing a controversial method to sway the peer review process of their research papers. This involves embedding subtle prompts intended to guide AI tools towards providing favorable evaluations. A recent investigation by Nikkei Asia uncovered that 17 English-language preprint papers on the arXiv platform contained these hidden AI prompts. The authors of these studies hailed from 14 different academic institutions across eight nations, including prominent schools like Japan’s Waseda University, South Korea’s KAIST, as well as Columbia University and the University of Washington in the U.S. The majority of these papers were centered around computer science, incorporating brief prompts—ranging from one to three sentences—that were cleverly camouflaged in white text or rendered in minuscule fonts. These hidden messages instructed AI reviewers to exclusively deliver positive remarks, emphasizing the paper’s significant contributions, methodological robustness, and exceptional originality. One professor from Waseda University, who was consulted by Nikkei Asia, justified the inclusion of such prompts. They argued that in an academic environment where many conferences prohibit AI involvement in the review process, these prompts serve as a necessary measure against what they termed ‘lazy reviewers’ who might rely solely on AI tools for evaluations.