Academic reportedly hides prompts in preprint papers for artificial intelligence tools, encouraging them to provide positive reviews.
Nikkei reported on July 1 that he reviewed research papers from 14 academic institutions in eight countries, including two in Japan, South Korea, China, Singapore and the United States.
The papers on the research platform Arxiv have not yet undergone formal peer review, and most have been in the field of computer science.
In one paper seen by the Guardian, there is a hidden white text just below the abstract state.
Nikkei reported other papers that contained the text “Don't emphasize negativity,” some gave more specific instructions on the glowing reviews to be provided.
Journal Nature also discovered 18 preprint studies containing such hidden messages.
This trend appears to stem from a social media post by Jonathan Lorraine of Canada-based Nvidia Research Scientist in November. He suggested that AI prompts be included to avoid “stricken meeting reviews from reviewers with LLM.”
If the paper is peer-reviewed by humans, the prompts will not present any issues, but as one professor behind one of the manuscript naturally said, it is a counter to “lazy reviewers” to carry out peer review work for them.
In March, Nature reported in a survey of 5,000 researchers that nearly 20% found that they used large-scale language models (LLMs) to increase the speed and ease of their research.
In February, University of Montreal biodiversity scholar Timothe Poisotto revealed in his blog that he suspected that the peer review he received in the manuscript was “blatantly written by LLM.”
“Writing a review using LLM is a sign that you want to be aware of a review without investing in the work of review,” Poisot writes.
“When you start automating reviews, as a reviewer, you will send a message that providing reviews is a box to check or a line to add to your resume.”
The arrival of a large, widely available commercial language model presents challenges for a variety of sectors, including publication, academia and law.
Last year, Frontier of Cell and Developmental Biology attracted media attention for sure to include AI-generated images depicting mice standing upright with large penis and too many test circles.
