How AI is disrupting scientific publications.
AI’s Growing Influence on Research Papers Raises Concerns
Initially, experts were amused. Then, they became genuinely worried. In August 2023, Guillaume Cabanac, a professor at the University of Toulouse-III and a keen detector of dishonesty in research articles, stumbled upon an oddity. In the middle of a physics article, he read a strange phrase, “Regenerate response,” which he immediately recognized. This text was, in fact, a copy-paste from the button on a website that everyone had been talking about for months: ChatGPT. He had just identified the first evidence of the famous chatbot being used to write research articles in seconds. The text, published by an independent and reputable publishing house, the Institute of Physics, was retracted in September 2023.
A few months later, an image of a rat with a giant penis, illustrating a biology publication and betraying the use of fanciful artificial intelligence (AI) image generation, also caused a stir. Retraction, again. “Upon discovering other examples, my first reaction was amusement. But the more serious consequences quickly became apparent to me. If this kind of content can pass the peer review filter, then that process is not doing its job,” says Alex Glynn, a documentalist at the University of Louisville (Kentucky). Since the arrival of generative AI, he has embarked on a collection of suspicious cases, looking for typical AI expressions, such as “as of my last update.” He has already identified more than 500, some of which involve major publishing houses – Elsevier, Nature Springer, Wiley…
In response, these publishers have issued guidelines for good practice: authors are not necessarily prohibited from using these tools, but they must specify it. Two publishers contacted for comment, Elsevier and Nature Springer, want to reassure. “We believe that AI can be beneficial to researchers,” says a Nature Springer spokesperson. “We see AI as a powerful tool that, when integrated responsibly, strengthens research integrity and accelerates innovation,” adds an Elsevier spokesperson. Both also specify that these uses must be done “ethically”, “with human supervision.” Like other publishers, they also have artificial intelligence tools to detect other AIs (images, plagiarism, etc.).
Enjoyed this post by Thibault Helle? Subscribe for more insights and updates straight from the source.
