"Man-made brainpower" instead of "artificial intelligence", "peril of the breast" for "breast cancer" or "nucleic corrosive" for "nucleic acid".
So many “tortured expressions” found in pseudo-scientific studies that smack of copy-paraphrase-paste.
It is these that Guillaume Cabanac hunts, in scientific publications using the software he designed, Problematic Paper Screener.
Work carried out in collaboration with the Grenoble researcher Cyril Labbé and the Russian computer scientist Alexander Magazinov which earned this lecturer at the University of Toulouse III - Paul Sabatier and researcher at the Toulouse Computer Research Institute (IRIT) d '' have been selected by the scientific reference journal "Nature" among the 10 scientists who will have marked the year 2021.
"I programmed an informant who gives me pipes"
Indeed, the publication of bogus scientific studies is a growing phenomenon, even in prestigious journals edited by Springer Nature or the Institute of Physics, which casts doubt on peer review.
The result, according to Guillaume Cabanac, of the misuse of scientific research by some, in an increasingly competitive and commercial world context.
“The safeguards have eroded.
Initially, we observe predatory journals, which publish articles, for a fee, without any evaluation being carried out.
In addition, companies offer for sale pre-written articles on specific topics.
"
To read also "I'm really sorry": the mea culpa of the coordinator of the study of "The Lancet" on hydroxychloroquine
And it is precisely to remedy this phenomenon and not to put dust under the carpet that the Irit researcher has developed this software which, every night, reviews the 120 million articles published and identifies these famous " tortured expressions ”.
In total, a little more than 3000 articles have already been reported, of which a third date from 2021. “In a way, summarizes Guillaume Cabanac, I programmed an indicator, who gives me tips.
And me, with my scientific training, I follow the trail like a detective and I check if the article is really problematic.
By snowballing, scientists around the world are also reporting problem articles.
"
"People have confidence in the scientific process"
Once these publications have been reported, they are shared on a PubPeer post-publication evaluation site, designed by two Parisian researchers, so that scientists who are experts in their field can come in turn to analyze these articles.
The Dutch Elisabeth Bik, specialized in this detection work, is one of these contributors.
And the Toulouse researcher invites the scientific community to take up this tool and this problem.
“This does not call scientific research into question, on the contrary.
This proves that we care about it and I really believe that people have confidence in the scientific process, ”concludes Guillaume Cabanac.