The pressure to publish scientific articles suffered by many researchers is not only causing fraud at the editorial level and complaints in the academic area, it is also causing changes, for the worse, in the published content. When neuropsychologist Bernhard Sabel put his new fake study detector to work, he was “shocked” by what he found. After reviewing some 5,000 articles, he estimates that up to 34% of neuroscience articles published in 2020 were likely fabricated or plagiarized; in medicine, the figure was 24%. Both numbers, as seen reported in a medRxiv preprintare well above the levels they calculated for 2010, and well above the 2% baseline estimated in a report from the group of editors of 2022.
“It’s incredible,” explains Sabel, from the Otto von Guericke University Magdeburg and editor-in-chief of Restorative Neurology and Neuroscience. It’s like someone told you that 30% of what you eat is toxic”.
Their findings underscore what has been widely suspected: Journals are inundated by a growing wave of scientific manuscripts from paper mills, secret businesses that allow researchers to pad their publication records by paying for fake articles or undue authorship. “The paper mills have made a fortune basically attacking a system that had no idea how to deal with these things,” says Dorothy Bishop, a psychologist at the University of Oxford who studies fraudulent publishing practices.
Sabel’s tool is based on just two indicators: authors using private, non-institutional email addresses and those listing an affiliation with a hospital. Not a perfect solution due to a high false positive rate. Other developers of counterfeit paper detectors, who often don’t disclose how their tools work, face similar problems.
Still, the detectors raise hopes of gaining an advantage over paper mills, which produce fake manuscripts containing partially or fully plagiarized or fabricated text, data and images, often doctored by ghostwriters. Some items have backed by lax reviewers requested by the authors. Such manuscripts threaten to corrupt the scientific literature, mislead readers, and potentially distort systematic reviews. The recent arrival of artificial intelligence tools like ChatGPT makes everything even more complex.
To fight back, the International Association of Scientific, Technical and Medical (STM) Publishers, which represents 120 publishers, is leading an effort called IntegrityHub to develop new tools. STM is not revealing much about detection methods, for avoid alerting these paper mills.
Examining suspicious documents can be time consuming: in 2021, the Springer Nature publication review of about 3,000 documents suspected of coming from paper mills it required up to 10 full-time and part-time employees, said Chris Graf, the company’s director of investigative integrity.
The pressure to “publish or perish” that institutions exert on scientists is also an obstacle. “We want to think about engaging with institutions on how to perhaps remove some of the professional incentives that can have these detrimental effects,” adds Sabel. Such pressures may push physicians with no research experience to turn to paper mills, which is why hospital affiliations may be a red flag.”
After a report 2020 that named magazines suspected of containing paper mill articles, an analysis using the Papermill Alarm automated detector found that the number of such articles in one of those magazines (which the analysis did not name) dropped rapidly and dramatically. The problem is that right now, with artificial intelligence tools, the number it could have increased even more than we thought.