Hundreds of thousands of research articles may be false

Research fraud

Article mills pump out fake research articles. According to the new analysis, the percentage of what is published out of control is increasing.

Not everything written in scientific journals is gold.

Every year, large quantities of fake research articles are pumped out of so-called “paper mills,” or “article mills,” which produce and sell literature to researchers seeking publications.

The problem is known, but no one knows exactly how widespread it is and how many fake articles sneak past quality control and are published.

Now, an unpublished analysis suggests It has been discussed in nature The research literature may be infected with hundreds of thousands of false articles. The amount varies from topic to topic.

Strong increase

The analysis was performed using machine learning software developed by Adam Day, director of data services company Clear Skies.

Dai reportedly used this software to analyze the titles and abstracts of more than 48 million articles published since 2000, flagging articles that bore a clear similarity to known fake articles from article mills. Nature wrote that the two retracted articles and the questionable articles were covered.

According to the analysis, more than 400,000 research articles published in the past two decades have strong textual similarities to known fake articles. Figures in the journal Nature show that the proportion has increased sharply. In 2022 alone, there were about 70,000 fake articles.

It is estimated that between 1.5 and 2 percent of articles published last year were very similar to such articles. In biology and medicine, it is estimated at a full three percent.

From 2 to 46 percent

Adam Day assures Nature that it is not possible to know whether all of these articles are actually fake. Analyzes are needed for each individual substance to check whether it comes from substance factories. But when it comes to the proportion of fake articles out of the total, he thinks that’s a reasonably conservative estimate.

Dutch microbiologist Elisabeth Beck, who was summoned “The Sherlock Holmes of Scientific Fraud.” Because of his work exposing research fraud, he told Nature that although the estimate from Day’s analysis is “astonishingly high,” it is “not far-fetched.”

Jennifer Byrne, an Australian cancer researcher also known for exposing research fraud, says she “unfortunately believes these estimates are reasonable.”

Nature points to one Report on “Paper Mills” from the Committee on Publication Ethics (COPE), published in June 2022, which was also previously featured in Khrono. It states that one of the biggest problems in combating fake articles is knowing how widespread they are.

After analyzing data from more than 53,000 articles, they concluded that the percentage of suspicious articles submitted to journals ranged from 2 to 46 percent. While most journals have about 2% of suspicious articles among their submissions, the percentage is higher for journals in which paper mills have successfully published articles.

Fake peer reviews

Something is discovered after it is published and withdrawn by the publisher. But most of them live on in the research literature.

Nature cites Retraction Watch, which has created a database that includes an overview of about 44,000 retracted articles. Fewer than 3,000 of them are linked to article mills, Ivan Oransky, co-founder of Retraction Watch, tells the magazine.

But he confirms that this number is too low. Firstly, they add more and more articles, secondly, publishers often avoid the word “paper mill”.

The latter was the case in a case reported by Chrono in 2022. The major publishing house Springer Nature withdrew an article with a research fellow at the University of Western Norway among the authors. In their justification, they point to several things, including compromised peer review. When Khrono contacted the publisher to find out why this was the case, it turned out that they suspected an article factory might be behind it.

“We can’t say for sure that this is a product from a paper mill, but given the issues we identified, it seems likely,” Chris Graf, director of research ethics at Springer Nature, told Chrono.

As Khronow recently wrote, there has been a significant increase in the number of articles being retracted due to manipulated peer reviews. Figures from Retraction Watch show that of the 5,447 articles retracted in 2022, a total of 2,748 were retracted due to faulty or questionable peer review.

– There are more and more of them, and they are responsible for a large percentage of the increase in withdrawals, Ivan Oransky of Retraction Watch told Khrono. He added that retractions due to manipulated peer review can often be linked to “paper mills” or material mills.