Up to 30% of Published Neuroscience Papers May Be Faked – And That's Before ChatGPT

The pressure to publish or perish has led some desperate researchers to pay for fake papers to pad their resumés.

Worse still, some of these sham papers are getting published in official scientific journals.

A computer program designed to detect these made-up studies suggests far too many are slipping past peer review.

The study was published as a preprint paper and still awaiting peer review itself, but if the results are confirmed, it's seriously concerning.

Using artificial intelligence, researchers trained a computer to look for several red flags commonly seen in fake papers submitted to scientific journals.

When the tool could pick out red flags with 90 percent accuracy, it was used to comb through roughly 5,000 neuroscience and medical papers published in 2020.

The tool marked 28 percent as probably made-up or plagiarized.

If this applies to all 1.3 million biomedical papers published in 2020, more than 300,000 would have been flagged.

Not all these flags are truly fakes, but they help identify the most suspicious studies that should receive extra scrutiny by reviewers.

For every 100 red-flagged papers the new tool identified, about 63 were actually fake, and 37 were authentic.

Neuropsychologist Bernhard Sabel from Otto-von-Guericke University of Magdeburg in Germany is one of the authors behind the study and an editor of a neurology journal.

He, like many others, has been dealing with a recent rise in fake papers. But even Sabel was shocked by his tool's initial numbers.

"It is just too hard to believe," he told Science.

Sabel and his colleagues blame 'paper mills' for the fraudulent activity. Paper mills bill themselves as 'academic support' services, but in reality, they use AI to scale and sell fake publications to researchers.

Prices for fake papers can range from US$1,000 up to US$25,000.

The quality of these studies is often poor but just good enough to pass peer review, even in established journals.

Publishers are aware that this is a serious issue that undermines their reputation. Scientists have even tricked publications into accepting laughably fake papers to bring attention to the problem.

Sometimes, paper mills will go so far as to pay publishers to accept their fake studies. In fact, an unsolicited email of this nature to the editor of a journal prompted the new study.

"Because the problem is still perceived to be small (an estimated 1 of 10,000 publications), publishers and learned societies are just beginning to adjust editorial, peer-review, and publishing procedures," researchers write.

"Yet the actual scale of fake publishing remains unknown, despite the fact that the number of reports on paper mills are increasing."

Between 2010 and 2020, the new tool revealed a 12 percentage point increase in the rate of potential fake papers published by some journals.

The nation with the highest number of potential fakes was China, contributing to just over half the red flags. Russia, Turkey, Egypt, and India were also significant contributors.

"Fake science publishing is possibly the biggest science scam of all times, wasting financial resources, slowing down medical progress, and possibly endangering lives," researchers argue.

And the rise of generative AI such as ChatGPT only makes the scam more of a threat.

To counter this emerging technology and uphold the reputation of science itself, researchers say a more rigorous review system is urgently needed.

The preprint was published in medRxiv.

Source