55,000 studies have already been retracted, and the number of unreported cases is likely much higher.
During a six-month investigation, two computer scientists and a former health reporter uncovered thousands of fake scientific studies written by AI, particularly in the areas of cancer and Covid-19 research.
In science, more publications usually lead to more funding, so there is an unspoken incentive to publish whatever the journal editors let pass (also known as “publish or perish”).
How these papers fare in the required peer review remains to be seen. Harry Sockel writes on Medium that it’s a system of personal relationships and favoritism. You write positive reviews for each other. In addition, about 17 percent of peer reviews are written by Chat GPT.
One of the engineers who co-authored the report developed the Problematic Paper Screener , a tool that scans millions of new scientific papers for evidence of fraud. (Read about it here).
About 119,000 scientific journal articles and conference proceedings are published worldwide every week, more than 6 million a year. Publishers estimate that about 2% of articles submitted to most journals—but not necessarily published—are likely to be fakes. That would be around 120,000 forgeries a year, and the trend is rising sharply.
There are some misleading phrasings, such as “raw information” instead of “raw data.” But a fraudster who reads his fake work before submitting it will eliminate such clumsy writing.
Image: Pixabay.com
