20260109 KI produziert Mehrarbeit

Fake research

Let's call "fake research" the research that spreads fake news instead of conducting actual research in science. The linked article has investigated why more and more scientific articles no longer meet the standards for this.

The tip of the iceberg

Scientific publications are not posted online with a "fake traffic light". This means that only a few readers immediately notice the fake nature of the articles. This is what happened to Dr. Liudmila Zavolokina, assistant professor of information systems and digital innovation at the University of Lausanne, when she used Google Scholar to read some citations to four scientific studies allegedly published under her name. However, all of the research papers mentioned turned out to be non-existent – which she knew, of course, but every other reader did not.

For the scientific community as a whole, this means that, in theory, no search result can be trusted without a great deal of effort to find the sources and study them carefully for errors. The supposed relief provided by AI programs is increasingly leading to more work.

Who invents fake research?

Professor Pablo Gómez from Western Michigan University is quoted by Telepolis.de as saying: In many cases, it's pure laziness. Some researchers don't want to spend time conducting a state-of-the-art search, so they ask an LLM tool to do it for them, and it hallucinates some cool-sounding technical articles.

The second group of "fake researchers" also has equally base motives. These are people who publish their (often pseudoscientific) articles in "predatory journals". These are journals that pretend to be scientific without actually being so. An AI naturally makes mistakes when assessing the scientific merit of a found article.

And the third group is AI itself – it hallucinates. AI invents additional studies that sound similar to real studies and author groups in order to substantiate its findings. According to the New York Times, the number of fake scientific articles is currently doubling every 18 months. Now, truly scientific publishers are beginning to “de-publish.” They are retracting publications in which they have found strange-looking wording or illustrations, AI-falsified citations and references, and nonsensical translations during targeted checks.

Hits: 37