Research: AI finds more than 1000 "questionable" science magazines
Anyone who does scientific work also wants to publish the results. Fraudsters take advantage of this. Thanks to AI, fraud can be detected more quickly.
(Image: PowerUp/Shutterstock.com)
With the help of AI technology, a group of computer scientists has discovered more than 1,000 alleged scientific journals that are of at least "questionable" quality. The University of Colorado Boulder has now made this public and explains that this is a considerable number considering the more than 15,000 titles examined. At the same time, however, the research team limits the hit rate of the AI tool, which initially classified more than 1,400 as potentially problematic. After a manual review, over 1000 of these were left over and classified as "dubious". Against the backdrop of increasing attacks on science, the increase in such products must be stopped, the team explains its own work.
Posting online for an expensive fee
As study leader Daniel Acuña reminds us, it is common practice for researchers to submit papers to renowned scientific journals, which are then independently reviewed by experts. This peer review is intended to ensure that the content of the work meets the quality standards of the magazine and the scientific community as a whole. Recently, however, more and more magazines have emerged that often approach researchers outside the USA and Europe directly and promise to publish their work in return for payment. Under pressure to publish as much as possible, they may accept. However, the promised review does not even take place and instead PDF files are simply posted online.
Videos by heise
To take action against these "pirated journals", Acuña's team has now trained an AI based on reviewed specialist journals and had it check a total of almost 15,200 titles. The AI searched for indications of the existence of an editorial team, spelling mistakes on the homepage, titles with a particularly large number of publications or authors who cite themselves particularly frequently. The particularly questionable ones were then marked and checked by hand. Acuña believes that the final decision must be made by humans. His team wants to release the tool for research, but it is not yet available. The work is presented in the renowned scientific journal Science Advances.
(mho)
