A team led by computer scientists from the University of Colorado Boulder, USA, has developed an artificial intelligence (AI) platform that seeks to automatically identify scientific journals that use questionable practices, such as publishing an unusually high number of articles, offering very short publication times, exhibiting high rates of self-citation, and approaching authors aggressively. Those responsible for these titles, also known as “predatory journals,” try to persuade scientists to pay a fee to publish their work quickly, often without a genuine peer-review process. “They will say, ‘If you pay US$500 or US$1,000, we will review your paper,’” Daniel Acuña, a researcher from the university’s computer science department, told the website ScienceDaily. “In reality, they do not provide any service. They just take the PDF and post it on their website.”
Acuña’s group trained an AI system with the editorial standards adopted by the Directory of Open Access Journals (DOAJ), a nonprofit organization that produces a list of reputable open-access journals. The DOAJ curators focus on good practices such as transparent reporting of the journal’s objectives, scope, peer-review policy, and the composition of its editorial board (including the institutional affiliations of all members).
Testing of the AI platform, the results of which were published in Science Advances in August, found that of almost 15,200 journals analyzed, 1,400 failed to meet at least part of the DOAJ criteria and were flagged as potentially problematic. Human experts reviewed the findings and found that the AI made mistakes—about 350 journals identified as questionable were likely legitimate. Acuña sees the tool as a promising resource to protect researchers looking for a journal to publish their work, describing it as a “firewall for science”—a reference to network security systems that prevent hacker attacks. “The system can be used to prescreen large numbers of journals,” he said, “but human professionals should do the final analysis.” The beta version is available online, with access currently limited to publishers and databases that index journals.
Republish