Morressier teams up with academic fraud sleuths to bring tortured-phrase detection to publishing workflows

There has been a sharp rise in academic fraud over recent years, driven in part by a publish-or-perish culture that rewards researchers based on oversimplified publication metrics. Fraudulent practices include “salami slicing”, where a single study is published in separate papers in order to boost a researcher’s publication metrics, and the use of “paper mills” that sell junk papers for a fee. There’s a relatively new arrival on this sordid scene: papers consisting entirely of, or padded with, content created using generative AI technology.

In a preprint published in 2021, academics Guillaume Cabanac (University of Toulouse), Cyril Labbé (Université Grenoble Alpes), and software engineer Alexander Magazinov (Yandex) outlined a new technique for identifying papers that may contain or consist of junk AI-generated text, based on identifying ‘tortured phrases’. These are unusual language formulations that appear instead of well-established phrases: for example, “cruel temperature” instead of “mean temperature” or “flag to clamor”’ instead of “signal to noise”.

These three researchers founded the Problematic Paper Screener, a platform that flags questionable articles in the scientific literature. It leverages a daily maintained public database of ‘fingerprint’ phrases that can be used to identify potentially problematic submissions. Morressier now uses this data to check submissions via its workflows, flagging potentially problematic content early for editors.

Guillaume Cabanac: “To address research integrity issues, we need collaboration across the scholarly ecosystem. Everyday I post suspected content on PubPeer, the leading post-publication peer review platform. Now we need to prevent unreliable articles from getting published in the first place. That’s why I’m happy to volunteer my time to work pro bono with infrastructure providers such as Morressier. Our goal is to widen the net by adding pre-publication screeners and ultimately minimise fraudulent behaviours in research.”

Mădălina Pop, leader of Morressier’s Research Integrity Team: “Morressier is leveraging the incredible work that has been done in the research community by Cabanac, Labbé, Magazinov, and many others, who give hope to the idea that science is self-regulating. Our responsibility is to bring this work seamlessly into submission workflows so that any potentially problematic content is identified early and at scale.”

About Morressier

Morressier provides societies and publishers with workflows for transforming scholarly communications. Driven by a mission to accelerate scientific breakthroughs, it is headquartered in Berlin and has offices in London and Boston. Visit morressier.com and follow @Morressier.

Further reading

Barbour, B., & Stell, B. M. (2020). PubPeer: Scientific Assessment Without Metrics.
In Gaming the Metrics: Misconduct and Manipulation in Academic Research (pp. 149–156). The MIT Press. https://doi.org/10.7551/mitpress/11087.003.0015

Cabanac, G., Labbé, C., Magazinov, A. (2021). “Bosom peril” is not “breast cancer”: How weird computer-generated phrases help researchers find scientific publishing fraud. Bulletin of the Atomic Scientists. https://thebulletin.org/?p=93550