Tackling paper mills and bogus research: Some things are better done together

Author: Chris Graf

Research Integrity Director

Today marks the launch of the first phase in a new initiative from the STM Association and its members to address publication manipulation: the STM Integrity Hub. In particular the initiative will target so-called paper mills, organisations that deliberately set out to publish dubious manuscripts by manipulating editorial and peer review processes. 

The paper mill problem

There are countless reasons why paper mills are a problem. Mills steal credentials from legitimate researchers and con their way into editorial positions at journals. They falsely use these credentials to lend authority to an author- or peer reviewer-list. They mis-direct use of research funds that, presumably, would otherwise be spent by bonafide researchers on legitimate research. They dupe journals into publishing manipulated material as if it were real research, which can mean months wasted from time-limited grants held by legitimate researchers who try to verify and use findings in their own research. They con insufficiently wary researchers into purchasing bogus services, and perhaps enable those who are less naïve to do the same knowingly. This all slows down research and threatens its credibility, which in the end not only harms research and researchers, but us all. 

The challenges, first-hand 

Springer Nature, like many publishers, has experienced the challenges of paper mills first-hand. Late last year, one (or perhaps more than one) mill gained access to our editorial processes, leading to the ‘passing off’ of fabricated manuscripts as legitimate research in four guest edited issues in the Arabian Journal of Geosciences and Personal & Ubiquitous Computing. Many of the papers in these issues showed clear signs that artificial-intelligence software had been used to synthesise the text, likely drawing from discipline-specific sources. Our assumption is that this technique was adopted by one or more paper mill to fabricate manuscripts more easily, to make them harder to detect and -- in combination with efforts by the mills to effectively remove editorial oversight by inserting fake or compromised guest editors and peer reviewers -- to increase the mill’s success rate. 

At Springer Nature we took a number of steps rapidly, and that work continues (see ‘Footnote: What we did’). 

Steps towards solutions

For publishers, solving the problem is a challenge. Our traditional ways of working are based on trust, and we want to see legitimate researchers continuing to benefit from that. So addressing the malicious work of the paper mills without making publishing harder or less trust-based for the vast majority of legitimate researchers, is going to take coordinated and sophisticated approaches. 

And here, let’s rewind to the starting point: Today marks the launch of the STM Integrity Hub, led by the STM Association, the trade association for academic and professional publishers. When it’s up and running, the hub will give all publishers, large and small, support in preventing paper mills and other forms of bogus research from being published. The ‘hub’ will be a cloud-based service where all publishers can check submitted manuscripts for integrity issues and share information and data about the integrity problems they themselves identify. 

That collaboration and sharing is important. Paper mills threaten something that’s fundamental: the quality and integrity of the work we collectively publish and that researchers rely on. However, the scale of their activities compared to the publishing enterprise as a whole is relatively small and, in actual fact, some publishers have little data about paper mills and other integrity problems with which to design technical solutions. Even if a publisher has the data, the investments needed to solve the problem are large and ongoing -- likely too large for some. Hence the ‘collaboration.’ The hub will enable information sharing in a way that brings access to the scale of data we collectively need for effective solutions and the speed and agility we need to respond to new (or evolved) concerns, with legal compliance for privacy and other regulations built-in. 

This week is the first step for the hub, with the launch of a ‘demonstrator’ to show how the theory works in practice. The start point for the demonstrator is simultaneous duplicate submissions, which are an indicator of paper-milled articles. 

I’m thrilled to be involved in this work. We’re already moving faster than I think any of us thought we would. And that’s down to the highly experienced and talented colleagues from across publishing houses and allied companies who are each making fabulous contributions, along with equally excellent and committed colleagues from the ‘STM Solutions’ arm of the STM Association. 

Some things are better done together… and this is one. 

Footnote: What we did so far at Springer Nature

This footnote explains how we responded to problems with the four guest edited issues in the Arabian Journal of Geosciences and Personal & Ubiquitous Computing mentioned above. We investigated and corrected the publication record, retracting all affected papers from these issues, and more broadly:

• Immediately put on hold all guest edited issues that were in progress across our entire portfolio of journals to conduct technical and manual checks for AI generated content, other indicators of manipulation and verification of guest editors’ credentials. 
• Contacted editors-in-chief to highlight concerns regarding guest edited issues, and provide guidance on how to rigorously assess unsolicited approaches to ensure they are confident that the guest editor is who he or she purports to be and will handle the submissions appropriately.
• Continued to invest into the development of tools that help identify problematic content and prevent it from being published (e.g., software to identify manufactured manuscripts, to identify manipulated images, and to identify other suspicious patterns in manuscript submissions).