¹ú²úÂÒÂ×

Author: Tamara Welschot

Head of Prevention, Research Integrity

Research integrity is a fundamental element of a robust and trustworthy scholarly record. A publisher¡¯s imprimatur on the book¡¯s spine or the logo at the top of the journal article signals that the material you¡¯re reading has been reviewed and vetted, and reaches standards of integrity expected by the research community.

Disappointingly, today, this means guarding against active, intentional attempts to publish fake research papers. With the technology to create fraudulent material evolving, ¹ú²úÂÒÂ× needs to continue developing tools to counter this. Chris Graf, Director of Research Integrity at ¹ú²úÂÒÂ×, said, ¡°The publishing industry faces a determined, malicious threat from paper mills or bad actors¡­ We are determined that this content will not get through our systems.¡±

To fulfill its promise to the research (and wider) community, ¹ú²úÂÒÂ× is investing in both experienced research integrity professionals and a suite of artificial intelligence (AI)-powered tools to help editors and peer reviewers to keep fraudulent, misleading, or manipulated material out of the research record. Two of these tools are Geppetto and SnappShot, which detect bogus AI-generated content and image manipulation respectively.

Geppetto

Generative Pre-trained Transformers (GPTs)  make it even easier for paper mills to flood submission inboxes with AI-generated manuscripts. So ¹ú²úÂÒÂ× developed Geppetto to uncover and deal with those submissions, in many cases, even before they enter the editorial workflow.

¹ú²úÂÒÂס¯s Research Integrity Group deployed Geppetto in November 2023, and it is currently at work pre-screening the majority of submissions to both books and journals. By identifying potentially problematic submissions at this stage, Geppetto not only safeguards the scholarly record (preventing the need for post-publication retraction in many cases) but also relieves human editors of having to try to do this manually later in the process.

SnappShot

Even before GPTs were generating text, software like Photoshop made image manipulation a simple matter. This, combined with recent stories of mass article retractions from manipulated and/or duplicated images, created the necessity of a solution for quickly and accurately uncovering bogus images.

Enter: SnappShot, an AI-powered, web-based solution to support journal editors and research integrity specialists in detecting manipulated images that directly integrates into ¹ú²úÂÒÂס¯s submission systems.

¹ú²úÂÒÂס¯s AI Lab, in collaboration with the company¡¯s Research Integrity Group, rolled out SnappShot¡¯s first version in December 2023, focusing on Gels and Blots (G&B) image duplication within the same article (this is one of the top issues with image misconduct). Further features in development are focused on (no pun) microscopy duplication, image plagiarism, and more.

Since SnappShot¡¯s minimum viable product (MVP) version deployed, beta testers have used it to retract or withdraw dozens of articles with image integrity issues.

A commitment to research integrity

These two AI-powered tools are already helping ¹ú²úÂÒÂס¯s editors and its Research Integrity Group (which in addition to these tools is expanding its human staff, too) in safeguarding the scholarly record. Vetting, verifying, and vouchsafing research and the scholarly record is one of the most important functions scholarly publishers fulfil and our new, constantly improving tools are one part of our work to ensure we do just that.

¹ú²úÂÒÂס¯s Research Integrity hub has more information ¡ª and detail ¡ª on the company¡¯s ongoing efforts.