Leslie McIntosh is Vice President of Research Integrity and Security at Digital Science, where she leads work to identify mis- and disinformation in scientific publications. She has spent her career building tools to assess and improve the credibility of scientific research, from leading biomedical informatics at Washington University in St. Louis to founding Ripeta in 2017, which now sits within Digital Science. She previously served as the inaugural Executive Director of the Research Data Alliance – US, the regional arm of a global organization advancing recommendations for data sharing and interoperability.

In this interview, she discusses automated manuscript screening, paper mills, authorship-for-sale, and how research integrity is increasingly converging with research security.

[SIA] In your 2021 interview with Research Outreach, you explained the motivations behind founding Ripeta with Cynthia Hudson Vitale, who is now at Johns Hopkins University and Lyrasis. You brought a background in biostatistics, epidemiology, and biomedical informatics, while Vitale brought expertise in research data infrastructure. How did this combination shape the design of a tool intended to automate manuscript checks?

[LM] The question we posed was, how could we check manuscripts in a way that could facilitate the ease of research? What we were attempting was to improve the quality of the research output, enabling researchers to get on with their work. These are people doing excellent work with very little time, so we wanted to make quality checks less burdensome for them.

The combination of expertise was key. Cynthia is a trained research and data librarian, which was critical to designing checks on science that became a feature of Ripeta. At the time, I was running the Center for Biomedical Informatics at Washington University in St. Louis, and my contribution was understanding what world-class researchers needed to make their lives easier, how to automate checks, and build processes around them. But the library perspective shouldn't be underestimated: a librarian thinks about data organization in ways I simply didn't, certainly not then.

[SIA] During the early years of the pandemic, you applied the RipetaScore framework to assess the credibility of COVID-19 publications and found wide variation in basic reporting practices. What did that exercise reveal about the speed-quality tradeoff in pandemic-era publishing, and which of those lessons do you think the publishing system has actually retained?

[LM] The biggest exercise for us at the time was to see how much information you could get from the data availability statements. While having data available was a good practice, it was a "nice to have"; there wasn't an urgency about it until the pandemic. And when the pandemic happened, you realized you needed the data to be easily accessible for quicker decision-making; data availability has come a long way since then.

But I almost think we have peaked, as now we have other challenges with having data available. There are security problems when people take that data and misappropriate it, using it for large language models (LLMs) or for paper mills. When you solve a problem, you end up figuring out what the next problem's going to be, and that's what we have done.