Our AI Journey

Research Integrity and AI: Navigating Challenges and Using Potential

Chris leads Springer Nature’s work on research integrity. Research integrity is a critical part of quality in research and research publishing, and it’s an aspect of quality that’s super meaningful to communities of researchers as well as to research publishers ... not to mention the citizens of the world who, in the end, all benefit.


Chris Graf

Springer Nature Group

In my role, nothing stands still - what we need to do at Springer Nature to ensure integrity and quality is always a work in progress. That means intentionally making careful, considered innovation happen in procedures and policies and training and also - of course - in technology. And in the last two years, with partners in teams across Springer Nature, we've done amazing things with large language models and with computer vision that solve precise problems in research integrity. To be honest, it felt like a slow burn for much of those two years, with rapid progress in the last few months not only in technology but also in the people-power needed to use that technology. Kind of like slow-release carbs for an athlete, rather than that fast sugar hit. And that's what I love about it: We had the space and expertise to build strength into the underlying platforms that enable us to then innovate rapidly and create powerful solutions, and then we deployed those innovations to maximum effect.

AI - its threat and its potential for ensuring reliable research

To solve the world's toughest challenges we all need research that's reliable, that demonstrates integrity, and that researchers, innovators, and policy makers can build on. And we need a healthy and robust way to support researchers with quality control and to communicate that research - that's research publishing, and is what Springer Nature does so well. It’s kind of like the “if a tree falls in a forest and no one hears it...” thing: Without enabling researchers to communicate trustworthy research effectively to other researchers, to policy makers, and to innovators then researchers’ work can't be used and won’t make an impact. So that's what gets me out of bed in the first place. Now let's talk about AI and its potential for partnering with the many subject matter experts who work in research publishing in roles like editors in chief, and the many members of research communities who also contribute as peer reviewers and authors, for enabling new kinds of quality and integrity in research publishing. I can see the great potential there, of course. But I also want to make sure that AI is used carefully and responsibly, to supplement human expertise and agency in the evaluation and publication of research, not to replace it. Frankly, while there's a lot of excitement there’s also a lot of “hot air” about AI in all sectors, including research publishing. I want to tap that excitement and potential, but also be realistic and sensible. That's a fine balance to find, and I think we’re getting it just about right at Springer Nature.

Setting a thief to catch a thief – a method to spot and remove fake papers

Many research publishers are concerned about the rise of "paper mills." Paper mills are what we call the people (or organisations) that create fake research papers and sell authorship to scientists, then break into publishers’ processes to ensure those fake research papers bypass publishers’ quality and integrity checks. Paper mills have created and successfully avoided publisher’s quality control processes to publish thousands of fake papers, many of which have now been retracted post-publication by those publishers. You can see how this undermines the integrity of research and research publishing. One of the innovations we've created with AI at Springer Nature addresses the clumsy attempts of paper mills using large language models to create those fake papers. We “set a thief to catch a thief” kind of like Batman vs Superman, or King Kong vs Godzilla: we developed a method to detect the clumsy use by paper mills of large language models, and spot and remove the fake papers. Nice.

Pattern recognition to ensure integrity and quality

Beyond "setting a thief to catch a thief” I'm really excited about our emerging capabilities in “pattern recognition.” Looking at the patterns in data that are in and around a piece of research and across multiple dimensions enables us to draw a “trust map” for that piece of research, and so understand where we need to focus our efforts to ensure integrity and quality. Because this is the real world, there are limitations, and we need that focus to put our resources to work where they’re most needed. That focus is critical to deliver the high quality that researchers expect of us. And the demands from researchers for publishing services from publishers keep growing year on year, so to keep pace and to deliver what researchers need from us we need to think in new ways and to work in new ways. Needless to say I’m very excited about our pattern recognition work.

Cultivating research integrity by innovative teamwork and backing from leadership

The leadership team at Springer Nature sets the right tone from the top. Members of the team ask the right questions, and have been ready to invest heavily in integrity. The result: My team's tripled in size in the last two years, we've unlocked talent and partnerships that have delivered truly innovative solutions to integrity problems, and that investment continues. If I did need to ask for more, then I'm sure I know who and how to ask. And my team - what a great team - we have superb collaboration skills and deep knowledge, and that makes partnering with colleagues challenging in a good way: we challenge and support each other. And we get great things done.

Back to overview