This story was originally published in our Nov/Dec 2023 issue as “Science Sleuth” Click here to subscribe to read more stories like this one.

Science has a problem: It’s conducted by humans. Many of these humans follow the rules and procedures outlined by the scientific method and conduct themselves with integrity. But as with any sector of society, there are bad actors who cheat to get ahead.

In 1912, an archaeologist named Charles Dawson combined and altered pieces of human and ape skulls, claiming to have found the evolutionary “missing link” between humans and apes. His sham wasn’t discovered until 1953, after the invention of new bonedating techniques. Though Dawson couldn’t get away with this today, for every new technology that uncovers a fake, there’s another that provides a new way to fake findings.

One place this crops up is in falsified images presented as results in research papers. Fortunately, Elisabeth Bik — a Dutch microbiologist turned “science integrity detective” — is fighting back against this kind of fraud. Working as a digital vigilante, she has amassed an audience of more than 140,000 followers on Twitter. She’s meticulous, with an incredible eye for detail and the patience to put that eye to serious work. And she’s become a master at spotting certain types of falsified images in research papers.

Stats estimating the frequency of fraud in scientific research are hard to come by, in part because we can only know about the incidents that have been caught. But work like Bik’s helps to provide a minimum estimate: Her screening of more than 20,000 papers published between 1995 and 2014 found 3.8 percent had problematic figures. About half of those had signs of deliberate manipulation.

Bik’s ultimate goal is to reduce these wrongs, or at least generate scrutiny and corrections from the journals publishing scientific research. So far, she has contributed to more than 1,000 retractions; in practical terms, this often means a warning label is added to the online version of the article. This doesn’t prevent researchers from unknowingly referring to an already downloaded or printed copy. But it’s still an important fail-safe to make sure scientists don’t base future work on false information.

As she logged her 7,163rd problematic paper find, Bik spoke with Discover about how she got into research fraud detection, what her job is like and what it all means for science.

Q: Your expertise is spotting image duplications. What does that mean?

EB: Research papers have figures that show results. And in some fields, these figures are actual images, actual photos. For instance, a microbiology paper might have a figure with multiple panels of photos showing cells that have been treated differently. When there’s a duplicated image, instead of each panel being unique as it should be, two of the panels are identical. Or maybe two of the images overlap, meaning the researcher took a second photo of the same sample under the microscope, but just moved it over a little bit so that some of the same cells appear in both images. Of course, if they had just moved the sample a little bit farther, there would be no overlap for me to find. So, I really only find the tip of the iceberg — I can only catch a small percentage of actual fraud that likely occurs.

Q: What do you think motivates people to commit research fraud?

EB: There is pressure to publish as scientists. And there are situations where there could be a bully professor. There are labs that are run by big egos who might say to a young researcher, “Why did your experiment fail? I will hire someone else who will make it work. I’ll fire you, I’ll send you back.” A lot of researchers might be working internationally on a visa and, if they get fired, they have to go back to their home country within weeks. I think in a situation like that, researchers will feel the pressure to make up results in order to produce what the professor wants, and everybody is afraid to speak up. The graduate students and the postdoctoral researchers might be the ones photoshopping, but who is responsible for the atmosphere and the integrity of the lab? That’s the professor.

Q: How did you get into this work?

EB: I was born and raised in the Netherlands, and then worked at Stanford University for 15 years in the human and dolphin microbiome field. While there I became interested in plagiarism — and by coincidence, one of the Ph.D. theses I was investigating had images that were duplicated. I realized I have a talent to look for these things, so I started scanning the biomedical literature for them.

I focused on papers that had photographic images, like in molecular biology and oncology, and that used techniques I was familiar with from my own research. I started scanning in a systematic way, because I wanted to know: How often can we find these types of images? Out of 20,000 papers, I found 4 percent of them had these duplicated images. I did that for a while, and then worked as an industry scientist for a few years while continuing my image duplication work as a hobby.

Q: Is this a full-time job now?

EB: Eventually, in 2019, I realized I talked more enthusiastically about my image duplication work than my job. I decided to quit the next day, and that actually turned out to be sustainable. I created a Patreon account so that supporters can donate to fund my work, which puts me in a unique position that allows me to call out these bad papers: I don’t need to worry about my job because I don’t have a job. And I’m late enough in my career to not have to worry about my career. It’s not that I’m super courageous, rather that most people are not in a situation where they can raise these issues without damaging their own careers.

Q: How do you decide what papers to check, and how do you go about checking them?

EB: I’ll often follow threads, since researchers and labs who do this tend to do it repeatedly. For instance, just last night I discovered a “lab of interest” in the U.S. who has worked with several labs of interest in China. I’ve already found several papers that are suspicious and have duplications. When I started out, I’d look for duplications by eye, but now I use a software called ImageTwin. It scans and extracts all the images in a paper and compares them to each other, as well as to images in other papers. It’s very good at finding overlaps.

When I find a duplication, I’ll take a screenshot and outline the duplication, then post it on PubPeer. That’s an online database for scientific publications where you can comment for authors, publishers and other scientists to see. I try not to make any judgments there, since some duplications might be honest errors. Even when it’s very likely that it has been done intentionally, I won’t say, “This is fraud.” I try to stay as objective as possible.

From there, the paper authors should get an automated email from PubPeer, and they can reply. Most don’t. Some journals or publishers will also get an email, though many screen manually for entries about their papers. The publishers can then decide whether to retract the paper.

Q: Do you think these duplications are typically intentional?

EB: When it’s just the same photo twice, I almost always think it’s probably an honest error. It’s easy to imagine that a scientist might take hundreds of photos and put them all in labeled folders — this is this sample, that is that sample — and one day instead of a two, they type a three, and they happen to use the mislabeled one for their paper. Yes, it’s sloppy. Yes, they should have noticed it, or another author might have noticed it, but nobody did.

But sometimes a paper has a lot of duplicated images. Or sometimes you see the same photo twice, but it’s rotated or mirrored. In these cases, it’s more likely that it was done intentionally. And sometimes the image itself has been altered, like a photo of cells in a tissue where the same group of cells has been stamped multiple times in the image. That will almost always be intentionally done.

Q: Do you ever come across entire papers that have been faked?

EB: Yes. There’s another situation where people can purchase an entirely fake paper from what’s called a paper mill. The paper mills we’re finding are mainly or almost exclusively active in China, because there are heightened expectations there: Medical students and clinicians are often pushed to publish scientific papers — even though these are medical doctors who are not interested in research and don’t have time to do research. So, they just buy an authorship from a paper mill. But the paper is totally fake, made by professionals. The patients are made up. The photos can be real, but they’re not representing what they say they are. And maybe they make use of artificial intelligence to generate fake images.

Q: How worried are you about AI being used for research fraud?

EB: I’m very worried. ImageTwin is looking for duplications, but AI-generated photos are unique — I’m not sure if I would recognize a well-done microscopy image that is completely fake. But we tend to believe photos, right? We are so reliant on our eyes. If you see a photo in a scientific paper, you tend to think it’s real, and it will pass peer review. But if we can no longer trust photos, I don’t know where we’re heading. Anyone could misuse AI to generate false information, photos and text. And that’s what I’m worried about, because we cannot distinguish it from what’s real.

Q: It seems like public trust in science is pretty shaky lately. How do you feel your work plays into that?

EB: Misconduct is bad for scientists. But it’s also bad for science, because it could be used for people with the agenda of denying science. I feel the work that I do is important because science should be self-correcting. Sometimes our hypothesis might not be quite right, and the answer is always much more complicated than we thought.

I’ve worked on over 7,000 papers; 980 have issued corrections and 1,028 have been retracted, but there are millions of papers published every year. Science fraud is only a tiny fraction of science. I always remind people that I’m not trying to be a science denier and that this is the only way we can solve the big problems we’re facing, like climate change and pandemics and whatever else. We need to be critical within science in order to make science better.