Mouse neurobehaviorist by day; research integrity watchdog by night

Mu Yang is happy with her gig at Columbia University, where by day she studies mouse neurobehavior. But by night, she turns into a research integrity sleuth, focusing on something entirely unrelated to her discipline: problems with images in chemistry papers. 

Yang began sleuthing in 2020, when she was helping a graduate student set up an experiment. The student handed Yang a paper on Alzheimer’s disease, Yang’s area of expertise, and asked if they could work together to replicate its experiments. After studying the data, she told the student that they were “too perfect” and that she doubted they were real. 

When the researcher who wrote the paper visited Columbia to give a talk, Yang went to hear him speak and challenged him about the data in question. “It didn’t go too well,” she says, noting that her Columbia colleagues also weren’t happy with her probing. That case was Yang’s first foray into the world of research integrity sleuthing, a pursuit that would turn into a hobby—and a consuming passion. 

But it was the only time Yang confronted in person a researcher whose work she has scrutinized. In that case, she ended up reporting it to the journal that had published the paper and the US National Institutes of Health, which funded the research. 

Yang struggled initially because she found a lot of papers reporting data that were too perfect, but few people seemed to care. “Pretty much nobody likes that kind of argument,” she says. “That is not considered hard evidence.” Yang says her comments were removed from the online site PubPeer, where scientists discuss, among other things, the rigor of academic research. 

So she pivoted to detecting image-related problems in life sciences papers. “My logic was that if this kind of thing goes hand in hand, and people don’t want to hear about data, then we’re better off catching the image manipulation,” Yang says. It’s a shame, she adds, since she estimates that image manipulation happens in 10% or fewer of data manipulation cases. “I think the vast majority of data manipulation will never be found out.” 

Image manipulation is also easier to point out to journal editors, who are often more willing to take action if sleuths do a good job highlighting the problem, Yang says. “It’s a capacity issue. If you don’t have the time and energy to pay attention to it, then it is not a morality issue.”

For the next 4 years, Yang continued to trawl through research and flag flawed papers. During that time, Yang collaborated with other prominent research integrity sleuths but declined to be part of lawsuits against alleged wrongdoers because she was concerned about restrictions to public discourse while litigation was ongoing. 

Change of direction

Eventually, Yang moved away from finding image-related problems in the life sciences after she saw that automated tools were becoming better at finding them. “I feel like I am subservient to [artificial intelligence], which is a really terrible feeling,” Yang says. 

After getting some pointers from Valentin Rodionov, a chemist at Case Western Reserve University who is developing AI tools to detect, among other issues, image-related problems in chemistry, Yang decided in April 2024 to make chemistry-related images her forte. She recalls making a lot of mistakes at the beginning. “I don’t even know how I didn’t quit,” Yang says. “It was hard.”

But she soon realized that she had a knack for spotting issues in images generated from techniques such as X-ray diffraction, Raman spectroscopy, and energy dispersive X-ray spectroscopy. Yang attributes her talent to time spent as a child doing Chinese hand painting and pencil sketching. 

Notably, Yang says, AI tools aren’t yet good at spotting issues with spectra from the chemistry-related images she is now focused on. “I think the squiggles are too context poor for AI to figure out.”

Still, in the past couple of years Yang has been approached by multiple scholarly services firms looking to train AI models on chemistry images. Although some of these tools have used her data, reports, and materials, she knows of none that have mastered the craft. 

Yang also receives some unorthodox requests. For instance, a researcher in Saudi Arabia messaged her on social media offering to pay her to investigate studies authored by their colleagues. “It sounded like a hit man job,” Yang says. She declined the payment but evaluated the papers after the researcher explained the situation in more detail. 

Since 2020, Yang has flagged more than 5,200 papers on PubPeer, and her efforts have led to around 330 retractions—roughly half of which were in journals of chemistry, physics, materials science, and related fields. Yang estimates that she spends around 30 h a week on sleuthing, including much of her weekends. 

At the moment, Yang is taking a closer look at the increasing role South Korean universities may be playing in paper mills—shady entities that sell authorship slots and citations on often nonsensical, plagiarized, or subpar academic papers. 

She has so far spotted and flagged close to 30 papers listing Yeungnam University School of Chemical Engineering as an affiliation. Some of those papers list collaborators from Saudi Arabia and India. Authors from the Philippines, Kazakhstan, Egypt, Germany, Tunisia, and Vietnam also feature as collaborators. “It looks like the Koreans are the sponsors because the Koreans are typically last authors,” Yang says, referring to the list.

A 2024 analysis found that unlikely international collaborations are often a telltale sign that a study originated from a paper mill. But Yang suspects that journals are becoming wiser and more often rejecting papers based on unlikely authorship networks. Yeungnam University didn’t respond to a request for comment. 

Two sides to the story

While Yang acknowledges that certain countries and regions appear more frequently on studies coming out of paper mills, she is concerned about practices that might risk reinforcing harmful stereotypes or profiling based on nationality.  

In fact, recently, she has found that empathy is starting to affect her sleuthing activities. For instance, since the US and Israel’s war with Iran began, she has—at least for now—stopped flagging papers by researchers from certain regions, even if high percentages of problematic papers emerge from them. “I feel like they might be struggling to accomplish something that we have a rigid and unrealistic standard for,” Yang says of the researchers. 

More generally, Yang worries that a cutthroat culture in academia tempts many scientists to cut corners. “Everybody is chasing after that stardom, as if you don’t chase after that, you’re some sort of failure.

“Most people are victims in this vicious cycle,” she adds, arguing that systematic changes are needed. “I have a newfound empathy for people that I used to think are cheaters and fraudsters.”

Yang is less generous toward academic publishers, particularly for-profit ones, which she argues aren’t doing enough to address research integrity issues. “They keep trying to publish more and more but aren’t keeping up with the quality control,” she says. 

Last year, Yang flagged nearly 1,600 chemistry-related papers to the publishing giant Elsevier alone. An Elsevier spokesperson says by email that the firm is investigating the papers on a journal-by-journal basis. “We are aware of the image integrity concerns raised by Mu Yang and appreciate her efforts in bringing these matters to our attention,” the spokesperson says. “Once the investigations are complete, we will take appropriate action, including corrections to the published record where warranted.”

Yang says she isn’t motivated by notoriety or by achieving a certain number of retractions. “I am doing it predominantly because it’s interesting and intellectually fun.” 

Still, she periodically takes a peek at how many papers other sleuths are highlighting. “It’s like any gamer society,” Yang says. “You have some sort of eye on each other’s numbers.”

“Highly prolific sleuths are all weird,” Yang says. “We made the choice to be weird.”