Visualization and Integrity
Nikolas Mariani
The Heidelberg Laureate Forum has a single purpose: To provide some of the brightest minds in mathematics and computer science with the space and time to make connections and find inspiration. Some of the connections made at the HLF will echo into collaborations and projects, with some of those efforts leading to concrete developments. The HLFF Spotlight series unpacks a few of those examples.
Lonni Besançon is an HLF alumnus and Assistant Professor of Data Visualization at Linköping University, who attended the 9th Heidelberg Laureate Forum in 2022. In addition to his work in volumetric data visualization, he is dedicated to providing more transparency in statistics. In recent years, he has also received considerable media attention for his work in conducting post-publication peer review, uncovering questionable research practices and finding fraud among a great number of academic papers. His sleuthing efforts and advocacy for science integrity have led to various media dubbing him a “troublemaker for truth” (ABC), a “white knight of academia” (Le Parisien) and “dogged critic” (Science).
Can You Visualize It?
Lonni’s major research endeavors concern volumetric data visualization. Put simply, he researches methods of improving interaction with spatial data – meaning information in at least three dimensions – or rendering data immersively; all in service of helping people understand complex data better. The key difficulty, Lonni explains, is in communicating 3D data in a 2D space, such as on any given screen where we would usually interact with information.
One example is in fluid dynamics research: placing seeding points in a fluid space and mapping their subsequent trajectory and behavior. One can of course do so by entering the coordinates of the seeding points manually. However, as Lonni suggests, a more immediate and immersive method is to visualize the space by transferring it to be viewed in augmented reality via a tablet or smartphone device, which is exactly what he achieved in one of his projects. This can allow the user to position himself in a three-dimensional space and intuitively place the seeding points in real time.
Another way of improving interaction with volumetric data comes in the medical world. In one of the projects Lonni helped supervise, researchers helped doctors conduct autopsies in a more streamlined fashion, by employing a mixed-reality approach. When examining a corpse, especially in the context of a police investigation, doctors wearing a mixed-reality headset are able to overlay pertinent information in real time, thus allowing them to, for example, cross-reference police evidence. They can also make notes of their own findings or save a screenshot of sections of the autopsy, all without needing to scrub out and record findings by hand. The entirety of the autopsy can then be summarized by a local large language model (LLM), which can organize the report and draw connections to relevant information automatically. Furthermore, such a session can create a 3D imprint of the procedure, which third parties would be able to review, allowing for maximum transparency. Finally, medical students are able to consult the entire session at their leisure for educational purposes.
This project, developed together with the Victorian Institute of Forensic Medicine (VIFM) and Monash University, promises to help make autopsies more transparent and verifiable after the fact. From a technical standpoint, it is currently viable, but at the moment is undergoing a review process until it is granted legal status.
Another interesting project for Lonni came about during his time as a PhD student, albeit separate from his normal research. A colleague had gone to a talk introducing a new form of eye surgery and reported: “It was so gory and hard to watch, but I wanted to get how they do it … there must be a way to keep the information that you want to get from such a video, but reduce the gory effect that it has on you.” Lonni and his colleague, a co-author of the paper, tried to come up with various algorithmic filters that would eliminate certain elements from the video – turning it black and white, softening the edges, transforming the video feed to a type of pop-art aesthetic. Just getting approval for the project was tough, says Lonni: “It took two years to get ethics approval, because we pretty much went to the Ethics Board and we said, ‘We want to show gory images to people and see how they react.’”
In the end, simply removing red hues was not sufficient, as the doctors he worked with pointed out that this led to too great a loss of information. The abstraction to a sort of comic book style had the most advantageous results – a viewer could still make sense of the procedure, but the gut-wrenching effect of the visuals was greatly diminished, even for the faint of heart. Lonni sees applications in many others fields as well, such as social media content moderators, who often need to sift through swaths of graphic imagery of a violent or sexual nature.
Transparency and Integrity
When not engaged in his regular research, Lonni is a fierce advocate for more transparency and integrity in the academic world. The use of statistics in many fields, he argues, tends to be overly reliant on P values to determine the statistical significance of results. Thus, he advocates for other inferential statistics to be employed, such as confidence intervals, in order to portray more nuance in results, and “think about statistics in a less binary way.” Ironically, this approach is somewhat in opposition to the spirit of Lonni’s regular research in visualization, as he admits: “I went against all the principles of my field. People are like ‘we should make it clearer, we should make it easier.’ And I’m like ‘no, let’s make it harder! Let’s make it blurrier.’” When we ask whether he thinks such an approach will catch on more, Lonni does not seem too hopeful: “People like to have a cookbook that they can just reuse over and over again.”
Lonni traces part of this dilemma to the pressures that particularly young researchers face in modern academia: “People need to publish a lot, so they put all their emphasis on publishing, publishing, publishing; and never take the time to do slower, more robust science … The academic system is making all academics – me included – rush things instead of taking their time.”
At the heart of this, Lonni sees the difficulty in how researchers and their work is evaluated: “Find me a job that’s [as] high-skilled, where there are so many different metrics of success, which are used to pressure academics … You need to have a high edge index, a lot of papers published, a huge number of citations, you need to bring in research money, you need to be good at teaching, you need to have a lot of PhD students, you need to have a lot of master’s students, you need to give a lot of talks … the skillset is this large.”
Lonni says that when he attended the Heidelberg Laurate Forum in 2022, he was happy to see that “the struggles I talk about are not just stuff that I’ve seen … In the case of the HLF … people coming from all over the world … They were all gathering around these issues and being like, yeah, this is really problematic.”
Post-Publication Peer Review
One of the ways Lonni himself is trying to foster some improvement in the integrity of academia, is through what might be called his “extra-curricular” work: investigating and reporting countless scientific papers of different fields for fraud, dubious publishing practices, and questionable ethics violations. In our conversation, Lonni says he has reported some 600 papers for fraud.
Many such cases were simply blatant, as he explains: “So they actually copy pasted from existing papers and then replaced every word by a synonym.”
Others cases concern the supposed citation of published papers where “if you look at Google Scholar, a paper will have no citation. But if you look at other dashboards like Scopus … you would see it’s been cited a hundred times. But all of the papers that supposedly cited it don’t cite it.” In a single case, “one researcher benefited from 3,700 citations” from “just three journals.”
This all started, he says, with stumbling upon a few dubious papers during the COVID-19 pandemic, leading Lonni to publish a paper where he had discovered some 270 papers where the time from submission to publication was less than 24 hours, a turnaround time unheard of in academia, where the reviewing process typically takes months, sometimes years.
This sleuthing dovetailed into Lonni and his collaborators debunking a considerable amount of work by French epidemiologist Didier Raoult – including his most cited COVID-19 paper – on methodological grounds. All told, Lonni and his collaborators have reported approximately 800 papers by Raoult, many of them for ethics violations. In fact, on the website Retractionwatch, which presents a ranking of scientists with the most paper retractions, Raoult holds the distinction of being number 19 worldwide – with 40 papers retracted at the time of writing. Lonni estimates 95% of their reportings of Raoult’s papers will result in retractions, which should ultimately bump him higher on the list.
Unfortunately, this has led to some backlash against Lonni, with Raoult and one of his collaborators lobbing accusations of “terrorism” and even “sexual harassment” at Lonni, all of which has been the subject of two defamation cases.
As for whether or not this has deterred him from carrying on, Lonni simply shrugs and says “If they get that upset, then I definitely must be doing something right.” Lonni is currently active with anywhere between 20-30 people in a small group that use some of their free time to track down instances of academic malpractice.
Where Do We Go from Here?
When we ask him how he thinks the academic world can do better to avoid some of the issues he has encountered, Lonni states:
“We should put less focus on the numbers of papers published, or the number of citations, and put more focus on other measures that are important, which is robustness, validity of findings, reusability of the findings, social impact …”
“If you do something that is barely cited but changes the lives of 2 million people, that’s worth something, right? Yet there’s no measure of that.”
“If you do something that is incredibly well done, but it took you three years to do it instead of 6 months to do it quickly and kind of “half-ass” it, I’d rather have three years of your time on something really robust, especially if it has social consequences.”
We ask Lonni if he is concerned that others will begin to return the favor by scrutinizing his own research: “I’m happy to see that people are trying to take apart all my papers. That has happened a lot, as retribution like ‘we’re gonna tear down your papers.’ Be my guest. If they deserve to be retracted, they should be retracted. We should stop seeing retractions of papers as failure because it will be a self-fulfilling prophecy. It shouldn’t be career destroying if you need to correct mistakes in papers.”
You can read more about Lonni’s projects on his website, https://lonnibesancon.me, or contact him at lonni.besancon@gmail.com.
The post Visualization and Integrity originally appeared on the HLFF SciLogs blog.