 but we can start with an introduction and presentation before I start I would just like to say that you can ask questions in the chat and in the Q&A during the talk but let's start with our jingle. So welcome everyone to this fall's first E9 monthly webinar and I'm so honored to have today with us Dr. Elizabeth Bick she's going to talk to us about double trouble so Elizabeth the floor is yours tell us what is the double trouble what do you mean by it and I'm really amazed here with us today I'll just introduce you you are Dutch American microbiologist you have worked with for 15 years at Stanford University and then later in industry but you are a science integrity volunteer and consultant and you did amazing job by scanning biomedical literature for images or data that have actually resulted in a reflection of a whole lot of papers so you have an eye for seeing problems in in the images and you have also received a 2021 John Maddox prize as well for your work so thank you for for being here thank you for doing this amazing work and yeah the floor is yours. Thank you so much Sonya for having me here I love the jingle also that was very nice all right let me set up my presentation right you hopefully should see my screen so yeah I'll talk about double trouble so what do I mean with that I'll talk about image duplications in biomedical papers and before I start my talk I will give my financial disclosures so because I did quit my job I work at Stanford I worked two years in industry and at some point I decided to just switch or basically quit my job and I work on finding image duplications in papers and but yeah how do I make my money currently because I'm not employed by a university or by a company I work for myself I do some consulting I get speakers fees to give talks this talk is for free if I remember this correctly but I occasionally will get speakers honorarium and consulting fees so I mainly work for universities and scientific publishers to investigate particular cases of allegations of misconduct and I also have a patreon account which is sort of a crowd funding site where people can give donations so the average donation per person is $6.50 if I at least in the latest time I checked so it's small amounts but together that gives me a basic income so I don't have to worry too much about income and I really appreciate that people support my work because there's not a lot of grants that will fund the type of work that I do which is basically raising concerns about scientific papers I also have worked for a fraudulent company called you buy and I still have four patents from that time although the company was rated by the FBI the founders have been charged with insurance fraud and they have fled from their fugitives for the US government and apparently they're somewhere in Europe but so you know you might run into them at some point but yeah the company the patents I think are not really worth anything the company is bankrupt and yeah it's sort of a you know I I embrace that I work for that company even you know it was involved in insurance fraud but I felt that the scientific tests that we were developing at least had some merit and I am still proud of the work I've done there and none of the employees have been charged it was the founders who have been found guilty but yeah my my criticizers usually want to you know let you know that I worked at a fraudulent company so I'll be the first to admit that I did work there indeed so what do I do I look at image duplications in scientific papers and it all started in 2014 as I was scanning a phd thesis that I suspected had plagiarized text and it had but as I was flipping through that phd thesis and the chapters have been published in scientific papers I found this particular image here shown here in in in blue I'm not sure if you can see markers or probably not but the the one that I marked here in the below in cyan color in that cyan box that had a little dot in it in the fourth lane and I sort of remembered that and I flipped through the phd thesis and I saw the same image twice in another chapter one time it was the exact same image but it was sort of cropped differently and stretched differently but another time it was even rotated 180 degrees and it was still but it was representing different experiments and I found another image marked here with red boxes that was also used twice again to represent a different experiment and so that that made me very angry because I thought especially with this rotated image the image that had been used twice but rotated that sort of suggested that this was done deliberately and it also made me realize that I have some talent for detecting these duplications it seems perhaps very obvious but both of these phd thesis chapters have been published in scientific papers and so they have past peer review but you know by accident you can sort of compare these things and remember these plots and yeah I recognize that they were the same and of course when you submit a paper that is not really part there's no at least at that time there was no good plagiarism detection for images and so now there are some tools being developed but yeah it's really hard to recognize then but I just happened to see them in the same place so I've reported both of these papers to the journal and both of them were retracted a couple of uh well it took still some some time but at some point they were retracted and so I I thought about like why is this bad because I was first working on plagiarism and you know obviously plagiarism think we can all agree that's not good but it doesn't necessarily create fake signs but representing the same photo using the same photo twice to represent different experiments that appears to be either falsification or fabrication because one of these experiments did not happen and so it made me angry because for me if if you think about it for me personally science is about finding the truth and representing the same image twice is not the truth it's it's cheating it's it's misconduct and yeah I also thought about like science publications this is this is also how we communicate with each other as scientists we build on each other's work and we use publications to to look for for inspiration for our own research and so researchers never just do research by themselves we always build on the work that other people have done before us and so every science publication is like bricks in a wall and I use this metaphor a lot but I feel if one of those bricks contains fraud or an error that means that part of the the other breaks the other publications that are resting on that work could be tumbling down and so science fraud is is everything that science should not be it's not finding the truth and it's it's hurting other people who try to replicate those work and we if we are scientists you might realize how much research is really hard to replicate and some of that there's many reasons that can cause that but science fraud could be one of them and unfortunately science historically has been built on trust as scientists we tend to trust each other's work if we do a peer review we tend to think this is really what happens but unfortunately there is fraud in science like in any other field you can think of think of banking financing there there is construction perhaps there is fraud everywhere and also in science and so I feel scientists and editors and and professors have not really thought perhaps too much about science fraud but but yeah it is fraud and it's more and more unfortunately also organized and I'll talk about that in the second half of my talk but yeah thinking about why scientists do fraud I like to say that behind every case of misconduct there's a sad story there's a case where people felt the need to cheat and so um one of the papers that uh yeah was retracted is is actually so I'm showing that here on the left that is one of the papers of the the image duplication that I first found that I just showed on one of the previous slides so that paper got retracted and I blurred out the names because I I feel I don't want to make it about the people who do misconduct I I will make it about the papers but there's multiple papers on every multiple authors on every paper who is responsible for doing the misconduct who did that image duplication now in biology usually the first author is the person who did the lab work the last author is the person overseeing the lab work so perhaps the professor but who is responsible for doing the misconduct and it's easy for a senior author to to put the blame it's that person who did it it wasn't me and that's often unfortunately what you hear but all authors on a paper are responsible for the for the content of the paper and you could actually argue that perhaps yes it was the the first author maybe a grad student or a postdoc who who cheated but the last author's role is to be the mentor to be the supervisor to make sure that the integrity of the paper is correct and unfortunately there's many situations where people for example in the US might be working on a visa because they're they're from a different country and they work on a temporary visa to do research in the lab in the US and if you happen to have a professor who's a bully and very demanding you could have a situation where perhaps the graduate students are being told that unless they give the professor the results that he or she wants they might be fired and if you're fired your visa immediately expires and you have to leave the US go back to your home country within a couple of weeks and so under a threat of being fired you can imagine that people want to please the professor and want to give him or her the results they want and so science misconduct is complicated it's not usually just one person you know deciding to cheat they're usually in a very complex complex situation where they feel the only way out is to do misconduct and unfortunately it's also you know the rewards are big because if you do misconduct you have better results and so you'll you'll you know you can publish this is the publisher perish mentality that many of us have to deal with and the consequences of for science misconduct at least in research seem to be very small there's a very small chance of being caught and it seems that a lot of professors who you know under whose responsibility there's lots of cases of suspected misconduct they get away with it because usually they will blame the persons who are the first author so the more junior researchers so it's a very complex situation but you could also argue or there can be many problems with a paper and having just one person being sorry there there's there's two things so Holden Thorpe the editor-in-chief of science wrote an editorial about it and they you can say okay there's a problem with the paper and then there's the question who is responsible and these are two separate questions hopefully papers can get retracted without the very lengthy investigations that institutions usually need to do to investigate who is actually responsible for the for the misconduct now focusing back on figures so when you think about figures in scientific papers this is what I look at but I mainly focus on photos I will look at photos and if you think about photos in scientific papers you can you can see there they're like the photos we can see here on the right photos of cells or gels or blots or even mice or tissues things like that there's lots of details and we can tell that the photos shown on this slide they're all good I don't have any suspicions that there's any misconduct on them but we could perhaps look for duplications now photos are yeah like I said have a lot of detail if you think about other types of figures in scientific papers such as line graphs so shown here on the left there is some detail there but it's very hard to know if these images are based on real experiments at least with a photo you can see oh there's a photo of a tissue so they really did some experiments but photos images of line graphs plots and spectra and an ordination plots and bar graphs and things like that it's actually very easy to cheat in that without leaving traces I mean you can type in some numbers in excel and make a bar graph that looks realistic you would not really be able to catch if there was any photoshopping or overlapping images but photos at least there's some detail there so that is what I focus on now there can be many things wrong with photos that are actually not detectable and as an example I think we're very used in this society of looking at images that are heavily manipulated just as a example this is a cover of a magazine and on this is Faith Hill she's a country singer and on the left is the actual photo that was taken of her and on the right is how she ended up on the cover of the magazine and just pay attention for example on how thick her arm is you can see it's like you know reduced to a very skinny arm and I mean this person is already you know perfect to look at but yeah somehow these photos get so much airbrushed and you know body parts are made slimmer and you know apparently what we think is more beautiful but you you can see how much we're used to being you know having looking at photos that are manipulated and so a lot of people have filters instagram filters and things like that and so it's very tempting to also manipulate your photos but it's very hard to detect like I would not be able to look at this right photo and know that it's manipulated because it doesn't leave any traces but in several cases there are traces to be found of duplication so I cannot really detect the manipulation but I can detect a duplication so here are three types of duplications that you could find in scientific papers so starting from the left and moving to the right there's three categories and category one on the left is a simple duplication where two images are exactly the same but being used to represent two different experiments so the ones boxed with red boxes are duplicates of each other and the panels marked with blue boxes are also duplicates of each other and so this could be an honest error people make maybe the researcher made thousands of photos and just happened to pick the wrong one that could happen it's sloppy obviously and if it's a lot of sloppiness maybe it is misconduct but in many cases these simple duplications the exact same photo being used twice could be the result of an honest error it's inappropriate though it's an inappropriate image duplication the example on the middle shows an example of a repositioned image duplication here there are four photos of cells being treated with four different amounts of radiation so each photo should be different but the two panels at the top are overlapping and I mark the overlapping areas in green boxes and the two panels on the right also overlap with each other marked with blue boxes so instead of looking at four photos at best yeah there are four photos but they appear to be only of two specimens I cannot find an overlap with the fourth image but maybe there is one but yeah is this is this an honest error is this done deliberately um you know we can have a long discussion about that it's hard to tell I think it's deliberate but who knows and then the example on the right is a is an example where photos have been manipulated to contain now duplicated elements within the photo so the photo itself has been digitally altered so you can see four blots a through d and in blot a you can see that lane one and lane three look identical I've marked those with blue boxes and in lane in blot d three lanes are identical marked with red boxes that is very likely to have been done deliberately with an intention to mislead and so these three types of duplications are helpful in determining whether or not something was done um yeah with intention to mislead now here's an example of an inappropriate image duplication of a type one and let me my cursor is not moving that's my cursor yes oh there we go so here's an example of a type one duplication where you can see seven panels you can test now your ability to spot these and I hope you haven't seen I tried to rotate through my examples but here's an example where you can see seven different photos you can see that they have seven different labels so they should represent seven different experiments but two of these photos are identical even though they have different labels and this was a paper published from a group in Canada um and I reported it to the journal in October 2015 and unfortunately this has not been addressed yet and let me show you maybe you have spotted the duplication yourself of some Sonya says and s and e2 and she is right that is correct this is a very easy one I hope you can all see that that these photos are the same now yay yes well done well done you you get an emoji award if you if you would play this game with me I play this game also on twitter and on blue sky called and I call it hashtag image forensic so you can win an emoji award and um usually the examples are a little bit harder though but um this one was fairly easy to spot I do want to say again want to stress that this is very likely an honest error I don't suspect any intention of fraud here it's inappropriate then it's worth raising but unfortunately neither the journal nor the authors replied I also posted this on pubpeer I'll talk about that later also but yeah nothing happened and of course this is not the end of the world this you know that they took the wrong photo but you know it's an error and it should have been corrected this one is a little bit harder so here's an example of a type 2 duplication we're looking at a time series of cells being treated by something and uh you know all of these photos should be slightly different although you could argue that cells over time perhaps um you know could look similar because maybe they're like floating around in the medium and so but you would not expect these photos even it's in a time series to look exactly the same or to show overlaps so in this case we're looking at overlapping repositioned images and there's several of them and this paper actually got retracted so you can perhaps already suspect that there's something bad going on here and there is so I'll I'll show you the example of the the duplications here and you can see there's a sort of mark them here with colored boxes so boxes of the same color show the exact same overlapping region and you can see it's not just an overlap or shift under the microscope it's also a rotation for example um yeah the two marked in green the two hour and eight hour in the top they overlap but there's a there's a rotation so um or a mirroring I think yeah a mirroring and so there's all kinds of repositioning going on here there's different zoom factors and so this seems very likely to have been done with the intention to mislead and the editors agreed with that so moving on to a type 3 duplication so now we're looking at one photo with duplicated elements and this one I don't know once you see it you're like oh well but maybe you can see it this one is pretty hard to see this is a trans I think a transmission electron microscopy photo of some I don't know nanoparticles or tissue it doesn't really matter what it is but there's duplicated elements here that once you see them um or maybe you don't see them but let's see let's see if we can find the answer and there it is there's lots of duplicated elements here with you know within the photo parts of this photo appear to have been duplicated and uh rotated even and and so this is very unlikely to have happened by accident so this seems very likely to have been done intentionally this was uh the duplications on the right part of the photo were first um found by Alexander Magazinov so I want to give him credit but I found all the others um and I'm using actually you know I see I see these things by eye I also use software to find it called image twin so it will find some of these duplications but some of them you just are not found by software you just have to see them especially when it's rotated a little bit these things are very hard to detect by software but yeah a lot of duplications going on uh very likely to have been done intentionally I cannot think of any technical explanation and uh here's a very extreme example these are western blots lots of duplicated elements here bands appear to have been duplicated backgrounds have been duplicated things have been spliced and moved around and you have to wonder why but I I'm not quite sure why but um this was um uh I only reported this online so far but no response from the authors and yeah I would if I was an author I wouldn't really know what to say either but um yeah how how did did this happen by accident I cannot think of any reason but yeah who knows maybe the author has a good explanation for this but we're still waiting for one and so sometimes you can find these duplications in plots so this is actually an example that is not a photo occasionally you will find these duplications in plots so here you have a plot of an nmr spectrum with peaks and noise and the noise appears to have been duplicated perhaps in an attempt to remove some of the peaks so perhaps the chemical compound wasn't as clean as the authors wanted us to believe and this paper got retracted from scientific reports and so I've done a lot of this screening on um biomedical papers and in 2016 I published this paper together with Arturo Casadoval and Farrig Feng and what we did is I screened 20 000 paper by I at that time there was no software yet so I scanned all these papers spending 20 different years 40 different journals and 14 different publishers I scanned them by I for duplications within the paper and I only scanned papers if they had at least one photo and so I uh I scanned all these papers if I found a duplication in a paper I would send the reports to my two co-authors who had to both agree if they didn't agree I took it out so we had at least it wasn't just me seeing these things both of my co-authors had to agree that a it was a duplication and b it was inappropriate and so together we sort of developed a consensus for for what we thought should be marked as a duplication so in that set of 20 000 we found 800 papers with duplicated figures within the paper I didn't scan specifically for a cross papers which is much harder so four percent of these papers contain duplications now some of these might be honest errors especially if they were just simple duplications and we made sort of an educated guess that about half of these could have been done intentionally and I think we were very mild and generous in that you know could have been more but you know let's say that about half of these were done intentionally that would mean that two percent of these papers contained at least visible intentionally misleading photos now does that mean that that's the percentage of misconduct is it two percent well it was of course only biomedical papers I focused on molecular biology papers that had photos and you know you cannot really extrapolate that to all other papers a lot of other papers in other fields might not even have photos a lot of papers would only maybe have a table and a line graph or so but alteration in so fraud let's say in other data types is actually much harder to detect as I said in the beginning you can just type in some numbers in a spreadsheet and make a nice line graph and you would never suspect that it was made up data fabricated data so it is much harder to detect duplications or even manipulations in data that are not photos and you know if a person is a good photoshopper if you move your sample under the microscope a little bit farther I would also not be able to detect it I really only detect the tip of the iceberg the visible problems and so the real percentage of misconduct in science papers has to be much higher than two percent and you know people have estimated it to be between five and ten percent and I think that's a more realistic range and it might be even higher in some fields or some journals that are targeted by fraudsters looking at the the correlation between the impact factor of these journals so I looked at 40 different journals so you see 40 different points here data points on this graph and I plotted the impact factor of the journal to against the percentage of problematic images and you see sort of a roughly it is statistically significant but you know it's only 40 papers so you could argue you know maybe it's not enough but there seems to be a negative correlation between impact factor and percentage of problematic images meaning that the higher the impact factor the lower the percent of these visible problems yeah you know again I'm really only catching the dumb fraudsters the tip of the iceberg so maybe you know screening in high impact journals such as science and nature which are the the two data points most most on on the right of the figure maybe uh these are more experienced fraudsters we don't really know that but there's all kinds of factors that could determine this negative correlation but there are some journals that do do really well with a low impact factor but there are some low impact factor journals um most notably that dot on the top you know that has 12 of problematic images which is a spanditos journal that really did very poorly and there were lots of problems in that in that paper in that journal so what you should do if you find such an image problem or any problem in a scientific paper is to report it to the journals and or to the institutions so that's what I did so back in 2015 16 when I started writing that paper I had found around 800 papers and I reported all of these to the journals five years later around 2020 I checked how many of these papers had been retracted or corrected and as you can see on the graph on the the bottom that big blue chunk of the the pie chart is all the papers that no action was taken and that's the majority so almost 66 of these papers there was no action taken after five years after reporting this to the journal and only one third had either had received a correction an expression of concern or a retraction and and so that was frustrating that you know only one third of these papers at best get get you know corrected or retracted and in some cases the correction was for what appeared to be image manipulation which was also frustrated frustrating but at least there was a correction but yeah that two thirds of these papers were not corrected that was very disturbing to me and so I started to post these things on papyr because my main goal is not to get papers retracted or to get people fired it's to warn other people that there might be a problem and so how if journals are not really responding if it takes at least five years and some of these retractions um it depends it's very variable because it's also a question how you know there's a question I see how long an average that's the process of reporting misconduct and retraction of a paper takes in my experience it is around five six seven eight years it's hard to know there's some sets that I know it seems that now journals are a bit faster in retracting but it would still at least take half a year you know usually these things take you know you of course you want some good and thorough investigation but expressions of concern are actually a nice way for a journal to mark a paper as you know there is a problem we're gonna tell the reader that there's an expression of concern and we're gonna you know do an investigation in the meantime or maybe refer it to the institution to do an investigation and then we'll take a decision and so I think that's a good tool that's not used a lot yet as you can see it's just a tiny sliver barely visible in my part chart but I wanted to have a more immediate way of alerting readers so I'm using pop here which is an online journal club as they call themselves it's run by volunteers and if you install their plugin and you do a literature shirts on the image on the right on this slide you can see how your pop peer pop med search might look like if there is a paper that has a pop here command you'll see these green banners and you can click on it and see what command has been left by other people so I feel that's the best way that people like me who scan the papers for all kinds of problems and there's many people like me doing it I do it under my full name but many people operate under pseudonyms you can see if people have left the comments and by now I have you know scanned many papers maybe even 100,000 or so I found over 7,000 papers with problems you know so far and my work has now resulted in over a thousand corrections and 999 as of today corrections so waiting for number a thousand but over a thousand retractions and almost a thousand corrections but yeah still I feel a lot of papers are not being addressed and are still just just marked on pop here and there's no action also institutions seem to vary widely in how they address these fraud these allegations of misconduct there's several articles where there are people there are professors who have dozens of papers marked on pop here with problems the first authors are all different but it's always the same lab that pops up and these people still are employed there's an article in the New York Times talking about one of these cases this person is still employed through best of my knowledge and has actually threatened to sue the person who was the whistleblower and you know there's a long case he also sued the New York Times for this particular headline and lost that case and now his lawyers are suing him because he's not paying his lawyer so there's a lot of drama a lot of fun stories behind the scenes but these people are often still employed and they don't seem to really you know be receiving any consequences for their for their misconduct and you know this image what you can see here on the bottom in the middle that's you know what appears to be heavily manipulated image but the Chinese government declared that there was no misconduct or plagiarism found even though this person had 63 papers on pop here and so you know you have to ask yourself how is this not misconduct and how is this person not responsible for what happened in in their lab now Stanford of course recently has been in the news because their president has under received allegations of misconduct not while he worked at Stanford he has also worked at Genentech and UCSF but yeah his papers are now under investigation Stanford initially wanted to investigate this case by the board of the board of trustees of which the president himself is a member and that was heavily criticized so then under public pressure they actually had an outside committee investigate his papers and so far he is now retracted three papers and he has stepped down as the president but he's still going to be active as a researcher and overseeing a research group so it's a very complex case but at least the university under public pressure decided to have this investigated by an outside committee which I feel when it involves the president himself that seems like a good outcome because yeah if the university itself would investigate it the outcome would not likely have been that he was responsible for for overseeing the research so very complex cases handled very differently by different universities there seems to be a lot of conflict of interest in investigating these cases especially when it involves not just a graduate student but suspicions of allegations in a lab where there might be a you know a bullying or very demanding professor so how do authors respond so there's a very interesting bunch of responses on pop here when you post these things sometimes the author will come back and give a response for how these cases could have possibly happened so one of the examples I just showed you shown here at the top the the author replied two days ago these similarities are entirely anticipated and are caused by a non-uniform potent applied current within the material so some some people call this on twitter somebody called this techno bebble which I thought was an excellent term like they they try to throw in a lot of explanations for what could have happened but you know this molecules don't arrange exactly in the same way every time and so it I can only think of this as digital manipulation I have looked at thousands and thousands of microscopy photos so but yeah the authors will try to have some explanation and unfortunately some editors will fall for these nonsense explanations another author sat for the photos shown here in the middle of these duplications within these overlapping images they said we have a whole tissues look alike they all have the same cell architecture so of course images look similar well they overlap so and they're representing different experiments so I have some serious questions but again the author just tried to try to minimize the problems and the bottom one was a very funny one where they said that yeah the software at that time the when you took a new photo it didn't erase the old photo it just had some parts over it that rearrange themselves in a different way and it also sounded like very technically and I think a lot of editors might fall for these explanations but um yeah I don't know I I'm just like how how on earth did these cells organize themselves in exactly the same structure so yeah I I'm very skeptical of these explanations but they're they're very funny sometimes to read um so let's switch gears a little bit um I've talked about perhaps photoshopping and and individuals who might do these things but um it's sort of the elephant in the room is artificial intelligence and I think this is this has already and will change the way we think about originality of authors and also of you know student written essays this is this is a big problem because AI can do wonderful things I'm sure but it can also create fake uh things that did not happen fake papers fake images so just as some examples you know there's there's a lot of AI chat box that will create fake information because they they will pick things up from the internet and they will tell them as truth they will create fake references and if you search for example of the text as an AI language model you will find a bunch of papers that have that text in it um and these are not papers describing chat GPT they're papers being accepted and published as real scientific papers but they're written by some chat GPT like chat box some generative AI and they are based on they're sort of plagiarized texts but because they're unique they cannot be found by the classic plagiarism detection tools and you know is this allowed or not and I think this is a complex question as a person for whom AI for a person for whom English is not my first language I would love to use AI to help me write better sentences and I can see that could be very useful for a lot of people but you don't want it to write the complete paper and include all kinds of fake things or or made up references that that don't exist so it's a very um hard to draw line what can we accept and what is not acceptable and and this is obviously also a problem in in uh academics where you know do students can they use these tools or not um is it okay as a scientist to you know if we do experiments maybe you could say it's fine for an AI chatbot to write the paper around it maybe that's not really part of science it's about experiments in the data but it's hard to draw the line here and I think this is all so new that we have not really dealt with the ethics of AI and I'm struggling with this myself because I think there's good applications and there's poor bad applications in the hands of the wrong person this could lead to all kinds of fake fake news stories and fake science papers and even more worrisome is the ability of generative AI to create fake images shown here on the left are three images that do not represent real events they're they're made by del e or some other generative AI image program so it there's AI tools that can create images of people who do not exist there's tools that can generate you know recognizable people wearing perhaps a designer coat but this is you know Pope Francis but yeah that that didn't happen and also Trump wasn't arrested by a bunch of police officers these images look very realistic there's still some things where you can recognize they're that they're fake but you know that next year these tools will even be better and so if we can create images that are recognizable of people or that can create faces of people who do not exist it's probably very easy to make AI generated photos of tissues and and cells and so some examples are shown on the right and this was a paper from almost two years ago of course now these tools are even better and it will probably be impossible to to distinguish fake from real images and this is going to be a massive problem and maybe it already has infiltrated scientific papers it's just impossible to recognize fake from real images anymore and this can be used in the hands of the wrong person it can be used to create fake papers and this brings me to the topic of paper mills so paper mills are I guess how do you call them networks of people who do who sell papers that do not exist or that are made up or are plagiarized to authors who need them so what they sell is basically the paper or you could argue they sell the authorship so they will say that you can find these advertisements on facebook very easily I found the two shown here on this image call for authorship like they have we have a paper and if you pay us you will get your name on this paper how does this work it seems that in some cases these paper mills are working together with editors to get them to get these papers published or they have an accepted paper and they add a couple of extra authors who will pay for it so they're sort of brokers of papers but in some cases these papers are generated by by some chat gpt type of AI or they're plagiarized and synonymous and there's all kinds of different scams going around so there's no no one scam fits all they're all different people coming up with the same idea like oh we can actually sell author ships and make money and it's very hard to recognize these although in some cases in the earlier years you can recognize some of these papers so just as an example this is a set of papers we recognized and I was part of a bigger team that recognized that all these papers were fake and so this we call this the tappel paper mill and they all had very similar images so these images of western blots all had the exact same background not just within the paper but across papers as well so we found a group of papers that about 600 papers that all had that same background and it seems that the bands these black stripes on it were generated you know by some some generative adversarial network GA and technology or some other more primitive form of AI and because they made the error of using all the same background we could recognize them and so these 600 papers we flagged them and and I have to give credits to Morty and smart Clyde Jennifer Byrne and Jana Christopher and many others who played a role in finding these papers I don't want to pass it off as my own but yeah credit is due to a lot of other people working on this problem so together we found 600 papers and many of these are being retracted and very often they have very similar title structures but they're published in different journals and so they were deliberately targeted to different journals and they're hard to recognize but many of these papers are being retracted and they're all in the same field of non-coding RNA some other papers that this is a set of papers I found that all appear to have the same source published in mainly in two different journals and these papers all had impossible coordinates of where plants were they all said that they were collecting plants in Iran and the plants were all connected collected at different locations but the locations were very similar across all these papers so the ones marked in red boxes all have the same location even though they represent different plants and the latitude and longitude also have very weird numbers that don't even exist because you know they're given in minutes and seconds and so if you if you have 93 seconds that seems to be impossible so these coordinates are not even possible but they're all copied pasted and used in different papers and this was also part of a citation paper mill so not only where there are probably the author ships being sold but also the citations seem to be very heavily favoring two particular authors who might have been the masterminds behind these papers and so there's all kinds of weird things problems that you could find once you are open to them um maybe you can spot the problem in this particular table and so there's um in this table um this is also part of a paper mill and why do we think that because they all had very similar tables and text but the cancers differed the non-coding RNA numbers differed but you can see there's a big problem because this is a paper about prostate cancer and look at the gender distribution half of the patients were female which is very unexpected with prostate cancer so we believe these are papers that have the numbers are changed but in this case they forgot to take out the the females and this is how you can recognize that these papers are often written on a template and some of the things are changed so it's hard to really find that these papers are all based on the same template and here's an example of a paper mill that you know another group of spammers that scammers that use what is called tortured phrases basically these are plagiarized papers but the text is hard to recognize it's not exactly the same what they do is they take text written by others and they synonymize all the expressions so they have some some way of sort of translating sometimes it's translating into another language and translating back into English sometimes it's synonymizing a lot of the expressions so in this particular paper it's about bosomic lignancy you're like what is bosomic malignancy and then you you know a sentence that you can see here in the abstract is that chest peril is a remarkable kind taking all into account in basic parts of ladies room worldwide well that's what does that mean well and then if you find the original it's actually about breast cancer and it's one of the main reasons for women's deaths globally and so you can see these synonymized text is sometimes funny it doesn't make sense at all and but you're cannot quite put your finger on it but so Guillaume Cabernac is one of the main people who have discovered these types of paper mill writing style and so these are there's massive amounts of these papers and he's trying to infantize that and there's a whole database of them that he wants to add to pop here and hopefully get retracted because these are all you know plagiarized text but it's hard to find the original text actually it's a lot of puzzling to try to find these yeah bosom peril is breast cancer or chest peril and it's it's really funny there's also counterfeit consciousness which actually is artificial intelligence and so he has Guillaume Cabernac has a whole database of these synonymized what he called tortured phrases and once you recognize them you can use them as a hook to find more of these papers and then some of these paper mills are actually targeting special issues so special issues are a sort of a money making way for a lot of publishers to make extra money so what they do is they assign a guest editor who can then look at papers and you know publish papers so usually they will not be very experienced and in some cases it appears that the guest editors of these special issues are working together with the paper mills perhaps receiving a kickback of the income that the paper mills make they might actually use that to buy in one of the editors the guest editors and so there was a massive problem recognized at Hindavi but also mdpi is one of those open access publishers that have a lot of these a lot of these special issues that appear to have been targeted by paper mill articles and if you look at Hindavi so there was this Hindavi has been bought by Wiley there was an announcement earlier this year in retraction watch that they were going to retract 1200 more papers but i just did an inventoryization of the number of papers that they retracted this year alone Hindavi has retracted almost 5000 papers and you know we're the year is not over yet so there's a massive amount of retractions this year from Hindavi special there are all special issues it was believed that a lot of these um yeah special issues were run by guest editors who either did not pay attention to the pay review process or who were in the loop and just receiving kickback from paper mills and so these are just massively being retracted and uh yeah there's there's probably more to follow but it's uh it's a very big problem and at least Wiley has recognized that and and yeah that some of these hidden a lot of these Hindavi articles are are not very useful and you know there might be some real papers in there but a lot of those appear to have been targeted by paper mills and mdpi unfortunately has not really seemed to have recognized this problem but is massively inundated with these papers as well so this brings me to my last slide um i've talked about what my personal belief is that science is about discovering the truth and that science misconduct is everything that science is not you know it's it's everything that science should not be science misconduct is about deliberately changing the outcome of a science experiment and yeah that's not what we should call science there should be more consequences because unfortunately the rewards of fraud are high uh the consequences are low and so there's a disbalance if the chance of getting caught is low you know imagine we would all be speeding we would all be crossing red lights because we would never get a ticket and i feel rules are there to be enforced without enforcement people are gonna yeah just see the rewards and will do fraud we focus probably way too much on quantity while we should focus more in science on reproducibility also it takes a village not just the role of people like me but reviewers journals institutions funders we should all care about research integrity and it seems that there's a growing um responsibility that is being taken but a couple of years ago there seems to be very little um very little action being taken if allegations about misconduct were raised so we need faster correction because that is the only thing that can serve the other readers and the other scientists reading these papers and also we need better legal protection for those who raise concerns there has been a very recent case where a researcher at Harvard was being accused of misconduct was even being put on leave and a couple of her papers were going to be retracted but she is now suing both her university as well as the people who raise the concerns for a large amount of money which these researchers do not have so you know you can be right to raise concerns but in the US at least if you're being sued you have to pay your an attorney to defend yourself and that will cost you you know perhaps $250,000 and so you know most people don't have that type of money so how do we legally protect people who just raise concerns in an objective way so that's a big problem um because you know legal this uh scientific discussions should not be held in the courtroom and then finally there's a tremendous cost of misconduct I mean people trying to replicate science papers that's one big problem but for science as a whole for you know bringing the deliberate false narratives into the public eye I think that's that's a big problem and we've all seen the consequences of that and fewer and fewer people seem to believe in science and unfortunately you could walk away from my talk thinking that all science is flawed and I hope you're not like we you know there's problems in science but hopefully we can do better so with that thank you and I'll be happy to take any questions oh thank you so much for this amazing webinar always a pleasure to listen to I think we have a question in the chat from a colleague uh I found 75 questions in citation review reports made by nine reviewers all of them co-authors themselves in about 20 journals all of them from the same publisher would you name it as a reviewer's ring I'm not sure what co-visitive means but um if all the reports have similar language yes that could be a reviewer's ring I mean I've seen cases where of course I don't even know what that means I don't even know what that means sorry but uh yeah I mean if it's all the same language and very short peer reviews um you know all have very similar wording uh yes that could be that could be a very big sign of a peer review ring uh in many cases we don't really know who the peer reviewers are for for journal but um journals like um Frontiers and MDPI actually sometimes name the peer reviewers and then if you look into who these peer reviewers are you can see they actually work together often with the papers they the authors of the paper they have reviewed and so sometimes you can find them but very often you don't know who the peer reviewers are but people have looked at wording of um scientific uh reviews and found peer review rings with that because yeah you would never use the same language it's similar as fake peer reviews on Amazon uh these things are often being misused and a lot of these special issues also have probably not been peer reviewed very well but it's you know if if you if you find something please post it on pop here yes wording is similar and a reviewer's asking for adding references yes that's a very um doesn't necessarily mean that it's a peer review ring but at least the reviewer who is very you know on the lookout of their own papers and that should you can sort of say maybe one paper you can put in but if they ask you to put in five papers of their themselves yes that a good editor should should refuse that i should immediately kick out that peer reviewer but yeah there's no consequences for these things there's people are just too naive unfortunately uh we have another question what do we do if you find scientific discontent yeah so the the real officially you should report it to the journal and you should so that you have to find the journal editor and write them a um an email and and you know making sure you forward your concerns you can if it's multiple cases of the same person i would also write to the journal to the institution or to the funder but the writing to the journal editor is the main thing that you should do but as i've shown unfortunately in many cases the journal editors do not seem to be willing to respond and so i would also put it on papier you know be objective create an account there you can create an account completely anonymously and you can uh but you have to be objective you cannot say uh you know that professor is a fraud you can but you can say this image looks remarkably similar to that image or this peer review is the same the wording is the same as in those 10 other papers or this uh you know you can see that that suddenly 10 papers were added to this to the list of references that have no real connection to the paper itself and so you can ask yourself why where you know why were these inserted so whatever is objective and uh worded yeah objectively that could end up on papier so the time is almost up and i would really like to thank you oh there is one more yeah that's the same person yeah the first step is to write to the corresponding author they never applied then the journals they also don't reply what do we do after that yeah i mean i'm facing the same problem and i'm you know hopefully a you know i've earned hopefully my my stripes in this field they very often the journals also don't respond so now i ask i usually write like i'm keeping track of which all with which journals do respond and i will post about this sort of as a warning like i'm keeping track of you if you respond to me so most journals nowadays will say okay we're looking to this but then you know you never hear anything and you know even some of the retractions that are featured in retraction watch i've written to the journals and i just hear it through retraction watch they didn't write back to me even though that is a cope guidelines but it's um yeah it is a problem um and i cannot really tell you what to do other than to post it on papier i feel that's my civil duty to to uh uh yeah to to report it there um so there's also a question about my email sure i'll be happy to shut up put it in the chat yes please do and during the time uh we have another question i'm not guaranteeing that i will answer you because i get a lot of requests it's elizbeck at gmail.com so i put it here on the webinar chat i hope everybody can see that yes and one more question if you had some advice to give to journals and how to address ai and the issues it brings to image manipulation what would it be it's a tough question because i'm not a computer scientist and i would not really know how we can nowadays distinguish real photos from ai photos um and so we can think about maybe having you know all ai tools have a um an imprint that says this is ai generated but i'm sure this could be removed um you can also think about the other way like that all microscopy photos or all scanners have some digital imprint that it's a real you know generated by a size microscope on that and that day at that location or something like that uh that could also be tampered with but um so i'm not actually quite sure because i think we're at a stage we can no longer distinguish you know footage or photos from uh that were generated from real ones like you know if you think about Jurassic Park the dinosaurs look pretty realistic now right like we know it's fake but man that looks very good yes i don't i i'm i'm at a loss actually i'm a bit pessimistic also oh one more question have you faced criticism for your work in reporting this misconduct yes i have faced a lot of criticism i have a loyal team of people on twitter who harass me uh almost daily um i have been threatened with lawsuit by some professors it hasn't come to a real lawsuit yet but the the data colada um uh lawsuit that i talked about the harvard professor who's suing those three professors criticizing her work that has uh given everybody pause because it will hinder it will uh it's sort of meant to silence critics and i feel that's not the root we could go so i'll keep on doing what i'll do but i hope people will support me if i ever been and being sued um and so far the data colada team is you know we organized the fundraiser and we organized you know we got 250 thousand dollars in 48 hours so we have that money for to help them defend themselves but um it's it's tough so yeah i have people calling me a failed scientist and an ugly fat pig like every day and death threats and so but yeah it you sort of get used to that but i do cry because some of these uh insults are really bad but yeah and i'm so sorry to hate that and i want you to know you have an enormous support i do thank you as well and you are a hero to so many of us and role models so you have to know that as well i know that yes yes so and we we are all hoping that you are going to keep up with this i will i will keep on doing what i do but if i have to cough up 250 thousand dollars to defend myself i do hope there's some you know everybody donates a dollar and i'll i'm almost halfway there yeah thank you so much thank you so much it's such a pleasure having your hair it was an outstanding talk as always and keep up with your go good work we are really grateful that you could be with us today thank you so much thank you sonja for inviting me and my pleasure to be here thank you bye everyone bye everyone see you next time in a month okay bye bye