Reaction roll genocide
A paper in a well-known journal from a major publisher on the subject of using nanogels for wound healing is packed with bizarre phrases that have nothing to do with the paper. It came to light this week on BlueSky via research integrity expert Leonid Schneider of For Better Science.
A quick scan of the Abstract seems fine…no real red flags in the Introduction either, but deeper into the Methods section is where the paper gets very weird, with nonsensical phrases appearing seemingly randomly:
“flag to commotion ratio”
and
“A phosphate safeguard answer was used to regulate the pH to 7.8, influencing the total response book to 30mL.”
and
“Subsequently, 1mL of the mass killing of an ethnic group was opposed to 20mL of the skin sample and unprotected to light for 7 min.”
This looks like an AI fail or maybe a really bad translator, or perhaps an attempt to embellish simple language to make it sound more important than it is. It could, however, simply be phrase swaps to mask something far worse. Hubert Wojtasek, Associate Professor of Biochemistry at Opole University in Poland, told me he had highlighted dozens of problems with this paper having been alerted to its existence by a third party.
I’ve had a look and have translated some of the phrases myself.
A technical word like “buffer”, which is pertinent to the acidity or alkalinity of a solution, has been switched for a non-science synonym “safeguard” rendering the phrase meaningless, even though a buffer in a lay setting is a kind of safeguard.
The word “solution” in this section was swapped for “answer”, which makes no sense.
The word “book”, well that was obviously “volume” originally and again makes the whole phrase nonsensical as it stands.
I think that sentence should read as follows:
“A phosphate buffer solution was used to control the pH to 7.8, changing the total reaction volume to 30mL.”
Another BlueSky user chemist Nessa Carson mentioned a quite amusing one phrase “flag-tocommotion”, which was is obviously “signal-to-noise” ratio!
What of the sentence with “mass killing of an ethnic group” and “unprotected to light”? What was that originally?
Well, let’s consider unprotected, the most obvious word that could have been is “exposed”. It was “exposed to light” rather than left “unprotected to light”.
But, the “mass killing”? How can that be in a paper about hydrogels and wound healing? Nobody seemed to be able to work out in the BlueSky thread.
It took me a few moments of thinking around it all, then it suddenly hit me. It was like finding the answer to a cryptic crossword clue. What that murderous phrase actually meant to say is something very sinister:
“1 mL of the final solution”
The AI or translator has presumably looked for a synonym phrase for “final solution” in some non-scientific phrasebook and found a very inappropriate reference. In a scientific research paper, the “final solution” is just the contents of the reaction flask once the chemistry is complete, but in this alternative reality, the final solution is the Holocaust! This is horrendous.
The original phrase before obfuscation must have been something like this:
“Subsequently, 1mL of the final solution was added to 20mL of the skin sample solution and exposed to light for 7 min.”
One has to assume that no one in the editorial office or among the referees or reviewers noticed these “tortured phrases”, and certainly no one among the authors, one of whom is listed in the paper as being responsible for “validation”, noticed either. The reason for this is that nobody could have actually read the paper at all, at any point. If they had, they would’ve pressed pause on its publication and confronted the authors. In fact, to put it even more bluntly, not only did nobody read it, it’s possible that nobody even wrote it, well, at least not those bits in the experimental section.
This paper’s figures also have some worrying characteristics, very obvious textual errors that older AI tools might make, for instance. There are non-words like “grarually” instead of “gradually” in one figure. And a wholly nonsensical string of letters, “magrauick graauad”, is a label for one of the reaction flasks in a diagram. In addition, the colours in the image don’t seem to match how they are described in the caption. Something described as orange is green, for instance.
The images were not looked at nor checked either, by anyone.
Wojtasek suggested to me that “Nowadays most papers are not read by anybody, either before or after publication. After all, reading is a completely ineffective activity, publishing papers isn’t.” He added that he has been fighting this kind of junk for almost two years and pointed out that Schneider has been battling against it for more than a decade.
Automated tools are being used to assimilate words and phrases from other sources and cobble these together to generate new papers for submission to journals. It might be that these inappropriate modifications of some stock scientific phrases are being made simply to mask the true origins of the text.
I asked the scientist who had first spotted this particular paper about it. What they told me, off-the-record, is a rather sorry tale of the fraudulent nature of these works and of the character of the authors and wider section of the scientific community that sees this kind of misconduct as somehow acceptable. Researchers are using slop to boost their careers.
None of this is really a new problem, it just seems to be the wildest of the most recent examples especially with that “final solution” swap. Indeed, this 2021 paper from Guillaume Cabanac of the University of Toulouse and colleagues, discusses the problem of tortured phrases and how it had been happening for at least ten years back then.
Papers over the years, they write, have had bizarre phrases like “subterranean entomological nation” instead of “ant colony”, and ironically enough, “counterfeit consciousness” where what is meant was “artificial intelligence”. It stands out a mile, it’s obviously been changed by someone or something with no understanding of the phrases nor their context.
I’ve personally seen the phrase “immense information” used to replace “big data” and maybe I imagined it, but I am sure I’ve also seen “mist calculations” when they actually meant “cloud computing”.
Fraudulent research papers, paper mill articles of which there are thousands and thousands now, with their tortured phrases and fake figures, are a major problem for science and for honest authors and publishers. They fill the literature and academic CVs with nonsense that somehow gets them a new job where they can stir the slop pot even more. Often they are called to account, but there are now so many that simply never get flagged and are added to the literature and CVs without anyone checking or really noticing. There are some infamous cases, where a publisher has been embarrassed into retracting thousands of papers from dubious journals it has acquired in a takeover. But, this is the tip of the proverbial frozen marine aqueous mountain [iceberg].
It is all only likely to get worse as AI gets more and more sophisticated. While the “final solution” paper is easily spotted, the AIs that some scientific fraudsters are now using are much more subtle and perhaps undetectable. We can but hope that such papers are detected earlier, but at best what happens is that they are retracted after they have already done their worst.
When I think about how much effort we put in when I was in the academic publishing industry back in the late 80s and early 90s, this kind of activity makes me feel sick to my stomach. We would receive and log paper manuscripts, decide on two referees to either accept or reject the paper or more often than not ask for revisions prior to acceptance. We sometimes had to deal with a third referees, and in extreme cases, the academic editorial board members, or the academic editors themselves!
We’d then edit the endless stream of accepted papers with a pen on paper and once complete send the edited manuscripts off to be typeset. We’d receive paper proofs back from the typesetters some time later (they would too would have checked the proofs for formating problems). We then read and corrected, again with pen and ink. The first proofs were also sent to the authors for their corrections.
Once all proof corrections were complete, the printers would send us a set of second proofs of the complete issue of the journal for a final check. They would be read by the boss, the technical (as opposed to academic) editor. Any minor mistakes or major problems would be taken into account at this point before the papers were printed and bound in the journals and sent to libraries and subscribers.
None of that happens now at any but the most fastidious publisher, as far as I can tell, at least given the terrible typos and mistakes that I see in papers from so many different sources. I suspect that most publishers perhaps do a cursory on-screen edit, but that’s definitely not happening for some papers and even whole journals. It’s scandalous.
Remind me again how much journals now charge for subscriptions to cover their purported costs of peer review, editorial, and gatekeeping?
Publishers everywhere take heed. Your authors are creating slop, your perhaps overworked referees and editors are not even bothering to wade through this slop, and you’re charging readers huge amounts of money for this slop.
The slop is soon going to hit the fan, and you will hopefully notice that when I write slop that word itself is almost a tortured phrase where the true meaning is a much more guttural and familiar term. It’s all SH*T!
Footnote
I have not shared the DOI or the full reference to the final solutiono paper to avoid giving it unwarranted clickthroughs. I have a reprint of the paper, you can contact me email or DM on social media if you wish to read it.
Not all tortured phrases are of this dubious nature. We in science and sci-comms all know the bizarre reference to vegetative electron microscopy, it’s not a thing, the word “vegetative” and the phrase “electron microscopy” were present in a single paper on two adjacent pages and were somehow inadvertently conflated when the document was scanned. There was no editor to notice, presumably.
And then there was the case of the abbreviated Copper Nanotubes, which gives us an NSFW acronym akin to carbon nanotubes’ CNT, but remember the symbol for copper is Cu. I don’t need to spell it out, do I?