Wikimedia Commons
A gibberish phrase that caught the attention of science sleuths after it slipped into several journals might trace its origin to a typo in Farsi rather than questionable use of AI, as we reported earlier this month.
Nearly two dozen scientific papers, including some in journals from major publishers, mysteriously refer to “vegetative electron microscopy” or “vegetative electron microscope.” As we wrote in our previous story, sleuth Alexander Magazinov speculated on PubPeer “the phrase could have originated through faulty digital processing of a two-column article from 1959 in which the word ‘vegetative’ appeared in the left column directly opposite ‘electron microscopy’ in the right.”
Most of the articles containing the strange wording included authors from Iran. Magazinov told us perhaps an AI model had picked up the phrase from the 1959 article and spit it back into machine-generated text that was later plagiarized in other papers by the same Iranian network of fraudsters.
After our story ran, however, a commenter suggested on Reddit the phrase might have a different origin:
According to Google translation, “scanning electron microscopy” in Persian is “mikroskop elektroni robeshi”, while “vegetative electron microscopy” is “mikroskop elektroni royashi”. They are only differed by a point in the Persian script:
میکروسکوپ الکترونی روبشی
vs.
میکروسکوپ الکترونی رویشی
Three Iranian scientists asked to comment on this potential explanation all found it plausible. Cina Foroutan-Nejad, a chemist at the Polish Academy of Sciences, told us:
The [Reddit commenter] is absolutely right. I have heard that some Iranians write their papers in Persian and ask an agency to translate it to English. So, the translator could have misread “روبشی” as “رويشی”. Just a single dot changes B to Y in Persian and the guy who most probably had no science background translated that to vegetative that is a translation of روبشی
Alternatively, the writer of the Farsi manuscript could have made a simple typo changing “scanning” to “vegetative,” because the two letters ب and یـ in the Farsi script are adjacent to each other on the keyboard, Foroutan-Nejad said.
The latter scenario would also be compatible with cases in which authors refer to both “vegetative” and “scanning” in the same paper, although other explanations are possible, too.
Foroutan-Nejad also said reuse of text were frequent in Iran, which could have perpetuated the use of the term:
Another problem with many Iranian teams is that they literally have a “template” for publication. If you publish the same type of paper over and over again on different molecules, biological samples, or materials, you can hypothetically have a template in which you can replace certain names etc. using [the Find and Replace function]! I have seen this and I am not joking at all!
Magazinov was quick to acknowledge the possibility of the new potential explanations. However, he said he felt “absolutely sure” or “quite sure,” depending on the case, that in three specific papers to which our earlier story linked – two coauthored by controversial Spanish chemist Rafael Luque and one by Elsevier editor Vijay Kumar Thakur – the appearance of the strange phrase was the result of copy-and-pasting from other papers or using generative AI.
Computer scientist Guillaume Cabanac reaffirmed the phrase remains a fingerprint in the Problematic Paper Screener, a tool he developed that combs through the scientific literature looking for various types of problematic content such as “tortured phrases.”
“A paper is listed in the Tortured tab under ‘potentially problematic papers’ when 5+ tortured phrases are found in the paper — or when the paper is also retracted,” Cabanac told us by email.
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on X or Bluesky, like us on Facebook, follow us on LinkedIn, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].
Many occurrences of “vegetative electron microscop(e/y)” are from 5+ years ago and pre-date generative AI.
The article notes it wasn’t due to AI, but a typo:
“might trace its origin to a typo in Farsi rather than questionable use of AI, as we reported earlier this month.”
Non-AI translation tools (really, non LLM tools) could have made this mistake before 2020. So could humans.a
I think it might be a translation problem. However, I do not understand what is the problem with using AI at all. AI is a tool. I believe that people are free to use it as long as they do not fabricate making sense data with that. I mean if you are not native you can give your text to it and polish it to make it more readable. Take this as an example, if someone said to you, you were not allowed to use the calculator. So, it is ridiculous. Any innovation has both pros and cons. It can be abused and use in useful ways. Another example is that, in my master research, I had to write a long code (14k lines), then I had to schedule a time for running that on super computer of the university. But, many times I got an error and spend days and weeks for that bug, just for a “;”. However, if I worked now, I could give my code to AI to be recheck for possible errors. Did you believe that made my code garbage? Definitely, no. It saved my time ans helped me. I do not understand some people think what makes life easier is forbidden. Please think a little bit open minded. As long as it is not used for fabricating data, AI is useful tool for scientist. Even, for better science group are active now thank to Dimension AI. Do you believe they can find the patterns of paper mills without AI? No, for tracking each person, it took one month. But, they track a person in just one day.
A shovel is a tool, too. Not that I am open-minded to its use in surgeries.
Well said.
Frankly, this is not a coherent comment. AI is not a shovel, and writing a scientific paper is not a surgery. Cute, but reminiscent of Scalia’s broccoli argument. It would be more useful if you provide a specific comment why you oppose use of AI in scientific writing.
Not expressing support for either argument, just pointing out that zingers should not be used in lieu of substance.
This sleuthing was a pleasure to read — it’s an example of the fairness that results from open-minded and rigorous investigation rather than trolling. A lesson for the wider world.
I was going to post a similar thought: this type of sleuthing is very interesting to read and is one of the things I particularly like about RW. The different skills people use to understand issues is fascinating.
It would seem if the phrase turns up in Iranian journals and the like it could be honest error, but if it turns up elsewhere its diagnostic of AI slop.
Hmm. Interesting that the larger lesson here isn’t “contact a native language speaker before making broad plagiarism accusations against numerous authors.” Especially when there is a corresponding political narrative demonizing people from those same countries. It’s almost like there’s a lesson on bias and prejudice, just out of reach…