53
votes
A nonsense phrase has been occurring in scientific papers, suggesting artificial intelligence data contamination
Link information
This data is scraped automatically and may be incorrect.
- Title
- A Strange Phrase Keeps Turning Up in Scientific Papers, But Why?
- Authors
- The Conversation
- Published
- Apr 25 2025
- Word count
- 1035 words
The term is vegetative electron microscopy
Huh. To an extent, I get why some scientists might use AI to aid in writing a paper because writing is a different skillset from whatever they're researching. Still, to not even proofread it? If this nonsense term keeps slipping through, that raises serious questions about the accuracy of the rest of the papers' content... and also the accuracy of papers without a telltale clue it was AI-written like that. Because the scientists who submitted those papers clearly aren't checking it that closely.
The pressures created by the publish or perish culture, and especially how it connects with funding opportunities, has been a problem in various corners of the academia for a long time now and it has, predictably, resulted in paper mills and citation rings, as well as just generally a lot of junk research that's done only because some quota somewhere mandates that you put something out with your name on it. It's a sad reality that a lot of what is published is put out there not to further our understanding of something, but to game the algorithm, so to speak.
Ironically, I wouldn't rule out this helping combatting that toxic aspect of academia. It slowly creeps enshitification in scientific papers. And while institutions may be oriented towards the publish or perish culture now, it's not like scientific progress does. Meaning it may have to shift institutions away or risk reputational damage. It also brings out some of the bad cultures of academia out in broad daylight more than ever.
It also implicates the review boards of journals. If they're just skimming the abstract and noting the status of the authors, they aren't fulfilling their very important function of ensuring validity through peer review.
Bear in mind English is the main science publication language worldwide. Even when publishing in local journals in one's native language, English abstracts and maybe captions are often included.
There was an entire little industry of academic editing, often simple proofreading to fix language errors, and i know the work dried up extremely quickly in the wake of chatgpt becoming popular.
These misses may be a language barrier issue in top of looking for a way to save precious funding dollars (by not having to hire an editor).
I think this is maybe a bigger part of it than people are realising. I come across a decent number of GitHub repos for research code that have Chinese language docs as the primary info and relatively shaky English in the issues, but the associated papers are always in polished academic English - because apart from anything else, they’re required to be.
Obviously I’m making a few fairly big leaps here, but if I were in their shoes and had to publish in something other than my first language with a particular expectation of style and tone I’d absolutely be looking at LLMs to help out with that.
Pretty sure I saw a reddit thread about this exact phrase, where some users figured out that the phrase is easily explained as a mistranslation from Persian. Basically, remove a single dot in the Persian phrase, and this is the proper translation of slightly misread Persian.
This, to me, is the real problem here. However your paper is written, not even proofreading a paper before you publish it is more than irresponible and lazy; it's actively harmful to the scientific endevour.