A publisher is retracting five papers from one of its conference series after discovering what it says was “clear evidence” that the articles were generated by a computer.
The five papers were published from 2018 to 2020 in IOP Publishing’s “Conference Series: Earth and Environmental Science.” According to an IOP spokesperson, the retraction notices will all read:
This article has been retracted by IOP Publishing in light of clear evidence that it was computer generated. IOP Publishing is investigating why this was not identified during the submission and peer review process by the conference. As a member of the Committee for Publication Ethics (COPE) this has been investigated in accordance with COPE guidelines and it was agreed the article should be retracted.
Here’s the abstract for one of the papers, “Neural Networks Considered Harmful:”
System administrators agree that omniscient modalities are an interesting new topic in the field of complexity theory, and researchers concur. Given the current status of lossless information, experts shockingly desire the visualization of model checking. In this position paper, we use biomorphic information to disconfirm that the little-known random algorithm for the simulation of journaling file systems by Van Jacobson is optimal.
It and the other four papers — “Financial Information Security Using Modular Communication,” “Singular Topoi of Countably Non-local, Continuously Cayley, Maximal Elements and the Continuity of Closed Elements,” “Newton Categories for a Solvable Element,” and “Decoupling Evolutionary Programming from Gigabit Switches in Neural Networks” — show telltale signs of being created by SciGen, “an automatic CS paper generator” developed by graduate students at MIT in 2005, or Mathgen and Physgen, which are similar.
Indeed, Kim Eggleton, research integrity and inclusion manager at IOP Publishing, tells us that initial findings from our investigation indicate that some of the papers were generated by SciGen, others by MathGen.” Eggleton says:
We were alerted to the papers by an external whistleblower and immediately began investigating the papers. We’re very grateful to Cyril Labbé, the creator of the SciDetect software, who has been advising on the legitimacy of these papers.
The investigation into these papers is still ongoing, and part of that investigation is understanding the legitimacy of the authors. What’s not in doubt is that the papers are not reflective of real research, and therefore it’s important for readers that we retract as soon as possible. Once the investigations are completed we may update the retraction notices with further information relevant to readers.
In 2014, Springer Nature and IEEE retracted more than 100 papers for the same reason. The following year, Springer Nature made Labbé’s software to detect such “gibberish” freely available.
None of the authors listed on the IOP Publishing papers, all of whom have affiliations in China, immediately responded to a request for comment, with one prompting an out-of-office holiday message.
In September, IOP Publishing retracted more than 20 papers by an author who plagiarized and used an alias.
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on Twitter, like us on Facebook, 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].
The “Neural Networks Considered Harmful” paper contains some really amusing parts – far funnier than what SciGen usually puts out:
“The question is, will HUD satisfy all of these assumptions? Yes, but only in theory. ”
“Reality aside, we would like to harness a methodology for how HUD might behave in theory.”
“We consider an approach consisting of n RPCs. ”
Did they get lucky, did SciGen improve, or did the authors edit the result to make it funnier?
Also, section 2.1 seems to just be some formatting guidelines from a call for papers.
If you wish, you may have a look at the source config of SCIgen:
https://github.com/strib/scigen/blob/master/scirules.in
It appears to me that the authors just got lucky 😉
The paper also mentions “SCSI disks” in the conclusion. Back in the 1990s they were a big thing. I think I still have one tied to my old Emulator Systems Emax II somewhere. I doubt that a lot of folks nowadays even know what a SCSI disk is. Or was.
Scuzzi drives are still a thing, though not typically in consumer level machines. They are mainly used in higher end workstation machines where speed is needed rather than capacity.
As far as the two mathematics papers (Singular topoi and Newton categories) are concerned, the primary reason for retraction is surely that they are gibberish, and gibberish which would have been detected the moment a competent person had seen the titles, let alone the contents. We have to presume that no such person looked at the papers at all. I hope that the publisher will investigate how that can have happened.
Re: “gibberish which would have been detected the moment a competent person had seen the titles, let alone the contents”. And then there are the References! My (competent) eye was drawn immediately to Z. Kobayashi, Non-Linear Knot Theory, purportedly from Springer (in 1954!). A Google search reveals at least half a dozen obviously fake publications with “non-linear knot theory” in their bibliographies. What a world.
That, it would seem, is the main point. Goof heavens, all but one of my papers had at least 1, usually 2-4, rejections.
I took a some text “amphibious” technology sledgehammer into a scalpel” from the paper “Decoupling Evolutionary Programming from Gigabit Switches in Neural Networks” and a simple google search found known SciGen papers but also ones not retracted like this:
http://paper.ijcsns.org/07_book/html/200605/200605111.html
http://ijeasrt.com/june-issue.html article 3 but I assume the rest are fake too
https://www.computer.org/csdl/proceedings-article/wgec/2008/3334a491/12OmNyp9Mht
https://www.ijariit.com/manuscripts/v1i2/V1I2-1152.pdf
and many more.
There are some other great lines in this paper to search too e.g.
“We scripted a trace, over the course of several years, disproving that our methodology is solidly grounded in reality.”
“Our evaluation method holds surprising results for patient reader.”
“The results come from only 7 trial runs, and were not reproducible.”
And these other quotes pull up other fake papers e.g. http://medcraveonline.com/MOJBOC/MOJBOC-02-00054.pdf
The IEEE article (the one whose abstract is hosted at computer.org) is actually retracted, here’s the link:
http://blogs.nature.com/news/files/2014/02/IEEE-wiped-articles.pdf
Most examples are from predatory publishers, but there are some exceptions (for those exceptions, look up “mathgen” / “scigen” / “computer generated” on PubPeer).
Saw that too and made me chuckle. Made me also wonder whether the whole thing isn’t really so much a paper mill issue but another sting operation by HSU (humurous scientists united).
The five examples are clearly bogus – even I have seen that although I’m not from this field. But I could feed the machine with a better (and more narrow) selection of papers as food… Combined with the next topic of rogue editors I wonder how many of the 100 000+ papers on Covid 19 from the last year are milled rather than worked on.