Remember those German academics who developed a fraud detector that actually did nothing but look at whether the authors used institutional email addresses, had international collaborators, and were affiliated with a hospital?
And then they concluded, based on this "detector", that a high fraction of the papers out of Asia and the Global South are fraudulent?
And then they got a glowing writeup in Science?
(If you missed it: https://fediscience.org/@ct_bergstrom/110357259338364341)
Turns out there's data on all this.
Easily accessible data, in fact, of the sort that one might say look at if one were going to use these criteria to make extreme claims about fraud coming out of particular nations.
Let's start with institutional email addresses.
Lo and behold, the countries flagged by the "detector" as the worse perpetrators have long been known to be places where the convention is not to use institutional email.
But wait, why would those academics not use their institutional addresses unless they had something to hide?
Lots of reasons, it turns out.
But what about international collaborations? That's part of the detector as well.
Unsurprisingly, there's a ton of data on this.
Ganzi et al (2012), for example, have data from 2000-2009.
Those data too old for you? Turns out the NSF has a detailed 2021 report that covers, among many other things, the same issue: https://ncses.nsf.gov/pubs/nsb20214/data/table/SPBS-35
Based on those data, are a number of countries in order of international collaboration fraction, lowest to highest.
Again, the countries that the authors claim are producing tons are fraud are countries that we know, in advance, have lower rates of international coauthorship.
Yet the authors of this paper concluded, based on their detector, that Russia, Turkey, China, Egypt, and India have the highest rate of potentially fake papers.
Yep, look where these countries fall on the tables I've provided.
Not even checking the existing bibliometric literature for how their classifier might be biased the attribution of blame — it's astoundingly shoddy scholarship.
To do that and then proceed despite an uncorrected false rate over a third? It defies imagination.
@ct_bergstrom many of my German, Swiss and Dutch colleagues, especially those in editorial roles, are very concerned with "paper mills", which they think exist for hire to influence citation measures. If those exist, the best way to get rid of them would be to not use citation measures, or at least reduce their importance, but that is the opposite of what European funding agencies, departments, etc are doing.
