Hacker News Reader

Comment by nl | Hacker News Reader

It's kind of bizarre that this article focuses on the plagiarism issue, but seems to miss the even bigger problem: It doesn't appear at all clear if these published studies are actually done!

If they aren't done and yet are published in fairly high impact journals, presumably people are relying on their conclusions. Even worse, increasingly automated systems are relying on journal articles for automated diagnosis.

OTOH, if the studies actually are done... I guess it's a new avenue for funding science!

View on HN

> Even worse, increasingly automated systems are relying on journal articles for automated diagnosis.

Where can I learn more about this? Without hearing more, it sounds like a terrible idea.

As a scientist, if my beliefs were based on an unweighted sample of claims made in journal articles, I can tell you, I would believe a lot of nonsense. I hope no one is practicing medicine that way.

On mobile ATM, but IBM Watson does it. It is generally more accurate than humans, and able to keep up with latest research much better.

"It's kind of bizarre that this article focuses on the plagiarism issue, but seems to miss the even bigger problem: It doesn't appear at all clear if these published studies are actually done!"

In fairness, I didn't get the impression that the article was "focused" on the plagiarism. Rather, it described using the plagiarism as a method to uncover dubious papers. If you perform a search for idiosyncratic phrases, such as the ones mentioned in the article, you turn up papers likely to have been faked entirely. The plagiarism is a giveaway. The more important point is that the studies themselves are phony -- and are being bought and sold into prestigious journals.

But that's the point. Authorship is being sold, but there is no indication for or against on whether the studies are actually being done by somebody. Is it possible that the people doing the actual research are simply not getting the credit? Maybe only some snippets are plagiarised and not the entire study?

"Authorship is being sold, but there is no indication for or against on whether the studies are actually being done by somebody."

I inferred that the studies are entirely bogus, i.e., no actual labwork is being done. I feel you can make this inference pretty safely from, for instance, the fact that the phony papers reference a nonexistent statistical methodology. If real work is being conducted, why bother citing a fictitious methodology?

I could be wrong, but I felt that soup-to-nuts fraud was strongly implied in the piece.

These are meta-analyses and as such don't require lab work by definition. My impression was that somebody has created some useful tools or amassed enough experience to do this kind of work efficiently.

Fair point about "labwork," but even putting that aside, it doesn't sound like any legitimate science is being done, one way or the other. I doubt they're even doing any meta-analysis. Sounds like they're just playing buzzword bingo, and citing previous studies to make it seem as if there's a legitimate basis for their "study." If you'll notice, most of the phony papers cited failed to establish any significant findings in their putative meta-analyses. If one were going to fake whole-cloth studies, it seems that this is the safest and easiest way to do it. For example: I find 20 previously published papers on a given topic, I claim I did a "Begger's funnel" meta-analysis of them, and I claim the Begger's funnel analysis found no statistically significant results. Done. I have a new paper to publish, from essentially nothing.

The giveaway, to me, is the fake statistical method that's repeatedly cited (in this case, the "Begger's funnel plot"). A legitimate meta-analysis wouldn't cite a fake methodology. It would have no need to do so. The fake method is used because it sounds arcane and advanced, and whoever is doing the peer review is going to assume it's legit.

Reply on news.ycombinator.com