Hacker News Reader

Comment by defrost | Hacker News Reader

I read the tone of this as essentially neutral, I would have liked to have seen some numbers on the propotion of retractions.

Key notes are:

* arXiv allows authors to withdraw papers whenever they’d like (with as much or little detail about their reasons as they care to provide).

* All prior versions remain online, without notation that the paper was withdrawn. (Without notation!!)

* Roughly two-thirds of biomedical, physics, math preprints are eventually published in peer reviewed journals.

The take aways are that a third of STEM preprints are never submitted (in arXiv form) for peer review, and an unknown proportion are withdrawn with no obvious indication that happened.

View on HN

Previous versions such as https://arxiv.org/abs/math/9810162v1 clearly link to the latest withdrawn version, with text in red and bold at the top. Researchers accustomed to the arxiv don't miss it.

Peer reviews don't catch some important mistakes. It does not guarantee much by itself.

I make no claim to peer reviewed journals being better, I simply noted that the linked article asserted that two thirds of arXiv STEM preprints are latter submitted to such journals.

> Previous versions such as https://arxiv.org/abs/math/9810162v1 clearly link to the latest withdrawn version, ...

Thanks for that note; that appears to apply to a succession of papers that each replace a prior version.

The statement made in the linked article that I quoted suggested that papers that are later withdrawn are NOT annotated to identify that have been withdrawn.

If we look to the latest version of that paper:

https://arxiv.org/abs/math/9810162

We can see that the author has added a comment:

> Comments: The paper has been withdrawn due to a crucial mistake in the arguments

but no other "meta" annotation has been added (such as the one you pointed out).

It seems the linked article is asserting that had the author NOT made that comment then there would be no other indication that paper had been withdrawn due to error.

In fact I'm not entirely sure what withdrawn means .. given the paper remains on the site.

> The take aways are that a third of STEM preprints are never submitted (in arXiv form) for peer review

Technically, putting it on arXiv is submitting it for peer review :-)

This is my read. arxiv + Twitter (or whatever other public platform) are the new peer review, and journals will slide into irrelevance. Probably for the best as far as truth is concerned, but the adjustment for both academia and science journalism will be rough, and I don't know which side will be worse.

This was a fresh take a decade ago, when altmetrics were all the rage. If this was going to happen, it would have happened by now. The arxiv and the web have existed for decades, "academic Twitter" has been a thing for a decade, and legacy journals are still as powerful as ever.

So-called "high-impact" journals perform a gatekeeping, credentialing, and attention-focusing function that is valued by the institutions of academia. What you and most people on HN probably see as a bug is a feature to them. It is part of why I left, to be honest.

A decade is far too short a time span for a shift like that. I'm aware that journals are quite entrenched; I've heard the phrase "publish or perish". If it happens, it'll be on more of a generational time scale. It has to be proven possible for someone to make a career without getting "published". Maybe by the time we're both almost dead, truth will be enough. :)

And yeah, all those features you mention that academia values are useful for the press, too. It's really handy to have a finite set of vetted sources, rather than trying to keep track of the reputations of various scientists and use that to parse their comment threads. That's why I said it will be rough.

Maybe the article is neutral, but the site does not come off as such.

From their homepage:

>The week at Retraction Watch featured:

> Mathematician withdraws preprint – 24 years after initial submission

> Penn says access to former Twitter employee’s thesis was ‘mistakenly closed off’ following Elon Musk tweets

> PLOS flags nearly 50 papers by controversial French COVID researcher for ethics concerns

> Cancer researcher banned from federal funding for faking data in nearly 400 images in 16 grant applications

> ‘I never asked or expected to be included as an author’: Retired Penn State prof has three retractions for manipulated peer review

The thing about mathematics papers is that, for the most part, social acceptance is _the_ test of research. I say social acceptance, because that is what it is, namely a culture of rigorous logical review of the arguments in a paper. As such, you don't really have ethics concerns relating to faked data or p-hacking in a math paper.

When a math paper is retracted, it is almost certainly due to an honest mistake. Papers with such an error are likely not a very significant paper to begin with--as the error likely escaped noticed because there were fewer eyes on it--or there is a very deep and subtle error that would be hard for anyone to catch.

What do you see as non-neutral?

It seems curated towards scientific misdeeds. Maybe non-neutral is the wrong word, but I could totally understand being upset having your name on the site.

People won’t feel comfortable retracting papers with errors if the practice in the industry is to highlight the retraction on a giant billboard next to items like “Scientist commits massive fraud.” If that’s the consequence for honestly correcting your own scientific record, better just to leave incorrect work out there.

For one thing, why is everything written as "scientist retracts paper" instead of "paper retracted" ?

Reply on news.ycombinator.com