Keira (She/Her)May 27

RE: https://mastodon.social/@kottke/116647849282379563

The thing I don't get here, is how?

I understand that AI hallucinates papers, it's very annoying and one of the reasons I don't use it for research.

But you're supposed to go read the paper before you cite it. Evaluate the paper, even. And you can't even skim or uncritically read just the abstract of a paper that doesn't exist.

So a huge number of published academics don't even read what they cite?

It's concerning that there's a higher standard asked of students than professionals.

#academic #academia #AI

Show threadFabian TranschelMay 27

@keira_reckons It's more like:

a) Papers need references to be taken seriously. So you add references that sound like they could be related.

b) Scientifically rigorous scholars read the literature even *before* writing their papers. Like, during doing the research.

I keep repeating ad nauseam that false incentives lead to problematic (to say the least) practices.

And measuring scientific success with citations is the *central* problem in modern academia, even before grant practice.

Show threadMarc A. KastnerMay 28

@ftranschel @keira_reckons for one recent conference I was asked to do a "LLM hallucination pre-check", between desk rejects and actual reviewer assignments basically checking all references of certain papers by hand and determine whether they exist or not.

I was kind of naively thinking it's just a formality, but then was pretty startled when I actually found one paper that had 4 or 5 completely made up references....

Show threadKeira (She/Her)
@mkasu @ftranschel oh wow, that's a lot
May 28 at 12:29pmWeb