@Drugmonkey has
a very good response to the storm over whether arxiv's proposal to ban all authors on a paper from future submissions if AI hallucinations are found in a paper.
https://drugmonkey.wordpress.com/2026/05/20/thoughts-on-each-author-taking-full-responsibility/
Relatedly, in the middle of the XX century the scientists that were responsible signed the paper as authors, and everyone else was listed in the Acknowledgements section.
That today we list everybody as "middle authors" is, on the one hand, a consequence of the "publish or perish" culture, and on the other, a way to bring visibility to people who did the actual work (I am thinking here of famous neuroscience papers were the actual work was done by non-PhD women technicians who were merely "acknowledged" instead of signing as first authors, which would have been the case today).
Further, in my own papers, I don't add anyone as author who hasn't read the whole paper and commented on it. As a bare minimum. Each contributor should know first hand what they are agreeing to endorse with their name, even if they couldn't do that work themselves without a lot of training.
But as the blog says, no one really is in the position of being able to vouch for every aspect of a paper. No one can watch every experiment to vouch for its accuracy.
And I’ll confess, with co authored papers I trust my colleagues to cite the relevant research in their sections. Just like I trust them to be doing the experiments as described.
I agree. Yet I ask co-authors to be comfortable with that, because if there is any problem with the paper, they are also going to be impacted.
Yes, always a wise path.
@albertcardona @MCDuncanLab @Drugmonkey on this note, after reading a few other people's reflections on the policy, I think my greatest concern is that coauthors, particularly middle authors, may not even be aware they're on a paper until after it's submitted, perhaps even after damning issues have been found by reviewers or editors and brought consequences down on them. I have more than once been told about a paper's existence after it was submitted and been invited to give any feedback in the meantime to be addressed during revisions.
If they have concerns, especially on papers with many coauthors, they may not be in any position to do anything about it short of take their name off. I have witnessed it happen and have great respect for that decision, but it's not an easy one. Coauthors likely don't have an opportunity to verify something they're only vaguely uncertain about. I think some journals' policies of collecting contribution statements from authors (X and Y carried out experiments, Y and Z did data analysis, Z wrote the manuscript) is a sensible partial remedy for this.
At the same time, I want to acknowledge that the absolutist stance has some merit -- if personal responsibility for the entire contents of a paper became, practically and enforceably, the norm, I imagine we'd shift other norms and practices to accommodate. I just don't think that's the case at the moment, and it seems unnecessarily destructive to the coauthors who contributed in good faith to jump to the level of enforcement that this policy does.
@iris @albertcardona @Drugmonkey
I too have been added to papers without my knowledge. Indeed, for one, I ended up being somewhat unhappy with the direction the paper took (this was a project I left unfinished when I left the lab) but the choice was either to stay as an author as is, or withdraw my name, and not get credit for work I'd done.
In this case, it was not anything falsified, just I felt they overinterpreted the data.
@Canageek @iris @albertcardona @Drugmonkey
😱 and for physics arxiv is like *the* publication of record.
My god, that's going to kill careers.
As I’ve said before, science runs on trust. I could so easily lie about what siRNA or antibodies I used to get exactly the image someone asked me to get without using AI. Anyone could.
However, keeping up the lie over multiple papers and with others potentially trying to advance on my false data is another thing entirely.
This is what’s great about modern science, most fields are robust and self correcting, one bad apple rarely diverts the whole field.
However, if we cross a tipping point, where more ppl are falsifying their data, science will grind to a halt.
But that’s kind of beyond the concerns about hallucinated references, which is itself quite trivial in my mind.
Scientific papers generally are supported by the merits of their data* not their citations. So I don’t so much get the handwringing.
(*Maybe this is different in other fields)
The shock seems to be to have discovered that some scientists don't think much of having the occasional wrong (or nowadays, inexistent) reference in their manuscripts.
@MCDuncanLab @Drugmonkey Sorry, strong disagreement on a couple of points. Before that, I’ll just say that Tom Dietterich is a well-known and highly respected scientist and Prof. Emeritus of Computer Science. He is the kind of person I would 100% trust to get it right on a co-authored paper and his conduct is unimpeachable. He sets a high bar, yes, but it is also consistent with excellence in the conduct of science.
If you co-author a paper with someone who is very junior or less than stellar in their scientific methodology or standards of writing a scientific paper, that’s on you because you made the decision to delegate more to them than was reasonable. I’m not saying you must avoid working them, only that you hold them to AT LEAST the standards you follow yourself.
For junior collaborators, finding bugs in their reports is a mentoring opportunity. For senior co-authors who provide slipshod work, it is an opportunity for you to take corrective action to protect your own reputation and also to reconsider the collaboration. Collaborators don’t need to be experts in each other’s area, but they do need to be able to interrogate each other’s contribution. From the readers’ perspective, co-authors are judged together on the content.
Can we all agree that citations must not be fabrications? That is easy enough to check. Accuracy and relevance are harder if you don’t know the subject matter, and this is where it gets dicey. People have different ideas about what to include and that is to be expected. It is fair to say that there is some burden on the reader to make this judgement. As a minimum, citations ought to be useful in some way, especially if they are not discussed in the body of the paper.
Finally, if somebody has not made an original and substantive contribution to the work described in a paper, they should not be a co-author. The responsibility is on them as well to decide if they should be listed or not. Padding their resume is not a good justification.
Glad to see that so many people are willing to sacrifice so many to the hill of 'Thou shalt not ever in your life be lazy and fail to check every reference on every paper...(that you may or may not have been consulted about prior to it being) uploaded in Axriv.
And no, it was not padding a resume when I didn't pull my name from a paper that represented 2 years of my PhD that was submitted without my knowledge. And I was not going to burn bridges with former PI over it.
MCDuncanLab
Albert Cardona
Iris Young (he/they/she) (PhD)
Canageek
David J. Atkinson