I found a significant flaw in my arXiv paper which quite impacts a major part of the conclusion, the paper has not been published anywhere but it has received citations to the arXiv version. Here are my questions:
Flaws in the analysis in a paper is a standard professional hazard in academia and it happens from time-to-time, particularly with preprint work that has not yet been subjected to peer review. What is expected of you here is that you will make timely corrections to your paper to do your best to mitigate any problem that might arise from having flawed analysis spread further in the literature.
I recommend you review your paper and see whether it is possible to correct the flaw in your analysis and revise your conclusion accordingly, and if so, how long this will take. There are three possible cases:
If you think this can be done quickly then undertake the relevant revision and post an updated version of the paper to arXiv.
If you think it will take a long time then you could create an interim update by using your existing (flawed) version of the paper but adding a page at the start alerting the reader to the flaw in the analysis and noting that you are presently undertaking revisions. (Also, if you don't yet know how to correct the flaw you can just say this, so that readers are not expecting a revision soon.) Once this is done, undertake the revision and post an updated version of the paper to arXiv.
If you think that the flaw is so fundamental that the paper cannot be fixed (e.g., the entire argument is flawed) then you should withdraw your paper from arXiv, ideally with a note explaining the reason for withdrawal.
As a secondary action, since your paper has already been cited, you could also do the academic world a favour and contact the relevant authors of the citing works to alert them to the flaw in the existing version of your paper and its implications. This is something that other authors will appreciate and it will prevent errors from proliferating in the literature.
- What should I do? I cannot delete the paper from arXiv, I can add a new version. Is it sufficient?
The arXiv has an option to withdraw papers, which seems appropriate for this type of situation. Providing a short explanation of the reason for the withdrawal (which is required when submitting the withdrawal request) will allow you to make it clear to anyone visiting the paper’s page what the issues with the paper are.
If you need to provide a more detailed explanation, then uploading a new version of the paper is another option to consider.
- What happens to the citations to my previous arXiv version given that I added a new version?
Nothing happens to them, they continue to exist. Anyone following up on the citations or linkbacks to your paper’s arXiv page will see that your paper was withdrawn and learn that its contents cannot be relied on to be correct.
If you are concerned that the papers citing your paper may be proving results based on it whose correctness is now in turn suspect (which is a very reasonable concern), the responsible way to handle this is to email the authors of those papers to let them know about the problems with your paper and that you have withdrawn it. They can then decide how to handle the issue.
- Generally how this can impact my academic reputation?
An academic reputation is built up from all the work you have made public. One paper with mistakes is embarrassing, but won't irreversibly destroy your reputation, particularly since you haven't published it, and everyone understands that arXiv preprints may represent work that's still in progress and can occasionally contain mistakes. If you keep producing more work with no mistakes of this type, I don't expect this incident will have a lasting impact on your reputation.
In any case, correcting the mistake promptly will have less of a negative effect on your reputation than if you neglect to address the issue and it's discovered much later.
My assumptions about academic publishing are:
Published research contains errors, some that are minor and some that make the papers junk. The most outright junk is easy to spot, but a lot of misleading or wrong research is indistinguishable from what is well done; if anything, the flawed stuff may look "better" because the authors have filtered out the imperfections inherent in data and research.
Some researchers care about the quality of what they publish. Others are either happy to publish fraud or sloppy work or too scared to correct their mistakes.
If a researcher publishes a correction to their work, particularly of their own accord, I've learned that they are capable of making errors. This doesn't surprise me, I know there are errors out there, and I know everyone is capable of making errors.
I've also learned that they care enough about the quality of what they publish to correct it and that they are not afraid of valid criticism. To a substantial degree, this increases my confidence in both the corrected work and any other work they publish.
Citations to your original paper are citations of your original paper; they aren't physically affected by anything you do next. You could consider reaching out to the authors if your correction would impact their work. If it were me, I would really appreciate hearing about it, though it's unlikely I would do anything about it unless it also affected the conclusions of my paper; you shouldn't expect people to change their own papers to point to your new version, all you can do is provide the information they would need to make their own decisions.
ArXiv is not part of the peer-reviewed literature, but a place to store papers that may be awaiting publication in such literature or, alternatively, as a kind of academese version of GitHub for papers. Papers have history logs and the sequence of revisions may be substantial. It was made with much the same philosophy as Git, and an evolution stream for a preprint is the expectation, not the exception.
You should not concern yourself too much with external citations, since they're not supposed to be used as peer-reviewed references in papers in the peer-reviewed literature. Even monographs shouldn't be. Their treatment would be more in line with that for technical reports and referees should actually subject the unvalidated sources to the review process, too. Otherwise, things slip into the peer-reviewed literature that were not subject to the process.
So, feel free to make revisions, even substantial ones.
An example of a preprint, that I was keeping close watch over during the course of its evolution between 2018 and 2023 was Oppenheim's recent publication of a General Relativity - Quantum Theory unification "A postquantum theory of classical gravity?" (https://arxiv.org/abs/1811.03116), which is now in the peer-reviewed open access literature under Phys. Rev. X 13, 041040 (https://journals.aps.org/prx/abstract/10.1103/PhysRevX.13.041040).
In its earliest stages, it took on more of a "gravity is not quantum" tack, even citing a couple potential "no graviton" links to support a "gravity is not quantum" stand. Later, however, the author decided to de-emphasize this possibility, removed the references, and re-stated the possibility that gravity may, indeed, be quantum.
If that was your cause, going into this, then it represents a blow: you're not establishing that conclusion. However, it does not detract from the key elements of the paper: (1) that of providing a cohesive formulation of classical-quantum hybrid dynamics, (2) a mash-up unification of classical General Relativity with Quantum Theory as a corollary of this and (3) a general framework for resolving the measurement theory / interpretation / classicalization problem of Quantum Theory, that will lead - and already has led - to a surge of papers that are now branching out this new framework (nay, new paradigm) to the different application areas where the foundations have for decades run into a wall.
So, even if the issue you're pushing doesn't pan out exactly like you expected it to, and even if there are errors, a paper still matters as a contribution in some fashion or another. Even authors may get blindsided by their own work, and nobody's perfect. At the very least, if you're on the other end, and are citing something on ArXiv, in preprints or other work outside the peer-reviewed literature, it's up to you to keep current on its evolution and make appropriate adjustments to your narrative, if its direction begins to stray away from yours. It is also your responsibility to review any pre-prints that you include in your work, if they haven't been published yet. Generally speaking, as I can personally assure you, they have errors - even the preprint I cited.
