16
$\begingroup$

I was reviewing a paper of a double-blind conference (ML/AI-based conference). The authors improved the approximation bounds for some special instances of a problem. To understand their proof better, I was thinking of a solution by myself while picking some hints from their proof. Using those hints, I came up with a better algorithm for the general instance of the problem. Moreover, my algorithm is much simpler than theirs.

I am feeling a bit greedy here; I want to suggest these changes to them and also want authorship in their paper. I can not ask them since I do not know them. I can not publish my own paper since a few ideas are borrowed from their proof and their manuscript is not online. What should I do?

$\endgroup$
4
  • $\begingroup$ Write a paper about the general instance of the problem to showcase your solution. $\endgroup$
    – Wei Wei
    Jul 12 at 8:30
  • $\begingroup$ This can work only after publishing their paper first. And sometimes at AI conferences like AAAI/IJCAI you need to wait a couple of months after the paper appears. $\endgroup$ Jul 12 at 8:58
  • 2
    $\begingroup$ Not the same situation, but could be relevant: academia.stackexchange.com/q/99487/64 $\endgroup$ Jul 12 at 9:46
  • 2
    $\begingroup$ Similar to your situation, but for a journal: academia.stackexchange.com/q/48074/64 $\endgroup$ Jul 12 at 9:50
19
$\begingroup$

I don’t think there’s a clear protocol. I’ve seen referees generously offer improvements. I’ve seen authors offer coauthorship for those improvements. I’ve seen the referee accept or decline the offer.

Of course, it’s perfectly legitimate for you to wait for them to publish and then submit your improvement. Delaying by a few months is a gamble (might get scooped, including by the authors). But the preceding options involve gambles as well.

$\endgroup$
2
  • $\begingroup$ Thanks! I agree with you. But, If the authors want to offer co-authorship, how can they do so? They do not know my name and I also do not their names. $\endgroup$ Jul 11 at 19:24
  • 9
    $\begingroup$ In a journal, the communication is through the editor. At a conference it’s trickier but presumably the Area Chair can mediate. $\endgroup$
    – Aryeh
    Jul 11 at 19:54
6
$\begingroup$

I made a flow chart describing what I think is the typical process in this case: flow chart

You will notice many articles end with "We would like to acknowledge the anonymous reviewer whose suggestions tremendously simplified the proof ..."

$\endgroup$
3
  • 2
    $\begingroup$ I like the flowchart, but some comments: "just" a few days may be already a long time in case of conferences were you rarely have more than a month. Also, I would personnally add an extra layer before the last no: "do you think this is a good enough contribution for a whole new paper / do you plan to work on this subject". I often find non trivial simplifications of proofs in papers I am reviewing and often give it to the authors because I do not have time/energy to write an incremental version of the paper later. And contributions done in a few days are rarely that interesting. $\endgroup$
    – holf
    Jul 13 at 13:27
  • 1
    $\begingroup$ @holf Good point about contributions that are too small for a new paper and which you don't have any plans to expand on. I'm not sure what should really be done with such ideas. If they don't fit into the current paper, the authors may not gain any benefits from them. On the other hand, it never hurts and they might get inspired. $\endgroup$ Jul 13 at 20:05
  • $\begingroup$ This seems like an excellent answer, I suspect that there are some similar questions on the AcademiaSE that would also benefit from this flowchart! $\endgroup$ Jul 14 at 9:11
1
$\begingroup$

Some reviewing guidelines state explicitly that referees are not supposed to communicate any original research as part of their review letter.

If this is not the case, there is the option of telling the authors (as well as the area chairs) openly about the dilemma you find yourself in and, at the same time, sketching your generalization in sufficient detail. Your trust in the authors' fairness in handling the situation might be rewarded by a fruitful collaboration.

$\endgroup$
1
  • 1
    $\begingroup$ I'm not sure there's a meaningful distinction between, say, finding a mistake in a proof or suggesting a considerable simplification and "original research"; both of the former could, under appropriate circumstances, be publishable as such. $\endgroup$
    – Aryeh
    Jul 13 at 8:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.