Tuesday, November 23, 2004

Anonymous submissions and ethical practice

Over in Lance-land, a post about chilling out after a deadline has mutated (in the comments section) into a discussion of anonymous submission to conferences. My question, which I posted there, is an ethical one.

Often I will go around giving talks about papers that are not yet published. My reason for doing so (and I have come recently to this way of thinking) is that as long as I am not fearful of being scooped, or preempted, giving a talk is how one disseminates work, and asssuming that the paper will get published eventually, this is OK, especially if the work is something I am excited about and want to talk about. Related activities might include submitting an e-print, or putting up a web page etc.

The ethical dilemma is then this: if I am submitting to a double blind review process (that many conferences adopt), is dissemination of the above kind an underhanded way of subverting the anonymity of the review process and should it be avoided at all costs ?

Obviously a way out is to just not talk about work under review. However, this doesn't tend to be common practice: many important results have spread long before they were submitted or accepted to a conference. I imagine that some nuance based on the "importance" of the paper might come into play. If I believe that the knowledge of this result will be of great interest to the world at large, then it is ok to talk about it. However, this gets us into a slippery area of reasoning: after all, one doesn't write a paper and submit it to a conference if one believes that the results are of no interest to anyone.

Comments ?

3 comments:

  1. My feeling is that if you submit double-blind then you must avoid revealing your identity as much as possible: no talks, no eprints. To me this is a reason for avoiding double-blind submission processes. What do they gain you, anyway? It's often possible to guess authors' identity (if they're a frequent contributor to the community, or their outsider status if they're not) with reasonable accuracy just from the content and style of the papers, and those can't be hidden.
    D. Eppstein

    ReplyDelete
  2. Right, it is sometime easy to guess the author. But say, you may fail 20% of the total guess. Better than 0%, isn't it?

    :)

    ReplyDelete
  3. The crypto community has had to deal with this exact issue. Nearly all the major crypto conferences are double-blind submission. (Incidentally, Goldreich has a letter arguing against this practice, but we continue to do so.) At the same time, the International Association for Cryptologic research sponsors an eprint archive, and it wasn't clear whether posting works under submission to the archive would be OK or not.

    Today, IACR official policy is that eprint postings have the same status as tech reports. They should not affect a submission one way or the other. In particular, posting work under submission is OK.

    Personally, I don't have a problem with the practice. Indeed, I prefer to read results earlier rather than later. As a reviewer, I try hard to treat each paper fairly and without regard to the authors. I appreciate the blinding process, because that makes it easier for me. As such, while I like e-print postings, I do mind the hypothetical author that tries to drop hints about his or her identity -- but I haven't seen that yet.

    -David Molnar

    ReplyDelete

Disqus for The Geomblog