« February 2009 | Main | May 2009 »

April 28, 2009

Dual submission - posters and conferences

As conferences experiment with new forms of presentation and formats, the old topic of dual submission deserves new attention.

Here, I am only concerned with concurrent submissions of the same or similar material to two conferences, i.e., submitting a paper covering the same ground to another venue while the first submission is still under review. While most authors would consider submission of the same full paper to two conferences to be clearly out of bounds, there has been less consensus on other scenarios, e.g., submitting a paper to a workshop as a poster submission and, concurrently, to a conference as a regular paper.

I don't think it much matters that a poster presentation has a different format and maybe a shorter representation in the proceedings. What matters is the core technical content, whether some implementation details and graphs are omitted or not.

This discussion is made more complicated by the fact that the treatment of poster submissions, in general, differs greatly between conferences. Some subject such submissions to very limited review, e.g., the student posters at CoNext, while others are reviewed at the same depth as regular papers. For some conferences, some sessions are more or less randomly designated as poster sessions, while others feature oral presentation, with Globecom being one such example. Other conferences move papers that fall just below the cut-off into poster sessions, converting them to short papers. Finally, the proceeding version of a regular and poster paper may be the same or the poster paper may get a one-paragraph abstract - and anything in between. With these distinctions, it is too simple to assume that all posters represent very early work.

One could argue that presentations in multiple venues ensures that good work gets to be known, particularly as the community is increasingly fragmented into ever more narrowly-targeted conferences.

Before considering this specific case, it might be worth going back to the question of why the community, in general, frowns upon double submission. I can think of four reasons:

(1) Avoid wasting reviewer resources, as we don't want to have very similar technical content reviewed six or more times, without the author having had a chance to incorporate the earlier feedback.

(2) Avoid "double credit" - authors shouldn't pad their publication count by republishing essentially the same content except under well-recognized circumstances.

(3) Listener boredom - nobody wants to attend a conference where half the presentations are re-runs.

(4) Wastes trees and library shelf space - why print the same or very similar material twice?

I believe that the last three reasons have become somewhat less important, and listener boredom doesn't factor if one of the formats is a poster, since the attendee can easily skip a poster that represents work they are already familiar with.

This leaves the first reason. Good reviewers remain a scarce resource, and we should husband such resources. Thus, I believe that, in general, such concurrent submissions are something we do not want to encourage or permit. Once it becomes known that such concurrent submissions are acceptable, there will be a strong temptation of more and more authors to submit papers to as many conferences as possible. However, if the poster does not consume any or minimal reviewer resources, the objections seem less strong. We routinely expect work under submission at a conference to be presented as posters at various local events, such as industrial affiliate or grad student research fairs, for example.

Recognizing that this may cause some confusion, one reasonable approach may be to clearly designate minimal-review poster sessions as such and for the poster event to indicate whether material under concurrent review elsewhere is acceptable for submission. The default assumption should be that it is not. Authors should also be encouraged by the conference organizers to cite such "publications" appropriately, such as "Conference X work-in-progress poster session" or "Conference X PhD research overview".

Any such policy should be coordinated with other organizations, such as TCCC. In particular, IEEE Infocom has a long-standing double submissions policy [http://www.comsoc.org/confs/ieee-infocom/policy.html] that is not entirely clear on this particular scenario.