Monday, June 21, 2010

Colocating Conferences -- What Will It Take?

In the ongoing future of STOC debate, one possibility that seems to have significant support is that we should do more colocation of conferences.  (See Suresh's latest post, for instance.)  That is, rather than expanding STOC itself, it is thought that the best way to make STOC more of an event that theorists will want to attend is to set up a "pan-TCS" conference, that has several related conferences or workshops the same week at the same location.  Arguably this would allow STOC to continue essentially the way it is -- which many people seem to want -- while leading to higher attendance and community building.

Given the success of the STOC/CCC/EC mix this year in Cambridge, it's hard to argue against this idea.  So I won't.  Heck, I think it's a fine idea.  The one thing that people seem to be ignoring (or perhaps just forgetting), however, is that this will require some non-trivial organization to make this work well consistently over time.  And I'm left wondering how this sort of organization will come about.  (Honestly, it's not quite clear to me how things all managed to work out this year!)  

Right now, separate conferences organize themselves.  Heck, even getting conferences following each other, like FOCS/SODA coming up, to organize their deadlines so people can send their rejects on to the next conference appears to be a strain.  (Anyone else think that the 1 week between when we're supposed to hear back from FOCS -- hopefully with actual reviews, but maybe not -- and the SODA deadline -- which includes the 4th of July weekend -- is maybe not enough to make worthwhile improvements and changes?)  Now let's think about what has to be done to make this sort of colocation work:

1)  Arranging conference hotels, and meeting space, in roughly the same place at the same time.  (On another post somewhere, some anonymous people complained that STOC and CCC/EC were not conveniently located -- they were miles apart, and you'd have to switch hotels to go to both easily.  Perish the thought!  Add "adjusing people's attitudes" to the general list of what has to get done to make this work...)
2)  Synching up schedules, including submission deadlines (will one conference in the set be able to take another's rejects for submissions and still have a workable timeline? ), and where possible avoiding scheduling important plenary/invited talks at the same time (sometimes this might not happen, as with this year's EC tutorials overlapping the STOC conference -- see above about "adjusting people's attitudes"...).   
3)  Managing these relationships and corresponding decisions years -- possibly multiple years -- in advance;  some conferences have different ideas about how often they will be run internationally, whether they should be connecting just to theory or other areas (e.g., EC is also connected to AI), etc.  So conferences will be joining and exiting this system regularly, requiring further high-level organization and significantly more advanced planning that we currently have.

I should be clear that I don't think any of these organizational issues can't be overcome.  It's just that they'll require, naturally, some organization.  Which translates into more time (at least up front -- maybe less time years down the line) spent by people in our community to put these sorts of things together, and/or more costs (as, like with FCRC, the organizational aspects are left to other organizations -- that charge for it).  So for all of you eager for this style of change -- as opposed to, for example, simply increasing the number of accepted papers -- I eagerly (very, very eagerly) await to hear your voices chime in when volunteers are sought in the future to help make this sort of thing happen.

8 comments:

Suresh said...

Very true. While in my post I was thinking more along the lines of workshops etc that are under the auspices of the conference, the general point remains true. In fact, if you look at these conferences, they don't just have a PC chair and local arrangements chair. They have chairs for awards, tutorials, workshops, panels, finance, and other topics. So there's a much greater degree of top down involvement.

Anonymous said...

The Europeans seem to do this rather successfully at least in their software engineering conferences.
Example, ETAPS (http://www.etaps.org) has a bunch of theory, compiler and applied SE conferences rolled in together. They typically do it over summer when the dorm rooms are available at a uni for hosting attendees.

Anonymous said...

You shouldn't take anonymous commenters so seriously. Nobody minds switching hotels.

For these issues, more conference colocation is not necessarily better. One other conference is a good idea, but two or more and the hassles outweigh the benefits. Not many people have the stamina to attend three conferences in a row.

As Suresh says, other activities are more important than only colocation. A poster session, for example, requires fairly little organization.

Jonathan Katz said...

I don't think it has to be very difficult, although I'm thinking here more in terms of having several smaller-size tracks (a 'core' STOC track along with specialized sub-tracks in other areas) rather than co-locating larger and already-existing conferences.

Here's how to do it: When forming the STOC committee, make sure to have sufficient representation from any sub-field that will have an associated workshop. (If there is a desire to keep the STOC PC small --- which I don't fully understand -- then you can have sub-committees whose members need not all be part of the STOC committee.) All papers get submitted at the same time, to the same site. Authors can mark which sub-track they find most appropriate. When decisions are made, papers can either be accepted to the 'core track' of STOC, or to one of the specialized sub-tracks (or rejected entirely). Presumably the way this would work in practice is that 'core' STOC decisions would be made, and then the borderline papers that have substantial support would be accepted to the sub-tracks.

The main disadvantage is that you would get more submissions overall. This could easily be dealt with (at the expense of a slightly longer review time) by having a 'filter' round at the beginning of the review process.

Paul Beame said...

Jonathan's notion of adding extra stuff to a core STOC is OK but misses a big part of the point: It isn't the program committee part that's the bottleneck. It's the local arrangements.

I've been involved now with the organization of a bunch of STOC/FOCS conferences on the logistics side but these new multi-mode conference ideas require much more energy and ability to navigate uncharted waters of conference organization than our typical local organizers have time for. We have the regular STOC/FOCS down enough that, with a little guidance, most members of our community ought to be able to handle being a local organizer. However, a conference format with many more moving parts would be another thing entirely.

Suresh said...

With workshops, the way it works is that the workshops are scheduled ahead of time in slots, and so the local arrangements folks sync with the workshops chair to make sure there are rooms etc. Then the workshops chair solicits proposals for workshops, vets them, and selects a few.

After this, it's upto the workshop organizers for each workshop to send out a CFP, get submissions, review them, and make a list of papers for inclusion in the program. In fact in many conferences, the workshops are hosted off different websites, with merely a link at the main conference page.

So the actual coordination effort is required, but is not as much as might seem.

Jack said...

I'm hosting ACM Symp. on Computational at UNC Chapel Hill in 2012, and am actively looking for workshop and co-location suggestions.

I hadn't actually thought of collocating with STOC, since that has been done at the federated conferences, which SCG has opted out of, primarily due to cost. (SCG seems a little price sensitive; some of our best have been held on college campuses, which give options for dorm accommodations.) I've been looking towards more applied conferences, but would be happy to talk about STOC...

Jack Snoeyink@cs.unc.edu

Anonymous said...

Mike,

surely you do not wait until you hear back from the FOCS PC before making improvements to your paper! At this point, enough time has elapsed since the submission deadline that your paper has probably been rewritten entirely, the proofs tightened and clarified, the results streamlined and generalized, the figures completed, and the typos eliminated. You wouldn't wait until you hear back before starting to do all the things which you knew, at the time of submission, were still work in progress. Right?