Monday, July 14, 2008

A BIG Theory Conference

Lance talks about TCS's lack of a big conference after attending GAMES. (I recently talked about something similar with regard to ISIT.)

I'm in favor of a BIG cs theory conference, but I don't see a clear path to it. One way of doing it would be to have a mini-FCRC especially for theory, with 6+ conferences/workshops together over a weeklong period. I'd imagine there'd be some small, hopefully nominal fee to allow access to everything, and lots of administration issues. (A CD with all papers would be nice, for instance.) Some conferences have moved this way-- ICALP, for sure, and even SODA has ALENEX and ANALCO set up before it regularly -- but nobody seems to have tried to set up something like this. One possible advantage of this approach is that it could reduce the "quality control" concerns that generally arise for larger conferences.

So what would you like to see strung together into a theory-super-conference? A summer conference including SPAA, PODC, STOC, SOCG, EC, and a few other things sounds like fun to me...

9 comments:

Anonymous said...

Seems like an interesting idea. This is what PODC and SPAA is doing from time to time, but it's requires lots of coordination (and no we even don't know how to coordinate the deadlines).

Another thing:
One reason why some big conferences are having so big attendance is that they have many lower quality talks. Combining EC, SoCG, SPAA, PODC, STOC and if you wish even FOCS and SODA (ha-ha) would be just too exhausting for the listeners.

GASARCH said...
This comment has been removed by the author.
GASARCH said...

Our model could be like AMS
which might also be what GAMES does. The AMS has a meeting once a year with
invited speakers, but the rest of the talks are basically unrefereed.
Since math does not have
high prestige conferences
some of these unrefereed talks are pretty good
(of course, some are not).
But the point is communication. I also don't see a path to this especially since we have high prestige conferences.

Paul Beame said...

We've had a significant chunk of this at FCRC with as many as 4 parallel sessions between STOC, CCC, EC, and COLT (and PODC, SOCG, and PADS instead of EC and COLT in the past).

It hasn't quite worked out the same way as ISIT. Why not? One is simply that the talk times don't coordinate since each conference works on its own scheduling norms. I don't believe that it is the presence of the non-theory conferences, though their presence does mean that this highly parallel structure only runs for one day rather than the full length of the conference. My sense from my brief visit was that ISIT has a narrower focus than these collections of conferences in aggregate or at least the techniques and terminology in different areas within ISIT have more in common with each other. Is that the explanation? Is it the difference in selectivity? Is it just that different subareas within CS theory have grown more insular?

Anonymous said...

My sense from my brief visit was that ISIT has a narrower focus than these collections of conferences in aggregate or at least the techniques and terminology in different areas within ISIT have more in common with each other. Is that the explanation? Is it the difference in selectivity? Is it just that different subareas within CS theory have grown more insular?

ISIT is probably way more general than you imply. For instance, one could envision a paper from most other theory conferences there, but probably not the other way round.

Michael Mitzenmacher said...

Paul --

1) I think that for the "co-scheduling conference" idea to work, it has to look more like ISIT. That means a uniform schedule, in a uniform location. Ideally, things would be scheduled centrally -- it could be that at certain times a conference doesn't have a session, leaving the attendees free to actually look at sessions from other conferences! (Or, that gives time for people to talk to each other.)

2) I don't believe ISIT has a narrower focus -- that list of topics is very broad! There are whole parts of ISIT far outside my experience.

3) I think Bill said it right that the main difference is our conferences are nominally "high prestige" while ISIT is not, in and of itself, high prestige in the same way. However, "high prestige" is in the eye of the beholder. I do think subareas in CS have become more insular, and arguably a lot of smaller events could be combined without losing the prestige factor while making a whole bigger and better that the parts.

David said...

As a complete outsider, it's interesting to hear that prestige is an issue. Do you talk at a conference to demonstrate how much better you are than your colleagues, or do you talk at a conference to circulate your ideas and get feedback on them from an intelligent, interested community?

Michael Mitzenmacher said...

David,

I think the answer is "both". Particularly for junior faculty, it seems that having papers at the prestige conferences is a big deal in terms of career path -- perhaps (probably?) exaggeratedly so.

David said...

When I go to conferences, they are industry conferences. Topics presented are so widely varied that the average response, "Hmm, that's nice," and then to mosey on your way. If there is any prestige, it's inherited from your sponsoring organization [employer].

This is actually a pretty interesting system engineering problem. There is some conflict about the purpose of conferences. You want to make them better, but you can't just cast aside the alternate ways that conferences are used. You have to provide some alternate route for people to get to an end.

My design aesthetic says that the shopping mall approach to conferencing is going to be organizationally costly, and not get you that much. Either the bazaar, where small vendors peddle small chunks of ideas, and with its inherent chaos, or the boutique, where everyone is selling the same high end stuff, and its inherent order, would better baseline points of departure.

Maybe I'll send you an email, if I can come up with enough meaningful anecdotes.