Monday, December 18, 2006

The Mega-Conferences

Chicago will be invaded by economists in early January, coming to the American Economic Association's Annual Meeting. At the same time the mathematicians meet in New Orleans. The physicists meet in March and April. We computer scientists all get together…never.

Most fields have their big annual get togethers with their plenary talks and many parallel sessions. New Ph.D's meet with potential employers often in a very organized way. Most importantly the entire community comes together to discuss the fundamental scientific and political issues of their discipline.

We don't have those meetings in computer science. The ACM has an annual get together where they give out awards but that is relatively small. Every four years we have the Federated Conference, a joint meeting of several conferences but they don't span the field, usually lacking a major AI presence.

So why don't we have a CS Annual Meeting drawing tens of thousands from across the discipline? Many of the other annual meeting started in a time when travel was more difficult and a single, or small number, of large general meetings made sense. We are a much more conference-oriented field and few of us would like to take yet another trip to a larger conference.

We lose something by not having a single regular meeting across computer science. We rarely meet people outside our field who are outside our departments. Different subfields in CS have developed different cultures. We lack the cohesiveness of other fields. When someone says "I am a Physicist" we know what that means. When someone says "I am a Computer Scientist", do we?


  1. I agree. It would have two effects:

    1. People will hear about research in other fields.

    2. People will think how to present their research to people from other fields.

    Co-locating conferences is a mild form of what you are suggesting, and is being done.

  2. There are two ways to look at this
    (probably alot more).

    I) Computer Science is a much more diverse
    field then Math or Phy or Economics.
    Hence a massive conference would not really
    work. People are somewhat insular and
    do not really care (that much) what is
    happening in other fields. This is
    intrinsic to CS.

    II) Computer science has BECOME insular
    because of the LACK of a massive conference. The insularity is NOT
    intrinsic to the field.

    As with most things the answer is a bit
    of both. The question NOW is would
    such a conference help NOW?

    The ACM and/or IEEE could help since
    most CS people ARE in one of them
    and they cover most (all?) of Computer

    bill g.

  3. The field would certainly benefit from having a large conference. I do not know if the federated conference is the way to go about it: every four years ACM "highjacks" the main conference in the field so you can mingle with other areas. Quite understandably this generates a lot of opposition from the people who run these conferences (witness SCG and CCC).

    The ACM would do better to organize a brand new biannual conference jointly with IEEE, SIAM, AMS-MAA, IACR and AAAI. There should be a large number of invited talks, some plenary talks plus a few dozen refereed contributions for each existing SIG.

  4. Bill's option (II) can't be right. ACM used to have an official Annual Conference (as opposed to FCRC which is a bunch of SIG conferences brought together) at which the Turing Award was given out, among other things. This conference withered and eventually was cancelled in the 1990's as a big money-loser.

    Why didn't the annual conference work in CS? (One 'problem' for CS is that we already get together too much and too intensely for reasons other than discipline-wide annual meetings and our travel money is used up (not to mention our interest because of the contrast between such meetngs and our other workshops/conferences). Can we expect to regenerate interest in a truly displine-wide conference when so many people who would rather attend even narrower workshops or conferences than go to broad conferences even within their sub-discplines?)

    However, the real reason is historical but in a different way from Bill's (II). CS grew out of a wide range of disciplines to begin with (EE, Physics, Math, OR, etc), initially held together by the common interest in the machines themselves (hence the ACM name). As the organization grew beyond that, the focus of the annual meeting became misplaced as a focus for research. The original 'core' of the field had lost its gravity and the sub-disciplines revolving around it broke off into largely independent planetary systems with modest attraction to each other. We remain a discipline that prizes diversity and branching into new fields: biology, economics, medicine, even physics, and may never find a similar core around which we can gravitate.

  5. While I am young and still new to the field, it seems to me that one great thing about computer science (especially theoretical computer science) is that it is largely a very friendly field. While there seem to be tuffs every once in a while, it seems nothing like what I hear from my fellow graduate-students who study other sciences. I don’t know what this has to do with an annual meeting (maybe people go there and fight, idunno), but it sure is nice.

  6. To be honest, it is not clear that the big meetings do play all the roles that you suggest anymore, and they are not necessarily worth imitating exactly. Here are some thoughts about the APS March meeting in particular.

    1) The meetings are so large that they can only be located in drab cities with huge conference centres, like Baltimore and Denver. It's true that the March meeting was in Las Vegas once, but they won't be going back there because the physicists didn't spend enough money in the casinos.

    2) You usually end up talking to far more people from your own field, who you could have met at a specialized meeting, than general physicists.

    3) You can go to a few sessions outside your field. These can be interesting, but the talks are so short (often around 10min) that it is hard to get a feel for the subject unless you already know something about it. They are more like adverts than research talks.

    4) Because the APS has a rule that any member with an appropriate affiliation has the right to give a talk, you have to sit through a LARGE number of weak talks to hear a few good ones.

    The main reason for the continued existence of these meetings is political. If you want your field to receive money then it helps to have a strong presence because you know that all the big shots and representatives from major funding agencies will be there. Also, if you can get a large number of people to join your group or division then you will have more money from the APS itself to use for organizing specialist meetings, giving research prizes, etc. The other main useful purpose is to talk to people from outside research altogether, like publishers, government representatives, etc. This can be useful if you're thinking about writing a book, want to start a new journal, change public policy on something, etc.

    The bottom line is that these meetings are not so effective for the communication research amongst peers, but they do serve a political purpose and help to give a sense of cohesion to the subject as a whole. It's inevitable that any meeting of a similar size and scope will be like this to some extent. Although this is not entirely a bad thing, it's important to have your eyes open if you try to create a similar meeting for CS.