Salil Vadhan


Here are my thoughts:

-          I do attend STOC/FOCS  on a regular basis, with the goal of keeping up on how other areas of TCS are evolving.  In fact, I have attended the last three STOCs.  What keeps me from attending every time is simply needing to limit my total amount of travel, considering family and teaching.   So I try to keep a balance between going to general theory conferences (ie FOCS/STOC), more specialized conferences (eg CCC and TCC), and focused workshops or one-time invitations.   What has led me going to STOC rather than FOCS more often recently is the fact that STOC is not during the semester, and that I’ve been organizing other conferences/events around STOC the last few years.  But, for the current discussion about overall attendance, perhaps the question of how often I attend is less relevant than whether or not I send my students to FOCS/STOC when they don’t have a paper – and the answer is that I do try to send each of them to one FOCS/STOC per year, typically the one that is closer to Boston.

-          The fact that other scientific disciplines have “traditional large meetings” does not imply that this is a better or more mature mode of operation for a field of research.  Scientific communication, publication, and dissemination and are all very different now than they were when any of these fields (or computer science, for that matter) established their meeting culture.  We should think about what’s best for TCS, given both the culture the field has established, and our assessment of present and future changes and opportunities.

-          While the conference system in CS certainly has some drawbacks, I think it is worth also reflecting on some of its benefits:

o   “Arbitration of competition of attention” (to quote Noam Nisan).  I view the role of a STOC/FOCS PC as selecting a collection of papers from across TCS (particularly outside my own areas) that seem worthwhile for me to hear about.  A traditional large scientific meeting will not have enough filtering to serve this purpose. Moreover, if it is broken into many parallel sessions, the talks will likely be aimed at narrower audiences.  And if I’m just going to attend sessions focused on my own areas, I’d prefer to do this in smaller, specialized conferences and workshops, which are better suited to real research interaction.
o    A level playing field for new people to get the attention of the community.  A new researcher (eg grad student) submitting a strong (but not necessarily breakthrough) paper has the guarantee that it will get a serious evaluation by the program committee, and if accepted, it will be heard by a broad cross-section of the TCS community.   With both the traditional large scientific meeting and internet dissemination, it seems to me that unknown researchers will have a much harder time getting the attention of the broader community.   The “invited plenary talk” model of large scientific meetings seems likely to focus attention on more senior, established people than new researchers.
o   Encouraging people to write up and disseminate results more quickly than they otherwise would.  Were it not for our regular conference deadlines, I know that I would probably sit on results much longer than I currently do.  While the infamous CS deadline rushes probably do lead to less polished and sometimes more incremental papers, I am not sure this is a bad thing.  The result is that ideas are communicated more quickly to the community, allowing other researchers to start building on them.

-          I have had little exposure to “traditional large science meetings,” but right now it sounds much less appealing to me than STOC/FOCS, for all the reasons mentioned above.   It feels as if the main community-wide interactions would become social rather than substantive.

-          All the above said, I too am concerned that the attendance at STOC has not grown as the community (and number of submissions) have grown.  But before proposing solutions, we should gather a lot more information about why people are and are not attending STOC, and what might attract them.   (Not to mention data about how many attendees are authors, students, faculty, etc.)  Without such information, I think that any adjustments should be incremental rather than revolutionary.


Addendum (6/18):

After seeing the well-attended co-location of STOC, CCC, and EC, I feel that a “Federated Theory Conference” may be a much better way of bringing the broader TCS community together than increasing the number of papers at STOC.  This would be like FCRC, but focused on theory, and hence able to be much broader in its coverage of theory.  One could imagine every few years trying to co-locate STOC with CCC, EC, SoCG, LICS, RANDOM-APPROX, TCC, PODC, and  more.  STOC could be in the middle of the meeting, and have nothing in parallel to it.  This would bring many more people together while retaining the advantages of the current conference structure:
-          The STOC program could remain small enough to serve a “highlighting” role with papers and talks aimed at the broader theory community,
-          Through the co-located conferences, people would have enough going on in their subfield to justify the trip.  Moroever the subfields can retain their own sense of community through the co-located conferences (where talks are aimed at people in the same subfield).
-          This seems to be simultaneously low-risk and high-reward.  It can easily be tried as an experiment once, with potentially large payoff for the community if it works well.  However, almost nothing needs to change as far as how the individual conferences work,  so it is easily reversible.

No comments:

Post a Comment