During the past year the SIGACT Executive committee has considered the issue that the STOC conference no longer attracts a large segment of the theoretical computer science community. Lance Fortnow wrote a proposal for restructuring the STOC conference to focus on bringing the theory community based on the model of large non-CS scientific conferences that accept all reasonable submissions, with many parallel sessions and plenary and invited talks. You can read the full proposal here.
The proposal had some support and some opposition in the executive committee. We did decide to circulate the proposal to approximately thirty members of the theory community. You can read many of these responses below.
Based on the responses, the SIGACT Executive Committee decided to start with a mild change, increasing the number of allowed submissions from about 80 to 90.
It seems the main objection is that many people like STOC and they see this proposal more as "losing" STOC rather than "gaining" something new.
ReplyDeleteJust doing a separate the poster session (as suggested by Noam Nisan, Aaron Sterling and Rocco Servedio) seems like a workable compromise that would allay any fears that traditional STOC is being eliminated.
This also seems to dovetail well with Dan Wallach's proposal for rebooting the entire CS conference system: www.cs.rice.edu/~dwallach/pub/reboot-2010-06-14.pdf.
As others have pointed out, rather than meddling with STOC which has its own constituency and traditions, perhaps it would be best to start a new conference called SIGACT as follows:
ReplyDeleteSIGACT-A accepts 60-to-90 papers refereed to a very high standard.
SIGACT-B accepts 120-200 papers refereed to a somewhat lower standard, but still a very high one. High quality rejects from SIGACT-A are given the option to present in SIGACT-B.
SIGACT-P takes 100-200 posters refereed to a more modest standard which are given the option of appearing or not appearing in the proceedings.
All three are held at the same place. Just from the number of presenters alone one should expect high attendance. Ideally the conference would be held during the summer months at a university campus, to facilitate attendance.
For me, the biggest concern about TCS ways of publishing is the prominence of conferences against journals. This raises the problem of checking proofs by the referees. It is not that uncommon to find some flaws in the proofs of papers published in STOC/FOCS or other good conferences (especially in the appendix part, as I suspect that we, reviewers, check more carefully "inside" proofs than "outside" proofs). Furthermore, contrary to what I read here, I do not find uncommon to see papers without a long version, therefore with missing proofs.
ReplyDeleteTwo conclusions about that:
1. It is a problem that conferences ask for appendices read "at the reviewer discretion" as you cannot know whether the proofs were carefully checked.
2. It is also a problem as reviewers do not have enough time to "carefully check" the proofs.
3. It is a problem as you have some papers without any proof in the published version, and full versions are not always available (or easy to find).
Some solutions:
1. Publish electronic proceedings, therefore no page limitations.
2. The only relevant solution for me: stop giving that importance to conferences, and switch to journals!
I am the previous anonymous, and forgot some points. I am very surprised by the arguments given in the other posts.
ReplyDeleteOne very recurrent comment is that we should keep STOC as it is because else it could be hard to compare CVs. But wait, isn't that more important to have some good ideas and good proofs instead? Moreover, it is often in conjunction with "how to compare old and new STOC?" But this has a name: side effects. So please try to see a bit more in advance than just your next recruitment!
The second argument I cannot accept is speed of dissemination. While almost everybody speaks about arXiv or ECCC, almost everybody seems to view dissemination as in the "good old days" when papers were typewritten and send by mail. Dissemination is done via those preprints servers. And I do not think conferences should disappear, but just lose their role of publication. So you could learn about new ideas in conferences just as now!
a) Given the current culture it will be impossible
ReplyDeleteto reach Lance's vision of a less prestigous STOC.
b) However, there are some things we can do to improve STOC. I propose these NOT as a way to get to Lance's vision, but as things that are good to do anyway. I suspect they are NOT controversial,
however may be hard
MORE INVITED TALKS
MORE TUTORIALS
SOME MORE PAPERS (The current notion of +10 seems good)
CO-LOCATE with CCC or SGC or BOTH
CHEAPER (this may be hard given the above)
Will this lead to more attendence?
Perhaps, but that is not why I propose them.
I propose them just to have a better conference.
STOC is better than FOCS for colocation since
people mostly don't teach over the summer.
Is 350 a low number? Where do you get the 1000 number? Could you provide a table of conference sizes for academic conferences?
ReplyDeleteWhat are the sizes, status, and roles of CCC, CIAA, COCOON, COLT, DCFS, FOCS, FSTTCS, ICALP, ISAAC, MFCS, LICS, SODA, STACS, STOC, WoLLIC, etc.? Why not a new conference or something like the SIGACT suggestion above?
ReplyDeleteHey, why not have a new conference that does not have high standards which welcomes everyone to present to create a giant party atmosphere with over a thousand people? You could hold it in New Orleans for four weeks overlapping all Spring Break weeks so that faculty could come when it fits their schedule. This would be no worse than conferences having many overlapping sessions that cause most people to miss talks they wanted to see.
ReplyDeleteParticipation, check.
Interaction, check.
What is it missing?
On a time-scale of decades, only three of the following futures can logically come-to-pass:
ReplyDelete• STOC grows much larger.
• STOC remains the premier conference for the theory of computing (TOC).
• The discipline of TOC remains no-more-fragmented than at present.
• Career opportunities in TOC expand commensurately with an increasingly populous and prosperous planet.
It might save time, if people stated explicitly which of these four alternatives they were willing to forgo.
Needless to say, these same four alternatives apply to all academic disciplines, not just TOC.
I will be a little harsher than my usual self: There are already conferences with an "everybody gets a trophy" acceptance policy, I find myself deleting them from my email every day. I understand that Lance does not want to go this radical, but I believe that serious referees will not be willing to invest the time required to really understand a paper when working with lower acceptance standards, so there will be an all or nothing effect here.
ReplyDeleteAs for making a conference large, I have been to extra-large mathematical conferences, and they felt a bit like an airport to me. Maybe they can work better for others, but I found myself lost in the hubbub.
Disclaimer: Recently I did not attend many FOCS/STOC conferences mainly because I am not an easy traveler, but I found the ones that I attended to be of just the right size for getting new collaborative research.
I will end with a constructive suggestion: It could benefit STOC to reserve 1-2 sessions for "Innovation" papers, ICS manifesto style. This could stir up things a bit in a positive way.
"The program committee for STOC would accept nearly all reasonable submissions, possibly limiting to at most one per attendee."
ReplyDeleteHow would this differ from the vanity conferences that are held? Realistically, who would actually come to this after a few years of the reputation being updated?
"There was a conference that was pretty good but then it started to accept more and more papers so that more and more people would attend until almost everyone attending was presenting work and the average quality of the work dropped to a very very very low level and the parallel sessions hit 10 per time slot. I now see the conference as a pointless waste of time."
Is that the future of STOC?
I said this on your June 22 blog and I will say it again:
ReplyDeleteSTOC & FOCS gained prominence by accurately evaluating the potential behind computational ideas and hence becoming trend setters. As their role diminishes to that of trend followers, their luster is rapidly eroding. Perhaps the most effective trend setter today is the Internet.
Summary: There is no way of saving these failed conferences with such a poor quality of paper evaluation.
Just doing a separate the poster session (as suggested by Noam Nisan, Aaron Sterling and Rocco Servedio) seems like a workable compromise
ReplyDeletePlease add me to the list of supporters of poster sessions.
Make science not politics.
ReplyDeleteSuggestion. To promote journals and encourage people to read journal articles, have some of the invited talks be an hour-long presentation of a result previously presented as a STOC conference paper and now published in a journal.
ReplyDeleteIsn't ICS the answer? While it was a small affair the first time around, there is no reason that it can't grow into the central theory conference, a la what is proposed here as the future of STOC (or a la ISIT for information theory). Still, I agree, that STOC and FOCS can be modified in some (less drastic) ways -- in particular by increasing the number of accepted papers.
ReplyDeleteSTOC reminds me of where General Motors was 25 years ago. There were a couple of clear flaws in GM's strategy that anyone with common sense could tell that needed to be fixed (more emphasis in quality, pay a bit more attention to gas consumption, less brands) but there were enough vested interests from all parts (management, dealers, unions) that nothing got done until it was too late.
ReplyDeleteAnyone can see that the impact, coverage and attendance of STOC has gone down over the last 25 years as percentage of audience. The solutions are obvious (increase conference size, collocate, reduce costs, increase coverage) but there are too many people who benefit from the status quo arguing against this necessary reforms.
STOC is far from going bankrupt, but if we give it enough time, these vested interests will take it there just as it happened to GM.
Here's an idea that just occurred to me:
ReplyDeleteWe could merge FOCS and STOC into a single annual conference.
The joint conference could accept roughly twice as many papers
(e.g., 150 or so). To accommodate the increase, it could be slightly
longer, e.g., 4 or 5 days, including a tutorial day, and/or have more
parallel sessions (4-6). It would have a single PC, but larger than the current ones, and perhaps more broad in terms of areas covered.
Advantages:
+ Attendance would probably be higher for the merged conference than for current FOCS/STOC.
+ Travel budgets reduced for people who normally attend both
+ Perhaps a broader coverage of ``marginalized'' areas.
+ Review process could be more leisurely and thorough
+ Hopefully, people would self-select ``breakthrough'' results for
FOCS/STOC, publishing more specialized results in specialized conferences with faster turn-arounds.
+ Less administrative work overall
Negatives that I see:
- Longer delay in disseminating good work; However, ECCC and other
archives fill this gap
- People would be likely to reduce the fraction of talks they attend. More chance for sub-areas to become separated into ``micro-conferences'' within the main conference.
Russell Impagliazzo
Why stop at unifying the *theoretical* computer science community? I'd love to see a unified computer science conference akin to the economists' AEA meeting or the mathematicians' Joint Mathematics Meetings.
ReplyDeleteI suggest the following variation on Lance’s suggestion:
ReplyDeleteFirst, there is a large meeting where *ALL STOC SUBMISSIONS* are presented. The STOC program committee decides *afterwards* which 80-or-so papers will be accepted to the final proceedings. So, if you want your paper to be accepted at STOC, you have to present at the large conference. This is the one scheme I can think of to get a large meeting that top people regularly attend.
I think that the laudable goals of the proposal could better be accomplished by:
ReplyDelete(1) making sure that STOC is held during the summer break for the overwhelming majority of universities.
(2) extending the duration of STOC to at least 4 days.
(3) adding to STOC several co-located day-long workshops, held in parallel to STOC paper presentations, on several topics (algorithms, complexity, crypto, topical). Each workshop should contain 10-15 invited [but not funded] presentations of the best work that was presented at non-STOC/FOCS venues in the past 1 year. Each presentation could be allotted 2 pages in the proceedings.
(4) having larger or hierarchical program committees, to improve review quality. (In the hierarchical case, 2nd tier PC members could be allowed to submit papers.)
(5) replacing the current best paper award with several (with no fixed upper bound) "best papers of STOC x-10 award" where x=current year.
Doing the above would help boost attendance, while retaining and indeed increasing the value of STOC as a way for to draw attention to the best works in TCS. (5) would also help to attract strong researchers who regularly attended STOC/FOCS 10 years ago to come back and talk with the young'ens.
I also strongly support Madhu's suggestion of not having a fixed upper bound in mind when the PC is deciding on papers.