Sunday, August 02, 2009

Response to Dr Lance Fortnow's CACM opinion

Dr. Lance Fortnow published a strong argument against the current 'strong conference' CS system in the August issue of CACM. His essential desire is that CS move to a system similar to the hard sciences and engineering. Conferences should accept any paper of reasonable quality, not publish proceedings and hold Journals as the sole vehicle for publications.

My Questions:

  1. While CS should have a stronger Journal system, why should that come at the expense of quality conferences?

  2. The reputation of CS conferences as a method of publication means that it is both acceptable to publish and cite papers from these confs. The result of this is that one can read a conf paper, compose a citing follow-up and publish it within 1 year. This fluidity of ideas without the sometimes 18 month to 2 year wait on Journal publication is a great advantage!

    Yes other disciplines publish pre-prints to is this really a solution when a paper has been rejected yet avail on for 18 months?

  3. Is this problem really that acute in all communities? Can't it be solved at the community level?

  4. Certainly in AI, Machine Learning, Data Mining and Evolutionary Computation I perceive that the Journals are held in high regard (all papers are great) and that the conference proceedings can be a mixed bag with strong and weak papers. I am getting strong pressure from my advisors and peers to consider Journal versions of some of my conf papers. EC specifically has a non-proceedings conference to meet and discuss less settled results.

    If Dr. Fortnow feels that his area of theoretical computer science is too fragmented, then a solution would be to found more Journals and push the best conference papers into those Journals more heavily. Perhaps that would mean that the conference system of that area would shrink in response as the Journals established themselves.

  5. Why would industrial researchers and scientists participate in publications with such long time cycles?

  6. Again the fluidity is an advantage. Were CS suddenly to switch to a soft conference system (low benefit to participation other than networking) I fear that participation of industry in publication venues would suffer. The time scales of Journals in general mean that at publication time the results were done an eternity ago in industry terms. Convincing your supervisor that participation and publication in a conference is a far easier sell than an extended Journal submission effort.

    One also wonders if the Journal system is not implicitly biased towards academic communities where participants are chasing tenure. This 'ranking of people' referenced by Dr. Fortnow is very much for academic institutions and not of much value to industry IMHO.

  7. Why do we want ANY such top-down forcing of CS organization?

  8. The culture of CS is much more aligned with self-organization and communities forming out of a bird-of-a-feather effect. This also aligns with the changing face of corporate cultures and culture in general. Such a top-down driven reorg would likely both fail and break the inherent fluidity of ideas and results in CS.

    I also have an unfounded suspicion that such a top-down forced re-org would result in a clustering of power and influence towards traditional centers of power in academia. If one picks up conference proceedings in my favorite CS areas and does a frequency count of the author's institutions the distribution is very much long tail. The 'elite' universities are not dominating the results meaning that the 'in crowd' effect is much lesser in CS.

    Feudal systems are dying fast for a reason.

  9. Obviously the current system is serving a need, doesn't that speak for itself?

  10. If CS researchers and scientists continue to attend, publish at and found conferences is this not evidence that it is serving a real need?

While Dr. Fortnow is correct on his points w.r.t. the problems faced by conference 'PC' committees... the correct response is best done within that community. Found some Journals and compete with the conferences for publications and reputation. I don't accept that a strong Journal system can only be created by first wiping away a very fluid and successful conference system.

My personal solution to strengthening the Journal system? I'll set a goal of submitting a Journal publication or two in the next year. I am completely remiss in not yet submitting anything to a Journal.

As a counter point to the above arguments.. if I could have a wish it would be a Journal system with fast review times, immediate publication of accepted papers and that markets itself by cherry picking great papers from conferences and encouraging those authors to submit to the Journal. Something like the Journal of Machine Learning Research. The publishing of referee reviews also sounds interesting.


Daniel Lemire said...

I am reproducing my comment from Daniel Tunkelang's blog:

1) Let the Journals compete for papers with the conference system and attain the proper audience.

Ah! But journals do not compete fairly because there are top-down rules stating that conference papers count more than journal papers.

Many CS researchers will discard journal publications as minor. This is a fact and the ACM article makes this clear: experts decided that journals were less important than conferences in Computer Science. This has been told to all University administrators.

2) In 2009, there is no reason for journals to be any slower than conferences. They used to run slower due to the printing press which we no longer need. If you need to go even faster, I’m afraid you’ll have to forgo peer review and… gulp!… post unreviewed papers on arxiv.

I have had several of my journal articles accepted within 3 months (including revisions). Others took 6 months. Some took longer, but I argue that the journals are mismanaged in such instances.

Of course, my papers were not printed within 3 months, but who cares? Most people download PDFs. If they work with printed copy, they are likely to print their own copies. This is particularly the case in industry.

3) Publishing in a journal is substantially cheaper than publishing in a conference if you include the cost of attending. Thus, journal publications are, if anything, easier for industry researchers. Not having a hard deadline means that legal issues can be dealt with more easily (industry researchers often must get legal approval before submitting a paper).

4) There is no problem with posting papers on arxiv and then having them rejected. If you post a lot of junk on arxiv, then eventually, it will hurt your reputation, but that is your problem. (In time, arxiv could block your uploads too!) If you use arxiv like a responsible researcher, then no ill will come to you. You might argue that “there is too much junk”, yet you are blogging with the rest of us… and there is much junk on the blogosphere.

5) If CS researchers and scientists continue to attend, publish at and found conferences is this not evidence that it is serving a real need?

It almost seems to me that your are advocating conformance: whatever people do right now must be fine because that is what they are doing.

Neal said...


[Cross posted to Daniel Tunkelang's the NoisyChannel]

I agree on point #1... though I'm not sure it's universally true in some CS communities. Aren't the JMLR and MLJ pretty high quality journals in Machine Learning? There are other examples in the AI field. My dissertation in evolutionary computation has loads of journal citations... so I don't 'see' the bad rep of CS Journals.

Agree on #2 and #3. I see myself submitting more to journals for the reasons you raise.

On #5: This is the basic response one should ask of any person advocating to radically change a system that is providing value. Nothing more. Doubt. Ask why and get clear answers and not hand-waving nirvana.

Claus said...

@Daniel Lemire

I think that your argument #1 does not hold universally true. Faculty I have been in contact with strongly favor journals over conference papers. My university has a minimum requirement of journal papers to give someone a PhD. And I have personally received comments from reviewers saying that I should cite more journal papers.

The comment in the ACM article says that the suggestion that conference papers were more important was donequite a few years ago.