On missing SIGIR

Just in time to justify my decision not to attend SIGIR comes another CACM article calling for an end to conferences as the primary publication venue for Computer Science. This one, by Lance Fortnow of Northwestern University, appears under the conciliatory, consensus-building title of Time for Computer Science to Grow Up (thanks to Shane Culpepper for forwarding this to me; a pre-print version is available for those without ACM access). Fortnow argues that conferences became central to publication in Computer Science due to a mix of historical coincidence (the field emerged during the jet age) and the fact that conferences used to offer the fastest turnaround for publication, important to an immature and rapidly growing field. Jets are still with us, but the seven-month delay between submission to and appearance at conferences no longer looks so rapid. Indeed, the conference timetable is actively slowing the pace of disemination of ideas, since many authors are reluctant to release onto pre-print servers like Arkiv a paper that has been submitted for conference reviewing -- especially if the reviewing is double-blind.

I've had sufficient conference publications and done sufficient conference reviewing to have experienced first-hand one of Fortnow's other criticisms of the conference publication system, namely the rushed publications it elicits from researchers, and the rushed reviews it obtains from program committee members. Just today I received a panicked request for inter-reviewer discussion of a borderline CIKM accept from a harried meta-reviewer, the day before the deadline for final announcements to go out! Researchers for their part work like medieval peasants, their schedules controlled not by the logic of their own projects but by endlessly revolving yearly cycles -- a SIGIR submission in January, a CIKM submission in June, a few weeks feverish activity before each, and often not much done in between. I've also read more than enough serial publications, where the results of a single research project, which would fit nicely into one journal article, are drip fed through three or four different conference publications. And this is not to mention all the other problems with the conference system as it currently works.

How is this going to change? I gather from Twitter that at the ACM business meeting at this year's SIGIR, the idea of a remote-only SIGIR conference was floated. But I think this is the wrong way around. The first step is to separate SIGIR publication from SIGIR conference attendance. This is being done at VLDB. There will be a quarterly, fast-turnaround journal publication, with mechanisms for reviewers to request amendments and authors to resubmit with memory of previous reviews. Papers for presentation at the annual conference will be selected from those published in the journal, and presentation will be by invitation.

Why could this not be done for SIGIR, too? What good reason is there not to adopt this model?

One Response to “On missing SIGIR”

  1. S. Very says:

    The full name of the new VLDB journal you refer to is "Proceedings of the VLDB Endowment". SIGIR may just lack the endowment for such an initiative - who knows?

Leave a Reply