Saturday, April 17, 2010

On computer science research publication

I have read one of the recent issues of the Communication of the ACM - as ACM member I am getting the hard copy of this magazine and from time to time I am browsing through it. One of the articles that caught my eye is an editorial article by the editor in chief, Moshe Vardi. It summarized some opinions about publications in computer science, it turns out that most disciplines make journals as the primary vehicle of publications, and view conferences more as a social gathering where most submitted papers are being accepted, while in the computer science area, conferences are a very dominant publication vehicle, and as such, the "good" conferences are very competitive, in fact, people are writing in their CV, what is the acceptance rate of each conference they published in. Vardi's article shows support in opinions that computer science has to mature, and behave like all other disciplines. This observation is true, in my previous incarnation, as faculty member in an inter-disciplinary school at the Technion, I quickly discovered that other disciplines don't really understand the term "refereed conferences", and that conference publication "does not count". One of the wrong (in my view) academic metrics is to count papers, but this is not the topic I am discussing now. From the industry research perspective, the current state of conference-centric, is very comfortable. Since (unlike our academic colleagues) writing papers is a secondary occupation, then it is comfortable to write relatively short papers based on the conferences's page restriction, it is comfortable to tell management that you have a deadline for publication then to dedicate time for undetermined deadline that always gets lower priority to other things that have deadline, and it is more comfortable for management to let the reviewers do the filtering of who should travel to conferences. Despite these facts, I think that getting computer science to act like a mature discipline is the right way to go forward. The journals and conferences will need to be adjusted. Academic journals will also are also going through a phase of coping with the Internet era, and various business models for journals are emerging. The event processing perspective of it is that we should consider to found an event processing journal, when we feel that we have enough quality content for it. The numbers of submissions to DEBS are indication that critical mass might exist, I guess that this will be discussed in some of the coming meetings (like the Dagstuhl seminar),

1 comment:

david said...

Opher,
I think this whole question of academic publication in CS is worthy of a full community discussion.
I believe publication in "refereed journals" is probably still the most important consideration in academic tenure decisions, even in CS.
But the question is whether it should be.
A similar process is often to be seen in funding decisions for research grants, where proposals are reviewed by a committee of ones peers, typically the NSF process for example. Funding agencies like DARPA do things differently, policy being left more to the individual project managers.

The upside of academia is that its very difficult to get rid of tenured faculty. This supposedly gives freedom from political manipulation. Up to a point! After which you are arrested and imprisoned in some societies!

I have been through the tenure process three times in my own career, in Mathematics, Computer Science and Electrical Engineering.
I later resigned my positions on two occasions.

And I know of at least one case where a man denied tenure embarked upon a career in industry and later on endowed a research building at that institution ... but not without a great deal of entreaties from its president!
- David Luckham