Wednesday, April 28, 2004

Conferences versus Journals

A reader asks why Gafni and Borowski did not publish their paper in a journal and become eligible for the Gödel Prize. I wish this was an isolated incident but it reflects on a sad state of affairs in computer science and theoretical computer science in particular. Too many papers in our field, including many great ones, do not get submitted to refereed journals. In an extreme case, Steve Cook received the Turing Award mostly for a STOC paper.

In most cases, conferences in computer science are more selective than journals. Your reputation in theoretical computer science is measured more by the conferences your papers appear than the journals. In other fields like mathematics, physics and biology, journals have a much greater reputation and most of their papers do appear in refereed form. I believe the reason is historical: computer science started as a quickly changing field and journals could not keep up with the rapidly emerging ideas.

Conference program committee cannot and do not produce full referee reports on conference submissions. Proofs are not verified. Papers are not proofread carefully for mistakes and suggested improvement of presentation. Computer science suffers by not having the permanency and stamp of approval of a journal publication on many of its best papers. The founders of the Gödel Prize put in the journal requirement to encourage potential award winning papers to go through the full refereeing process.

Many papers in our field do appear in journals and some researchers are extremely diligent in making sure all of their work appears in refereed form. Also I know of no computer scientist who purposely avoids sending their papers to a journal. But when we have a system that does not value journal publications, a computer scientist pressed for time often will not make the effort to take their papers past the conference version.

No comments:

Post a Comment