About the GPEMjournal blog

This is the editor's blog for the journal Genetic Programming and Evolvable Machines. The official web site for the journal, maintained by the publisher (Springer) is here. The GPEMjournal blog is authored and maintained by Lee Spector.

Tuesday, August 11, 2009

Journal Publication versus Conference Contribution?

In a recent issue of the Communications of the ACM, Moshe Vardi discusses the pros and cons of journal archival publications versus conference contributions. The upshot of his statement, which points to two recent contributions to the viewpoint columns of the journal [1], [2] is that perhaps it is time for Computer Scientists to shift emphasis away from conference and workshop contributions, and start publishing in journal as all other sciences do. A lively discussion followed, see among others, the opinion piece of Lance Fortnow.

As an editor myself of Genetic Programming and Evolvable Machines I have always wondered why it would be more attractive for people in our discipline to publish in conference venues than in archival journals. Are there not enough journals to allow for scientific progress? Or is there a dire need to communicate with colleagues in spatial co-location? Well, to my mind, none of the two! We are not the types of people that wanted to discuss our results to extreme length. Our conferences and workshops usually operate under tight time constraints, and one to three questions is about the average a presenter receives, anything else would eat into the next presenter's time and is discouraged. Also, the number of journals now accepting work from our field has grown over the years to a very reasonable number so that there is no shortage of places where quality work could find a home.

What is it then, that makes us submit and publish so much at conferences? Possible explanations are the existence of deadlines and the incremental nature of much of the work published. The existence of deadlines is a valuable selection pressure in our hectic times where everything is under the dictate of time-driven priorities. It can only be mimicked by journals through the introduction of regular "special issues" which also come with this requirement, and usually are successful in attracting work. As for the second possible explanation, I'd like to cite from [1] on the pitfalls of program committee work: "And arguably it is the more innovative papers that suffer because they are time consuming to read and understand, so they are the most likely to be either completely misunderstood or underappreciated by an increasingly error-prone process." So while innovative work has a harder time at conferences, "our culture creates more units to review with a lower density of new ideas." It is not only that we get to review smaller pieces of work, we are also more busy, with all the workshops and conferences that make us look at these papers. "Genuinely innovative papers that have issues, but could have been conditionally accepted, are all too often rejected in this climate of negativism. So the less ambitious, but well-executed work trumps what could have been the more exciting result." Those would have to be revised and revised and revised again, and there is no time to do this for conferences. Journal articles, on the other hand, can be worked on for a long time, if need be, and there is no time pressure except for the fact that delays could be unbearable and make results obsolete.

In the end, however, it is the impact of the work that counts most. And it is my experience that a carefully edited journal paper is worth the effort, as it produces impact on a scale that conference papers have diffulty to achieve.

[1] K. Birman and F.B. Schneider. Comm. ACM, 52(5) 2009, p. 34
[2] J. Crowcroft, S. Keshav, and N. McKeown, Comm. ACM, 52(1) 2009, p. 27


  1. There's also the fact that people like travelling to conferences. If one could always go to a conference, regardless of having a paper accepted or not, one might choose to submit more papers to journals rather than conferences.

    This being said, I don't see a problem with our field being conference- rather than journal-centric. There's nothing magic about journals.

  2. Togelius: I don't know if I'd call it magic, but the reviewing/editing/revision process is quite different between conferences and journals. In my experience with computer science conferences (although this varies from field to field) once a paper is accepted (often through a one-step review process that overrates safety, as pointed out by Lance Fortnow) the author can submit a final version that will be published without being reviewed further. The author may ignore reviewer comments or even add entirely new, unreviewed material. With journals the path to the initial acceptance often involves several cycles of review, revision, and reconsideration and then the final version is checked again by the editor prior to publication. Coupled with the more flexible length and review time limits that journals provide, I think this makes journals a superior forum in many respects. But of course there are many issues here -- as the linked articles attest -- and as a journal editor I am not unbiased!

  3. While deadlines, as well as all the follow-up review process which is needed to have a work considered for publication in journals, might directly affect the publishing preferences of many computer scientists towards conferences and workshops, these factors alone don't help me to understand why the same effect does not occur in other related areas. In other words, if these were two of the primary causes behind, we would see the same happening among, say, physicists and mathematicians. Yet this does not seem to be the case.

    I sincerely believe that there is one simpler factor that could also be considered in the specific case of CS: the relative lack of maturity of the field regarding consolidated research methodologies and data analysis. Of course, working in a scientific field with less than 60 years do not help us much. We're often borrowing ideas and methodologies from other areas and, still, there's no concensus on, say, which statistical tests are more appropriate.

    So, this may perhaps explain why someone in the CS field might think that it's no use considering to fufill journal reviewers and editors' requests - who are mainly (and correctly) concerned with methodological issues - when one can bypass this process and have his results published immediately in somewhat respectful and prestigious conferences.