Evalautaion of CS papers for GEB

Games and Economic Behavior (GEB) is the leading academic journal in Game Theory, and as such is a natural place for submission of your Algorithmic Game Theory papers.  (Other natural alternatives include TCS journals, AI journals, Economics journals, and OR journals.)  GEB has been very “CS-friendly” for quite some time, to the point of having added several CS researchers to its editorial board.

Judging papers on an interdisciplinary border is always a tricky business, and in the case of CS there is an additional complication with the conference publication culture in CS.  GEB has recently undergone some internal discussions reaching general guidelines for evaluation of CS papers for publication in GEB.  These have been written by Editor David Parkes and recently sent to the editorial board.  With his permission, I am publishing them verbatim.

The evaluation of CS papers for GEB: Guidelines

* On the Role of GEB

To be published in GEB, a paper should be of interest to game theorists; for CS-themed papers it should meet the standards set by the relevant leading field journal. Papers should be original, make a substantial contribution, and provide a broadly accessible exposition.

For CS papers in particular, a unique role that GEB plays is in certification that a piece of work is correct and is of interest to the game-theory community. This is important in meeting GEB’s mission of communicating game theoretic ideas across disciplines.

For the moment, many submitted papers will have previously appeared in “quasi-archival” CS conferences, such as FOCS/STOC, AAAI/IJCAI, EC and WINE. These are heavily reviewed and high quality venues, but publication does not certify interest to the GT community, nor necessarily the correctness of proofs, and the final version may also differ from a reviewed version.

* Originality of Contribution

In considering the role of GEB in certification of interest and correctness and its mission of communicating ideas, and the role of conferences within CS, we can expect the incremental contribution of the GEB paper over a CS conference paper to be much less than that required in other fields.

But the stated policy will remain that there should still be some incremental contribution, judged to be of some technical significance and game-theoretic interest. Authors should be expected to discuss differences from a conference version in a cover letter.

There has been an extensive discussion on whether or not to require an incremental contribution over a conference paper and there is not unanimity here. But this policy seems to best balance considerations about this role of GEB, with other considerations (e.g., that the GEB version be the reference version, and in making efficient use of reviewer resources.)

* On Scope

GEB should be a welcoming place for appropriate papers at the CS/GT interface, and we need to be as clear as possible in communicating with authors about what is considered in scope. It is this issue about which we have received most questions in the past couple of years.

The journal aims to communicate game-theoretic ideas across fields and applications, and CS constitutes a significant group of contributors to both GT and its applications. But what about a paper that doesn’t contribute to GT or its applications (i.e., it does not develop new game-theory per se, no new game-theoretic approach, does not test existing GT, etc.), but rather is computer-science centric in its contributions, be they theoretical, computational, or algorithmic?

This is a difficult area, but a broad guideline is that papers should be in scope if they are computational, but on a topic “of interest to game theory.” For example, new results related to core solution concepts, equilibrium analysis, or pertaining to issues of bounded rationality, would be in scope. On the other hand, results on more applied problems, such as faster algorithms for winner determination in auctions, or for clearing prediction markets, may be out of scope and may be better submitted to a CS or OR journal.  And as already stated, to be acceptable to GEB a paper must also be of a quality acceptable by the relevant leading field journal.

* Most importantly, be Flexible

Again, our overall mission is to communicate GT ideas across disciplines, and the guidelines above were designed towards this goal. But we suggest that every paper be judged with the overall mission in mind, even if it does not meet the guidelines above.  A reviewer should weigh all relevant variables: the importance of the results, the nature of the earlier publication (the refereeing process, visibility, the level of exposition in terms of technical details and applicability, etc.), the marginal contribution to knowledge in the GEB publication beyond the earlier publications, and so forth.

Advertisement