Crowdsourcing Scientific Paper Review ∗ WORKING PAPER

The current system of peer-reviewed publications is governed by a cycle — write a paper, submit it, have it undergo review, and be published in an archival journal or conference — that was shaped by the medium used for disseminating the results. Printing and shipping journals or conference proceedings is expensive, and takes time; this gave rise to a publishing process with long lead times and a review process centered on selecting, rather than ranking or commenting. Information nowadays can be disseminated immediately essentially at no cost, and furthermore, in a manner that makes it open to social interaction: in blogs, wikis, forums, social networks, and other venues, people can both share and comment on information. As the medium and its capabilities have changed, so, we argue, should the process of scientific peer reviewing and publishing. The current process has several drawbacks. One of the most salient is the delay imposed on the dissemination of results. In a typical computer science conference, six months may elapse from submission to publication in the proceedings, and this assumes that the conference deadline came just when the paper was ready for submission, and more importantly, that the paper was accepted. To avoid this delay, many authors submit the paper to open repositories such as Arxiv1 at the same time as they submit it to a conference or journal. While this makes it available to