Write. ArXiv. Repeat.

As discussed in earlier posts, I believe that two simple and effective ways to improve the publication process are to require that only papers available on the arxiv can be be submitted for publication and to eliminate all formatting requirements. (Throughout this blog I use the arxiv for concreteness only; several other repositories should work just as well.) In this post I want to consider a broader, radical publishing reform, and discuss several related issues.

Here’s how I would like the publication process to be:

As an author, you write your paper. When you are done, you post it on the arxiv. Period. You now move to your next paper.

Forget reverse-engineering the chronology of progress: there would now exist a unique citation for your paper: its arxix entry. Forget bibtex and its BeaST. And also forget trying to pick the best venue. Forget “are they going to invite me for the special issue? In fact, is there even going to be a special issue?” Forget the conference vs. journal debate. Forget a lengthy camera-ready production process whose goal is to put your paper in an electronic format that is only read by library computers.

Papers would be ranked by a system of badges. For starters, the badges will correspond to the current entities. So we have the STOC badge, the JACM badge, etc. We also have some badges like ECCC, which are assigned to papers that satisfy minimal requirements, such as not making sweeping unsupported claims. Badges cannot be removed but can be added. This last aspect makes the new system more flexible. Today, it is a bit funny to find out that a seminal paper appeared in an obscure venue, but it is hard to update that paper’s status. With the new system one could just add another badge.

Q: Which papers are the committees supposed to evaluate?

A: Committees will need to monitor papers like many people already do. Note that for the year 2014 the ECCC repository lists 184 reports, for the year 2013 191. These are fairly small numbers, comparable to the number of submissions to a top conference.

Q: What if a paper does not get noticed? What mechanism would there be for giving it additional chances?

A: The current default mechanism is that the author resubmits the paper, to signal that venue’s committee that they should give the paper an n+1 chance. The same can be done with the new system, for example by posting an arxiv revision with the comment “no changes from previous version”. In both systems, what prevents authors from flooding committees with resubmissions is the reputation loss, so I expect this aspect to work in roughly the same way. A more rare current mechanism is that the paper gets invited or is selected for an award. This would work in the same way in the new system.

Q: What about the cycle of getting feedback from the reviewers and revising the paper accordingly?

A: In the spirit of “only papers available on the arxiv can be be submitted for publication”, I would like the public to have access to the same information that is given to authors and referees. So I would like this cycle to take place in a public forum. If there is a serious issue with an arxiv paper, I would like to see a comment pointing this out right away, instead of having to wait for authors and referees to converge on a new version to release to the public. I also believe that the feedback/revision cycle is less prominent in theoretical fields than it is elsewhere. In other fields, it is common to receive feedback of the type: “Result X is interesting but not enough. Please run experiment Y. If you get outcome Z then we’ll talk.” With theoretical papers you hardly get a request for obtaining better results. If there is any feedback it is mostly about presentation, references, and correctness. Also, especially with conferences, it is not uncommon to get inessential feedback.

As a first step towards implementation, one could keep the publication venues as they are, but replace the cumbersome submission process with an email containing a link to the arxiv record. The production of a camera-ready version and copyright transfers are eliminated. Conferences going solo have a great opportunity to implement this first step. Alas, the Computational Complexity Conference did not quite go for it; think about it next time you get an email about an overfull hbox. Once this step is taken, one can ask if even the submission email is required.

NSF now requires its grantees to make their peer-reviewed research papers freely available within 12 months of publication in a journal. This move by NSF is the answer, at least in part, to this petition, which I signed. (Incidentally, my inability to advertise this petition through the available channels is one of the factors that eventually led me to start my own blog.) However, I don’t find this change very significant. For one thing, 12 months after the publication time is a very long time for research.