ACL Submission Policies – Whence and Whither?
As NAACL 2018 is the first conference to apply the new ACL policies for submission, review and citation, there are naturally many questions about how and why these policies were put in place. The NAACL Program Chairs have therefore asked me, as President of the ACL, to give a little background on the new policies. In doing this, I will concentrate on describing the process that eventually resulted in the ACL Executive Committee adopting the new policies. For the rationale of the policies themselves, I have little new to add and instead refer to the report of the working group that drafted the policies.
Our models of scientific publishing are constantly evolving, and there has been a lot of discussion recently about the increasing impact of preprints and the way this interacts with the double-blind reviewing that has been the norm for conferences and (most) journals in our field. The ACL Exec has been following this discussion for several years – my predecessor Chris Manning highlighted some of the issues in his presidential address in Beijing in 2015 – and during the spring of 2017 it became clear that there was a need to review our policies and guidelines. In discussions on the ACL 2017 Program Chairs’ blog and elsewhere, opinions clearly diverged on the pros and cons of preprint publishing and of our current reviewing model, but there seemed to be general agreement that the existing ACL policies and guidelines needed to be at least clarified and possibly revised.
In order to gather more information on current practices and opinions, we organized a survey in the community. The full report on the survey is available in the ACL Portal, but here are the most important conclusions that we draw from the survey:
There is strong support for maintaining double-blind reviewing for our conferences (and for doing what we can to protect its integrity by discouraging non-anonymized preprint versions of submissions).
There is only weak support for completely banning preprint versions of conference submissions (although a majority say they are prepared to live with it).
Many people think the policies and guidelines for submitting, reviewing and citing preprint papers need to be clarified.
The results of the survey were presented at the ACL conference in Vancouver in July-August, first in my presidential address and later in the ACL business meeting. The business meeting also featured a discussion with invited position statements from a panel consisting of Jason Eisner, Chris Manning, and Paola Merlo, followed by questions and statements from the floor. Immediately after the conference in Vancouver, the ACL Executive Committee appointed a working group to draft new policies and guidelines. The working group included the Program Chairs for NAACL 2018 and ACL 2018, editors for the CL and TACL journals, the members of the panel mentioned above, and three representatives of the ACL Exec (with me as chair).
The working group delivered its report and recommendations in early October, and the new policies and guidelines were approved by the ACL Exec shortly afterwards. The new policies will be applied at all ACL, NAACL and EACL conferences, starting with NAACL 2018, and they have also been adopted by the TACL journal. Other conferences and workshops are free to adopt the policies as well, and conferences and workshops that are co-located with ACL, NAACL or EACL are encouraged to do so if they use double-blind reviewing (as long as this does not conflict with other policies they have).
The content of the policies can to a large extent be seen as clarifications of best practice for citation and review in the rapidly changing landscape of scientific publishing. The main novelty is an anonymity period, which prevents authors from posting or updating non-anonymized preprint versions of their submissions during a period starting one month before the submission deadline and ending at notification time. (There is no restriction on anonymized preprints.) The purpose of this policy is to eliminate some of the preprints that directly compromise double-blind reviewing and where the only inconvenience for the authors is that they have to wait a short time before posting them. This is in line with the view that submission and reviewing should be organized in such a way that double-blind review is maintained and protected without sacrificing the positive effects of preprint publishing. Based on the survey results, we take this view to be shared by a broad majority in our community.
For further motivation and discussion of the new policies, I again refer to the report of the working group. Needless to say, the effect of the new policies will have to be monitored and evaluated, and the ACL Exec welcomes feedback on all aspects of our publication and review processes. Meanwhile, we continue to work to facilitate scientific communication without undermining the quality of our review processes. To mention just one example, we are currently engaged in a dialog with the scientific board of arXiv about embargoed authorship, which would allow authors to post preprints that remain anonymous for a limited time period while they are under review, a functionality that currently exists only on some platforms like OpenReview. I hope you will all join us in the work to further improve our models for review and publication.