Comments on the social media motion at CVPR 202115 Jun 2021
Written by Amy Tabb and Dmytro Mishkin, June 15, 2021.
Recently we have seen a motion to be voted on during the PAMI TC meeting at CVPR 2021. The idea is the following:
Social-media promotion of papers is prohibited during the review period. Explicitly excluded from this ban is the automatic posting of new preprints by arXiv.
And the motion, background and a FAQ are available here (version avaliable prior to CVPR, Google Doc), version for the vote, DropBox.
We are opposed to this motion. The problem, that the motion aims to solve, is the bias in peer review problem. However, we predict negative consequences to the communication within our community that will not improve the bias in peer review problem. In addition, the motion as currently written is not well-defined (see the Appendix at the end of the post) and introduces new, additional ambiguities and loopholes to existing processes.
The largest negative communication problems we predict are:
Extreme paranoia and finger-pointing among people, who respect rules, about collusion rings, where a set of authors has gotten another set of authors to post on their behalf. There are already some concerns about collusion rings in review [1, 2, 3], and to move these concerns into yet another realm is not productive.
A chilling effect on talking about work under review, whose natural progression will be not to post on arXiv at all except prior to an anonymity period, similar to what the *ACL conferences do. We have already written extensively on why the availability of arXiv is important [ 4, 5, 6]; the ability to talk openly about one’s own work to other scientists as well as the general public, no matter the medium, is equally important.
People, who do not respect rules and were to find loopholes, would benefit much more than they have now from promoting their work. For instance, anonymous social media accounts could be used to evade this policy. MachineLearning on Reddit is often called toxic and one of the stated reasons is anonymity. Do we want to have it on Twitter as well?
Finally, this motion may be focused on large labs or people with a large number of followers, but it will penalize small labs the most – the same as with a possible arXiv ban. We believe that science and the community would not benefit from a restriction on paper discussion in any form and ask PAMI TC members to vote against this motion.
What do we mean that “the motion is not well defined”?
Questions that arise from reading the motion that are not answered in the FAQ:
- What constitutes social media? Are YouTube and Twitch social media, given that many people live-stream coding sessions there now? WeChat? Clubhouse?
- If another person mentions your paper on social media, and you re-share / re-tweet their mention, is this promotion of the paper and a violation of the proposed policy?
- An author provides a link to their paper as a resource to a talk, is this considered a violation of the proposed policy?
- An account with a large number of followers profiles your paper. Is this a violation?
- A scientist in medicine asks an author a question about their paper on social media, that has implications in medicine. The author answers on social media. Is this a violation?
- Is liking / favoriting a social media item that mentions your paper considered promoting your paper, given that some social media algorithms promote content to networks of users?
- How would this policy be enforced? A dedicated team? What about fraud with screenshots? If a post is deleted quickly, does it count?
1 Michael Littman. Collusion Rings Threaten the Integrity of Computer Science Research. Communications of the ACM, June 2021, Vol. 64 No. 6, Pages 43-44. 10.1145/3429776
2 Reddit. Collusion rings, noncommittal weak rejects and some paranoia. Jun 11 2021.
3 Jacob Buckman. Please Commit More Blatant Academic Fraud. May 29, 2021. jacobbuckman.com .
4 Dmytro Mishkin and Amy Tabb. Hands off arXiv! June 29, 2020. amytabb.com .
5 Dmytro Mishkin and Amy Tabb. What does it mean to publish your scientific paper in 2020?: Benefits to authors of non-anonymous preprints. August 21, 2020. amytabb.com
6 Dmytro Mishkin, Amy Tabb, Jiri Matas, “ArXiving Before Submission Helps Everyone,” October 2020. arXiv:2010.05365[cs.DL]
June 25, 2021: This motion passed at CVPR 2021, vote tallies: Walter Scheirer, Twitter.
- June 27, 2021: added link to the final version of the motion for the vote at CVPR 2021 from DropBox.
- June 27, 2021: added the outcome of the vote.