CVPR authors abandon double-blind review

So it seems (and I have no way of checking) that some CVPR submissions are being trailed in the New York Times the day after the submission deadline.

The implication is that the authors and the community are in favour of abandoning double-blind review.

The problem is, though, that this is not our choice to make. By “our” here, I mean those of us who are famous authors at famous institutions. Bigshots if you will. Of course, there are arguments for and against double-blind, but we (the bigshots) are those whom it is designed to protect against, so if we unilaterally abandon it, we strengthen any case for its existence.

Note that this is not a question of what the CVPR conference chairs decided. This is a decision each individual makes when they post their unblinded paper on any indexed website.

Rekindling this on twitter… https://twitter.com/Awfidius/status/1346948648249282560

Researchers with privilege: here's one thing you can do this year to balance the playing field for emerging researchers. DON'T VIOLATE DOUBLE-BLIND REVIEW. This is something YOU can do, not something that is decided by conference committees, or multilateral discussions.

If you submit to any venue where your paper is anonymized, don't post it on arXiv during the review period. Don't put it on your webpage. Wait.

If you're giving a talk at any venue during the review period, you can mention the new paper, but you must ask everyone present to recuse themselves from reviewing it. If you already submitted the paper to a non-blind venue, or gave a talk in a public forum, sorry, you can't submit it to blind venues.

If you're giving a talk at any venue during the review period, and talk about the new paper, you must ask everyone present to recuse themselves from reviewing it.

Q: “But the conference website clearly says it's OK” A: Doesn't matter, this is a moral position YOU are taking, to check your privilege (pun intended). That website wording was written by people just as confused as you used to be. Just think it through yourself – the following handy FAQ may help.

Q: “But what about science? Progress will halt if we wait three months to disseminate our fantastic idea” A: Oh I can hardly count the ways this is wrong. First, are you really that important? OK, yes, I see you are. So if it’s that great an idea, just publish it. Don't wait for peer review. Put it on arXiv, tweet about it. If it's that great, folk will take it up, cite it, and you don't need the NeurIPS paper. And if it's wrong, folk will learn that over time, and won't read your next paper quite so trustingly.

Q. “But I only gave the talk at a tiny conference in Cork.” A. First, I'm from Cork, so be careful what you say. Second, great, lucky you, lucky Cork. Was the talk recorded? Can I see it on de internet? If so, you have chosen not to avail of blind review. Don't worry, the Cork people will definitely spread the word. Particularly if it was wrong.

Q. “OK, I am privileged, I'm a full professor at MIT. But I can't commit to this, my students are not so privileged.” A. Well, let's explore that. Your students will suffer because they had to wait 3 months to publish? So they are at the very end of their PhD? Then see “job talks” below. Otherwise see “science” above.

Q. “But I need a NeurIPS acceptance to get my PhD” A. Ah, well then. You should definitely not violate double blind review before submitting. Just submit, get the reviews, and then release the paper.

Q. “But I need a job, and this work is the best I've done – I need to talk about it.” A. Absolutely you do - and you should. But tell the audience that you're talking about work under blind review, and that they should decline any requests to review it if it comes their way, and that they should politely ask for it to be reassigned if it's already on their stack. If your job talk is a week before the PC meeting, this will be tricky, I'll do the full flowchart for you some time. The point remains for the 99.9% of other situations.

Q. “But, but, but.. mumble mumble scooped.” A. Yes, that's likely to happen if you're doing the blindingly obvious next thing in a crowded field. Don't do that. Do something harder. Embrace scoopage – it means you can save time writing up that trivial stuff, and get on with the real science.

Q. “But Andrew, didn't you hear? There's no need for double blind – that's all hokum. Me and my good ol' boys always saw through it, and nothing bad happened.” A. Oh, of course, you're right. Forget I said anything. Privilege is fake. Double blind is only for weaklings. Silly me. But no, seriously: people of privilege, you are exactly the people who are not permitted to say double blind is optional.

Q. “But I'm not privileged, I'm a lowly PhD student.” A. Where? MIT? You may be privileged. Q. “No, it's not MIT. It's a terrible place. Near Cork.” A. Are your parents rich? You may be privileged. Particularly if you say “it depends how you define rich”. Q. No, we were so poor I had to watch TV in the dark. A. Do you have a platform? Hundreds of twitter followers? Etc… My point is, you might be.

no-more-double-blind-cvpr.txt · Last modified: 2021/01/07 09:37 by awf
CC Attribution 4.0 International
Driven by DokuWiki Recent changes RSS feed Valid CSS Valid XHTML 1.0