We've seen that go terribly wrong in other fields, and in the ML field.
I think a system where area chairs skim abstracts and solicit these ethics paragraphs where they seem necessary -- so, not for the billionth paper obtaining pretty bounds for some variant of convex optimization -- makes more sense.
I know that "think of the machine learning researchers!" is not a take that engenders sympathy, but we're probably talking thousands of human hours spent writing these things. It's not nothing.
But the subtext of this letter goes fairly deep (and to be clear this isn't meant as a response to you specifically):
- While it doesn't specifically say it, this letter does appear to be at least somewhat anti-ethical concerns in general, which isn't good.
- At least one of the signatories of this letter has not, in my opinion at least, followed the guidelines in this letter. Insinuating that a colleague of holds an opinion because they watch too much online porn is in no way civil, and is absolutely a personal attack.
- Given the "disagreement" between Domingos and Anandkumar, who as others mention, sits on the ACM editorial board, this could be an attempt to censure or "cancel" her for her personal views. This is antithetical to the values held in the letter itself, and leaves a sour taste.
Shouldn't her actions at least be censured? Her behavior has corrupted the science community, limited speech, created a culture of fear, and is ultimately inhibiting others' from freely disseminating their findings to the rest of the community (at NeurIPS and elsewhere) unless the author and/or their work passes a political purity test, effectively. If that's not discouraged, then I think science and open discourse will continue to be under attack.
What specifically has she done that "corrupted the science community", that "limited speech", that "created a culture of fear", and that is ultimately "inhibiting others from freely disseminating their findings"? List the particular actions she took that did such things. I don't think there are any.
The closest you can get, I feel, is that Domingos and perhaps some anonymous others felt that her list of people who she wished to educate and discuss with was going to be used to attack them. This never actually happened though. Domingos claimed that the list would be used in such a way, and then it was deleted.
I could just as easily claim that Domingos created a culture of fear by making hyperbolistic claims about the things that Anandkumar would do.
Don't know what you were hoping to show with this comment, maybe you were unaware that there were archive links of her tweets.
In any case, Domingos wasn't wrong, yet for some reason you seem far less willing to give him the benefit of the doubt, while asking others to do so for people you seem to agree with. Even in this case, for a person who clearly doesn't deserve it. I am curious as to why.
And yes, I think this is pretty clear since in other contexts Anima notes that she's trying to use this as a list of people to engage with and educate, and someone asks well what if I don't want to spend effort educating people, and this is her response. The idea that she should spend time trying to educate and engage with these people, but if you don't want to do that, you should try and get them fired doesn't really make much sense.
> In any case, Domingos wasn't wrong
I certainly think he was, and unlike Anandkumar, he's made no attempt to apologize and continues to grandstand.
Did you see anyone harass anyone on the list? Was that behavior condoned?
What part of that suggests harassment? I saw more than one case of positive interactions between Anandkumar and others who initially disagreed or were on the fence who then discussed things offline. That certainly didn't appear to be harassment.
In response to your below comment:
> The part you didn't quote. The part about "re-educating" them
The full tweet I'm quoting is
> I am looking for volunteers to try and change the minds of fanboys of Pedro (real people and not bots, everyone I saw is male). Especially junior people. We need to get them away from fanaticism. I can share my blocked list. We need #ALLY who can engage with them. Please DM me.
Heck, in another tweet she noted "My blocked list is not meant to be punitive."
There's nothing about "re-educating", and even searching up discussions here and on reddit that contained quotes of her tweets, I couldn't find anything mentioning "re-education", except your comment and one other in this thread. Neither references an originating tweet, and, not for lack of trying, I can't find anything on waybackmachine that includes the term, and I'd have assumed that someone would have taken the time to either comment on it or capture it at the time.
Could you please point out her use of the term, because I'm unaware of it. So the rest of your comment doesn't make much sense as it appears to be based significantly on words that weren't actually said, and further based on subtext and implications of words that, again, were never actually written.
> And if all that is fine, why did she apologize?
Perhaps because she's willing to apologize when she's made others feel uncomfortable. That's essentially what her apology says.
So: First try to re-educate them, and if that doesn't work, then cancel them. Said to a relatively large group of supporters.
And if all that is fine, why did she apologize?
I saw "re-education" in (at least) one of the posts here. It seemed to be a direct quote. I took that as being accurate. I have not gone groveling through every possible tweet that she sent to check the wording.
So either the quote was accurate, in which case it was bad wording (at least) and the apology is perhaps a bit lacking. Or the quote isn't accurate, but is someone's projecting their own opinion on what she said, and I'm a sucker for falling for it. Naturally, I'd prefer to believe the former, because it makes me not the bad guy, but... you very well could be right.
I am not so sure the letter attempts to cancel Anandkumar. Although her blacklist clearly went too far, it's quite difficult to find the context that led to the letter (as several comments show in the thread).
The sad truth is that expressing publicly any opinion on diversity/politics/AI ethics is currently a minefield for anyone's career and online safety. It shouldn't be the case and that's how I understand the letter.
> challenging and debating ideas is always acceptable and ought to be encouraged. Marginalizing, intimidating, or attacking the holders of those ideas is not.
If anything, that statement reads as supportive of Gebru. But I don't know.
For a while he had "anti-woke" in his Twitter bio, and his recent tweets[0] are lamenting AI ethicists and negatively commenting on how "science is under attack from politics"[1]
[0] https://twitter.com/pmddomingos
[1] https://twitter.com/pmddomingos/status/1341966137576198144
Ultimately though, they're irrelevant to my point, which is that even if you personally dislike the conclusions of the paper, even if you disagree, it is an existence proof of thoughtful and deep analysis of broader impact.
You are more than welcome to dislike the concept. But it still remains present.
If they didn't have an amazing canned answer it's because they weren't thinking about it enough to have something interesting to say.
Maybe not needed for every venue every year forever, but I like this better than requiring it just from ethics sensitive papers. Presumably the people writing about swarms of killer drones have some justification or are fine with being supervillains, the field is trying to figure out where the other blindspots are and that requires broader search efforts.
Edit: I just want to add this is the first I'm hearing about this letter and I'm not entirely sure what it's even in response to. Since my work is health adjacent, I haven't had time this year to know much about conferences I wasn't involved in.
As far as I know, those are for experiments that directly interact with humans as part of the experiment itself. They exist because of a long history of unethical researchers testing interventions (or lack thereof) on people without their knowledge or informed consent - e.g. Tuskegee Syphilis Experiment, MK-ULTRA, Stanford prison experiment (there are a lot). The closest equivalent in CS of directly-interacting experiments would be HCI research as you said. I could also see a strong argument being made for research that uses the creative or copyrighted output of a person without their consent - for example, their face as part of training facial recognition software, their written words as part of training GPT-3, their voice as part of training voice recognition software, etc.
However the incident in question is really about a different kind of thing - it's asking researchers to speculate on the future ramifications of their research as it pertains to progressive ideals - in practice this means, "How could this negatively affect minorities or the environment?" These aren't inherently bad things to think about, but as you get further and further away from concrete applications of ML, it begins to look more and more like a religious ritual than something that is actually trying to address the stated problems.
Or use certain kinds of data from humans. For example, a machine learning project on medical data (even anonymized medical data) probably needs an IRB approval to gain access to, say, a dataset of cancer imaging. A facial recognition model doesn't. What makes those fundamentally different (arguably the facial image is more identifying).
> However the incident in question is really about a different kind of thing - it's asking researchers to speculate on the future ramifications of their research as it pertains to progressive ideals
I see no actual support for this claim in any of the actual documentation on the broader impact statement or its use.
The Neurips docs link to 4 blog posts and papers for inspiration/advice to a broader-impact-statemet-author[0][1][2][3][4]. None focus on minorities or the environment in the way you describe, unless you mean "environment" in the broadest sense of "society".
[0]: https://nips.cc/Conferences/2020/PaperInformation/NeurIPS-FA... ("How should I write the Broader Impact section?"_
[1]: https://brenthecht.com/papers/FCADIscussions_NegativeImpacts...
[2]: https://brenthecht.medium.com/suggestions-for-writing-neurip...
[3]: https://medium.com/@GovAI/a-guide-to-writing-the-neurips-imp...
>it's asking researchers speculate on the future ramifications of their research as it pertains to progressive ideals - in practice this means, "How could this negatively affect minorities or the environment?"
The framing appears to be far broader than just that, and the "ritualistic" concern you note appears to be unsubstantiated given the other areas for broad impact, and in practice given that for a lot of concrete ML applications, the broader impact statements appeared to be mostly focused on things like efficiency (or just platitudinal), as noted by other commenters.
If they're just platitudes or remarks about efficiency, why are some people so strongly opposed to including them, and some people so strongly in favor of including them? Basically, why does anyone care?
And a lot of people gave a huge amount of backlash to this experiment for, as far as I can tell, reasons that amount to this being a form of liberal/leftist censorship.
We're talking about, in this case, at least one person who is so scared of "wokism" that after Neurips changed its name from Nips to Neurips, he advocated for them to change it back, to, uhh, stick it to woke people I think?