No need for substantial commentary, here are emails he’s sent to large groups of people within UW CSE, on UW mailing lists.
In response to an article with evidence that Machine Learning algorithms that rely on biased training data are biased and may perpetuate social injustices (e.g., when used for law enforcement):
No, [other professor] is wrong. This is a well studied issue. If an ML system is designed to maximize accuracy in predicting (say) criminality and there is a correlation between criminality and skin color, it’s quite likely that there will be more false positives of some skin colors than others. It’s not discrimination. Rather, the problem we have today is that some people are projecting onto ML systems biases that they don’t have. Attempts to make ML “fairer” often wind up having the opposite effect. This is also a well studied issue. What ML really needs is protection from attempts to politicize it.
P.S. Some references:
https://www.nature.com/articles/d41586-018-05469-3
https://arxiv.org/abs/1609.05807
https://arxiv.org/abs/1703.00056
https://arxiv.org/abs/1803.04383
(BTW, [other professor’s] reply is a good example of the politicization I’m talking about. He wants to replace accuracy, the standard evaluation measure in ML, with something more in agreement with his beliefs. He thinks there’s a problem with the officer/stoppee example, when in fact there isn’t. He points to the volume of research on fairness in ML, ignoring that much of it is politically motivated — whether the authors realize it or not — and of doubtful quality. We all have a choice between sinking further into the echo chamber or actually paying attention to the facts. I hope that as scientists we take the lead in doing the latter.)
Responding to a poorly argued article about why there aren’t (and shouldn’t be) more women in Computer Science:
Or rather, I would ask folks to keep the discussion focused on the issues raised by Stuart’s article, e.g., whether our department’s approach to diversity needs to change to be fairer to everyone, including men.
If Stuart’s article is correct, then one implication is that current policies aiming to boost recruitment of women beyond what is justified discriminate against men. (I’m not saying this is a problem specifically in our department, but it certainly is a prominent one nationally — e.g., the Damore case that Stuart cites.)
I highly recommend reading the original Damore memo, which is precisely about discrimination against men in promotion and hiring at Google (a problem which you may yet have to face if you go to work there, or at many other tech companies):
https://assets.documentcloud.org/documents/3914586/Googles-Ideological-Echo-Chamber.pdf
I hate to disappoint you, but I’ve seen many events that could be viewed as examples of bias against men in our department, and heard complaints of more. (People are understandably reluctant to voice them publicly, and even writing an editorial like Stuart’s takes tremendous courage. Future generations will thank him.)
Race and gender-based discrimination can take many forms besides explicit quotas. You’re correct that affirmative action is illegal, and I hope everyone knows that. Whether we are within the law is something I’ve seen reasonable people disagree on.
https://assets.documentcloud.org/documents/3914586/Googles-Ideological-Echo-Chamber.pdf
To reiterate, because it doesn’t seem to be getting across to some of us, the Damore memo that motivated Stuart’s article is about discrimination against men. We can declare that it doesn’t exist and that we should only focus on women’s concerns, but that doesn’t seem wise, or indeed fair. It’s a little disconcerting (Orwellian, even) that a plea to not discriminate would be viewed as discriminatory.
One of Stuart’s (and Damore’s) main points is that perhaps there aren’t more women in CS because they’re not interested, and we shouldn’t force them to be. We can press on regardless, but whether that’s best for either women or men is a question that surely deserves consideration.
When a woman is hired or admitted over a more qualified man, the man was discriminated against. Referring to general notions of oppression, benefits, etc. doesn’t make it legal or ethical.