I’d be delighted if someone could explain to me how it’s even in the hypothesis that schminux could be considering himself privileged or having some sort of advantage for making or that lets him make the observation “slow/cautious drivers are 80%/mostly female”.
Sure. Take the outside view. Male commenters on the internet have some distribution of attitudes towards gender. Some of those attitudes are more or less likely to lead to comments like “most slow drivers are female.” One of those attitudes (simplifying considerably) is having and being unaware of male privilege, call it M. A comment C on the internet is Bayesian evidence for M if and only if M-commenters are more likely to make comments like C than non-M commenters. My claim is that a comment like shminux’s is more likely to come from someone who has and is unaware of their male privilege than from someone who is at least aware of their male privilege.
In this particular case, you have inside view reasons to believe that shminux’s statement was generated via some other process than the one I’m suggesting that shminux’s statement provides weak evidence for (emphasis on the weak). There’s no contradiction here.
Forget what I said about the typical outside female observer. The point of that device was to 1) enforce outside view and 2) make gender issues more salient.
I notice schminux originally didn’t say he makes the observation aloud, only that he observes and thinks. Now the claim is apparently that he says this in public
He did say it in public. It’s right here on LessWrong.
It’s now obvious to me that the outside viewer does see P(M|C) > P(M|¬C), and why.
However, what schminux gave is not C, but R(C) + Q, and to any attentive reader I assume that P(M|R(C)+Q+C) < P(M|¬C) < P(M) < P(M|C); where R(C) is a meta-observation about one’s own (past?) observations/comments C¹ , and Q is a question about the evidence carried by C and R(C).
Any external reader who misses this is, as far as I can tell, simply wrong. They were most likely themselves taken by confirmation bias or some other undesirable effect.
Confirmation bias among self-proclaimed feminists is actually something I have a rather high prior for, since they train themselves to see gender bias everywhere in many cases.
(and R(C) does include C as a component once unwrapped, but for clarity I added C in the chain above; assume appropriate anti-double-counting measures are taken)
Sure. Take the outside view. Male commenters on the internet have some distribution of attitudes towards gender. Some of those attitudes are more or less likely to lead to comments like “most slow drivers are female.” One of those attitudes (simplifying considerably) is having and being unaware of male privilege, call it M. A comment C on the internet is Bayesian evidence for M if and only if M-commenters are more likely to make comments like C than non-M commenters. My claim is that a comment like shminux’s is more likely to come from someone who has and is unaware of their male privilege than from someone who is at least aware of their male privilege.
In this particular case, you have inside view reasons to believe that shminux’s statement was generated via some other process than the one I’m suggesting that shminux’s statement provides weak evidence for (emphasis on the weak). There’s no contradiction here.
Forget what I said about the typical outside female observer. The point of that device was to 1) enforce outside view and 2) make gender issues more salient.
He did say it in public. It’s right here on LessWrong.
Okay. That makes it much clearer.
It’s now obvious to me that the outside viewer does see P(M|C) > P(M|¬C), and why.
However, what schminux gave is not C, but R(C) + Q, and to any attentive reader I assume that P(M|R(C)+Q+C) < P(M|¬C) < P(M) < P(M|C); where R(C) is a meta-observation about one’s own (past?) observations/comments C¹ , and Q is a question about the evidence carried by C and R(C).
Any external reader who misses this is, as far as I can tell, simply wrong. They were most likely themselves taken by confirmation bias or some other undesirable effect.
Confirmation bias among self-proclaimed feminists is actually something I have a rather high prior for, since they train themselves to see gender bias everywhere in many cases.
(and R(C) does include C as a component once unwrapped, but for clarity I added C in the chain above; assume appropriate anti-double-counting measures are taken)