You didn’t point out the problems with all the other things in the list, you made a claim about one thing in the list. My reply did not (as I have already pointed out) amount to “you haven’t made any objections to GRE scores”.
Regardless, no metric is perfect, and no one has been claiming otherwise. Accordingly, it is possible in principle (as I have already said more than once in this discussion) that there might be male/female differences that are either really huge, or highly relevant to scientific competence but startlingly uncorrelated with all the information provided to the faculty in this study, and that render the assessments they made rational given the information they had.
However, it seems pretty unlikely to me.
What do you think? Is the best explanation for these very different assessment of identical applications from differently-named candidates, in your opinion, that the faculty making the assessment are aware of such huge differences between men and women and have weighed them roughly correctly (not necessarily consciously and explicitly) to arrive at the results that have? If so, could you sketch for me roughly what these differences are and how they lead to that result? Because I’m having real trouble thinking of any hypothesis of this sort that’s consistent with what I think I’ve observed of the relative abilities of men and women.
Accordingly, it is possible in principle (as I have already said more than once in this discussion) that there might be male/female differences that are either really huge, or highly relevant to scientific competence but startlingly uncorrelated with all the information provided to the faculty in this study
Do I believe that Americans are generally more intelligent than Europeans? No, I don’t. At the same time in the LW census the average American has somthing like a 10 point higher IQ. In the data set there a strong correlation.
I think the intuition that the factor of the name should be zero is wrong even if there no causal effect because gender simply interacts in complex ways with many other things. I’m not sure in what direction the factor is going to correct, which might also be different in different situations but assuming that it contains no information at all doesn’t seem to be right.
I just grabbed the latest LW survey data I could find, selected (1) rows with “United States” as country and something other than a null for IQ and (2) all rows with something other than a null for IQ. (Note that this doesn’t include any sort of selection on the basis of reliability of IQ score.) I got means of 138.3 for the larger dataset (472 numbers, stddev=13.6) and 140.7 for the smaller (249 numbers, stddev=13.5). I wouldn’t call that “something like a 10 point higher IQ”.
the intuition that the factor of the name should be zero
What intuition that it should be zero? The question is whether it should be very large, not whether it should be exactly zero.
I’ve already explained why the difference would need to be very large for these results to be correctly explained by saying that the rating faculty made accurate allowance for real male/female competence differences. If you missed that, or you think I got it wrong, or it didn’t make sense, do let me know.
If so, could you sketch for me roughly what these differences are and how they lead to that result?
Let’s see: there are numerous ones the most relevant are: women have less variation in intelligence then men and so there fewer unusually smart women. Women are worse at taking criticism. There is also a lot of stuff about the kind of hierarchies men and women tend to form.
Because I’m having real trouble thinking of any hypothesis of this sort that’s consistent with what I think I’ve observed of the relative abilities of men and women.
Have you actually been observing the relative abilities between men and women, or is your reaction whenever you notice a woman doing something badly or acting emotionally to hit yourself for having a “sexist” thought?
women have less variation in intelligence than men
That could indeed (if the numbers work out) explain a difference in success at the very highest levels in the absence of prejudice. But this sort of effect is far weaker away from the very tails of the distribution, and the particular study we’re taking as an example in this discussion is not concerned with the very tails of the distribution. Further, my understanding is that GRE scores correlate somewhat better with intelligence than they do with job performance (see, e.g., this post which has a few references to the primary literature), and I would expect them to do a pretty good job of screening off differences in raw intelligence in this case.
Women are worse at taking criticism.
Evidence? (I have to say it looks to me as if people are bad at taking criticism, and I haven’t noticed a big difference between men and women; but I’ve not studied this and will be glad to learn.)
a lot of stuff about the kind of hierarchies men and women tend to form.
I’m afraid that’s not specific enough for me to form any idea of how it would justify a drastically lower assessment of the likely competence of a woman than an identically-credentialed man as a scientific lab manager.
Have you actually been observing the relative abilities
Relative abilities as such are pretty much unobservable. I’ve been observing the relative performance. But only casually and qualitatively; if you have a pile of useful data then by all means point me at it.
is your reaction [...] to hit yourself for having a “sexist” thought?
No, not at all. I notice both men and women doing things badly and acting emotionally all the time, and feel no particular impulse to self-punishment when I do so. -- Is it your usual practice to assume that people who disagree with you are off their heads, or have I said something to give you that impression particularly strongly in my case?
(Note for the avoidance of doubt: I am assuming that you didn’t mean “hit yourself” literally; of course if you did then it’s an even weirder thing to think I might do.)
You didn’t point out the problems with all the other things in the list, you made a claim about one thing in the list. My reply did not (as I have already pointed out) amount to “you haven’t made any objections to GRE scores”.
Regardless, no metric is perfect, and no one has been claiming otherwise. Accordingly, it is possible in principle (as I have already said more than once in this discussion) that there might be male/female differences that are either really huge, or highly relevant to scientific competence but startlingly uncorrelated with all the information provided to the faculty in this study, and that render the assessments they made rational given the information they had.
However, it seems pretty unlikely to me.
What do you think? Is the best explanation for these very different assessment of identical applications from differently-named candidates, in your opinion, that the faculty making the assessment are aware of such huge differences between men and women and have weighed them roughly correctly (not necessarily consciously and explicitly) to arrive at the results that have? If so, could you sketch for me roughly what these differences are and how they lead to that result? Because I’m having real trouble thinking of any hypothesis of this sort that’s consistent with what I think I’ve observed of the relative abilities of men and women.
Do I believe that Americans are generally more intelligent than Europeans? No, I don’t. At the same time in the LW census the average American has somthing like a 10 point higher IQ. In the data set there a strong correlation.
I think the intuition that the factor of the name should be zero is wrong even if there no causal effect because gender simply interacts in complex ways with many other things. I’m not sure in what direction the factor is going to correct, which might also be different in different situations but assuming that it contains no information at all doesn’t seem to be right.
I just grabbed the latest LW survey data I could find, selected (1) rows with “United States” as country and something other than a null for IQ and (2) all rows with something other than a null for IQ. (Note that this doesn’t include any sort of selection on the basis of reliability of IQ score.) I got means of 138.3 for the larger dataset (472 numbers, stddev=13.6) and 140.7 for the smaller (249 numbers, stddev=13.5). I wouldn’t call that “something like a 10 point higher IQ”.
What intuition that it should be zero? The question is whether it should be very large, not whether it should be exactly zero.
I’ve already explained why the difference would need to be very large for these results to be correctly explained by saying that the rating faculty made accurate allowance for real male/female competence differences. If you missed that, or you think I got it wrong, or it didn’t make sense, do let me know.
Let’s see: there are numerous ones the most relevant are: women have less variation in intelligence then men and so there fewer unusually smart women. Women are worse at taking criticism. There is also a lot of stuff about the kind of hierarchies men and women tend to form.
Have you actually been observing the relative abilities between men and women, or is your reaction whenever you notice a woman doing something badly or acting emotionally to hit yourself for having a “sexist” thought?
That could indeed (if the numbers work out) explain a difference in success at the very highest levels in the absence of prejudice. But this sort of effect is far weaker away from the very tails of the distribution, and the particular study we’re taking as an example in this discussion is not concerned with the very tails of the distribution. Further, my understanding is that GRE scores correlate somewhat better with intelligence than they do with job performance (see, e.g., this post which has a few references to the primary literature), and I would expect them to do a pretty good job of screening off differences in raw intelligence in this case.
Evidence? (I have to say it looks to me as if people are bad at taking criticism, and I haven’t noticed a big difference between men and women; but I’ve not studied this and will be glad to learn.)
I’m afraid that’s not specific enough for me to form any idea of how it would justify a drastically lower assessment of the likely competence of a woman than an identically-credentialed man as a scientific lab manager.
Relative abilities as such are pretty much unobservable. I’ve been observing the relative performance. But only casually and qualitatively; if you have a pile of useful data then by all means point me at it.
No, not at all. I notice both men and women doing things badly and acting emotionally all the time, and feel no particular impulse to self-punishment when I do so. -- Is it your usual practice to assume that people who disagree with you are off their heads, or have I said something to give you that impression particularly strongly in my case?
(Note for the avoidance of doubt: I am assuming that you didn’t mean “hit yourself” literally; of course if you did then it’s an even weirder thing to think I might do.)