= 27d567bc2d48d9f40e9ccc20ea370b15)
gjm comments on Link: quotas-microaggression-and-meritocracy - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (163)
They took the exact same application, sometimes with male-looking names and sometimes with female-looking names, and asked faculty for their opinions about them. The female versions were rated substantially (and significantly at the 0.001 level) worse for "competence", "hireability" and "willingness to mentor this student". The gap in estimated competence was about the same in size and significance as the gaps in the other metrics, which to me seems to indicate that differences in fear of a wrongful termination suit didn't contribute much if at all. (On looking at the relevant bit of the paper, the authors agree and have some statistical analysis that allegedly supports this view.)
When asked roughly what starting salary they'd offer the applicants, the "female" applications attracted ~12% lower figures.
(The details are all there at the other end of the link I gave.)
I'm not completely sure of anything, ever. But: The information included: age, degree granted and university that granted it, GPA, GRE scores, extracts from application letter and faculty letters of recommendation, etc. If there's any residual information to speak of in knowing whether the applicant was male or female, I'd be rather surprised; if there's enough to justify the differences found in the study, I'd be flabbergasted.
[EDITED to add: While affirmative action may be "a thing in college admissions", to the best of my knowledge it is not "a thing" in the awarding of college degrees, the calculation of GPAs, etc.]
If women are more likely to use maternity leave or otherwise devote more resources to family and less to the job than men are, and if they are more likely to sue for sexual harassment than men, then most of these assessments could be correct; seeing a female name actually does give information.
As I have said a few times already in this thread, the numbers make it look very much as if the dominating factor was an assessment that the "female" candidates were less competent than the "male" ones. Lack of commitment and increased lawsuit risk don't seem to me like matters of competence and I would expect the faculty surveyed to share that opinion.
Do you have a rough estimate of (1) how much more likely women would have to be than men to do those things, in order to justify a difference in evaluation of the magnitude found by this study, and (2) how much more likely women actually are to do those things?
(Two remarks in regard to sexual harassment lawsuits. 1: I think the relevant figure isn't how much more likely women are to file such suits but how much more likely they are to file them when no harassment has really occurred. But perhaps not: suppose women are more likely to be victims of sexual harassment sufficient to justify a lawsuit, and therefore more likely to file such lawsuits; then one possible position would be to consider women less desirable employees on those grounds and rate them as less competent. Personally, I think that would be odious, but I can imagine that some people might disagree. 2: My understanding is that actually such lawsuits are really rather rare, much too rare for rational consideration of their risk to yield the reported difference in evaluation even if (a) all such lawsuits are assumed groundless but successful and (b) the resulting losses in productivity and collegiality are assigned to lack of "competence" by the person filing the lawsuit. However, I don't have extensive statistics on this and will be happy to be corrected if wrong.)
The problem is that "whether sexual harassment has occurred" isn't all that well-defined. You can of course define "sexual harassment" however you want but then you have to establish you it's a bad thing. For example, from a briefing at the company I work at the examples of "sexual harassment" was:
1) a woman goes to work in somewhat provocative/revealing clothing and a male coworker complements her on her appearance.
2) a manager used the phrase "guys and gals".
Frankly if these examples are typical of "sexual harassment", I'd say sexual harassment isn't a problem.
Did either of these examples result in lawsuits?
I don't know, the presenter didn't say. Although the fact that these were presented as examples of behaviors not to engage in, is telling. Also even if they don't bring a lawsuit, the fact that they make an issue out of these kinds of things is not conducive to a good work environment.
Of course I wasn't there. But it occurs to me that there are several reasons why "marginal" examples might actually be the most useful:
I'd put the examples you give in the category of (not typical examples of sexual harassment, but) things that are frequently harmless but (1) might cause easily-avoided annoyance or upset in some cases and so should maybe be avoided and (2) in some cases might indicate, or be thought to indicate, an underlying bad attitude (women in the workplace being seen primarily as eye candy; women being seen as lower-status and akin to children).
I repeat: of course I wasn't there and don't know exactly what your presenter said about these examples. If s/he said "these things are definitely harassment and you could get in serious trouble for doing them" then I'd regard that as unreasonable; if s/he said "these things may seem harmless, and often they are, but you should still avoid them", I'd agree.
Anyway, I mention all this just in the interests of mutual understanding; it's all kinda irrelevant to the question of whether "greater risk of sexual harassment lawsuits" is a good justification for rating an identically-described person as substantially more "competent" if they have a male name than a female name. Do you really think it is?
The problem is that it causes people to treat it as an archetypical example.
I fail to see why it should be policy to cater to people who are clearly being unreasonable.
For one thing, because being unreasonable is simply What People Do and it seems better to care about outcomes in the real world than outcomes in some imaginary world where everyone is always reasonable. So if doing something predictably results in a bunch of people being upset, then it might be better to avoid it even if it would be better for everyone if they weren't upset by it.
For another, because what's "clearly unreasonable" to one person may be "clearly reasonable" to another. It may seem "clearly unreasonable" for a woman to have a problem with having her appearance complimented by her male colleagues. But if what she's found is that over and over again her male colleagues comment on her (and other women's) appearance, and never on their ideas, while the reverse happens to the men around her ... why, then, I have some sympathy if she gets frustrated by yet another compliment on her appearance. (It might in some sense be better for her to focus not on the compliments on her appearance but on the absence of response to her work. But actual things that actually happen are easier to see and more psychologically salient than absences, even when the absence is the bigger underlying problem.)
Only people who are -- how shall I put it? -- clearly being unreasonable. One might prefer not to make policy on the basis of people who are clearly being unreasonable :-).
Seriously: yes, I agree that that's a potential problem. The obvious solution seems to me to be to make it as clear as you possibly can when you're talking about central examples and when you're sketching the boundaries. Unfortunately, I bet there will always be (clearly unreasonable) people who don't take any notice and either mix the two up or pretend to. I'm not sure much can be done about that.
I mostly agree (and upvoted), but...
Well, complimenting people wearing attractive clothes is is simply What People Do and it seems better to care about outcomes in the real world than outcomes in some imaginary world where no-one ever notices other people's clothes. So if wearing certain clothes predictably results in a bunch of people commenting on your appearance (and it annoys you), then it might be better to wear more modest clothes yadda yadda yadda.
;-)
While lawsuits may be rare, they are expensive, and people are risk-averse.
Also, the range of behavior that has to be avoided to avoid an unjustified lawsuit is much wider than the range of behavior that has to be avoided to avoid a justified lawsuit, and since even unjustified lawsuits are expensive, the former category is what really matters.
For a lot of colleges the hard part is getting in, and getting the degree isn't that hard conditional on getting in.
That would be why the application also included the applicant's GPA. And also both GRE scores. And a bunch of other things.
GPA is meaningless without knowing how hard the classes the applicant took were.
So is your claim that the scores on a single GRE test completely capture everything about an applicant relevant to job performance?
I find your question absolutely bewildering, given that the very sentence I wrote that mentioned GRE scores mentioned them only as part of a list of things.
Yes and when I pointed out the problems with all the other things in the list your reply basically amounted to "you haven't made any objections to GRE scores".
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.
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.)
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.