http://blogs.discovermagazine.com/80beats/2009/06/02/more-evidence-that-girls-kick-ass-at-math-just-like-boys/
To copy the following older post off the lab blog (from 10/4/2007):
I vented about this at lab meeting the other day. I now think it's time to actually organize some information on it because the idea seems more pernicious than I initially realized.
I think of this as the "Larry Summers" hypothesis, although this is actually a bit inaccurate (although people would probably recognize it by that name). The core idea is that a reason women are underrepresented at the highest levels of success in math and science (e.g., faculty positions at top universities) is that the distributions of inherent ability betwen men and women are different. The mean inherent ability may be identical, but greater variance in the male distribution puts more men in the extremes of high and low ability. Thus at the highest levels of success, you would expect to find less women because there are fewer women at the very upper end of the distribution.
This seems like a question of science and statistics, but there's a significant danger here. If the core hypothesis is believed, it argues against gender-based affirmative action at top universities. If the existing difference in representation of men and women in top universities is based on a genetic difference, increasing representation women will actually make those departments stupider on average.
The alternate hypothesis is that the existing differential representation is due to cultural (social, environmental) factors that can be ameliorated by affirmitive action policies aimed at overcoming a historical cultural bias against women in these fields.
So we have a particularly difficult situation: a scientific question that is very hard to assess that has a direct and immediate policy impact. My personal opinion is that in these cases, it is important to weigh the costs of error. If we have to make a binary policy decision (affirmative action, yes or no) that will be based on an evaluation of the balance of unclear evidence, which error is more costly? Is more damage done by in appropriately implementing or eliminating affirmative action? I will note that this kind of consideration is broadly unpopular with scientists who see themselves as pure seekers of truth. But I'm not going to argue that philosophical point here since this particular theory can just be evaluated on a balance of evidence basis and there doesn't seem to be much to it.
Some background
The original Summers situation is described evenhandedly on Wikipedia.
An excerpt:
Controversy
Another study performed by the American Psychological Association in response to the book The Bell Curve, which investigated the difference in intelligence between different social classes (strongly correlated with race in the U.S.), determined (as did the authors of the book) that the studies available in 1995 showed no major difference between males and females in regard to IQ scores.[24]
In January 2005, Lawrence Summers, president of Harvard University, unintentionally provoked a public controversy when MIT biologist Nancy Hopkins leaked comments he made at a closed economics conference at the National Bureau of Economic Research.[25] [26] [27] In analyzing the disproportionate numbers of men over women in high-end science and engineering jobs, he suggested that, after the conflict between employers' demands for high time commitments and women's disproportionate role in the raising of children, the next most important factor might be the above-mentioned greater variance in intelligence among men than women, and that this difference in variance might be intrinsic,[28], adding that he "would like nothing better than to be proved wrong". The controversy generated a great deal of media attention, forced Summers to make a number of apologies, and led Harvard to commit $50 million to the recruitment and hiring of women faculty.[29]
In May 2005, Harvard University psychology professors Steven Pinker and Elizabeth Spelke debated "The Science of Gender and Science".[30]
In July 2006, Stanford University neurobiologist Ben Barres, a transsexual man, wrote a provocative piece in Nature on his own experiences as both a male and female scientist.[31] Barres argued that prior to transition, he had succeeded as a female despite pervasive sexism. Barres wrote that numerous studies show female scientists are consistently rated lower than their male counterparts with the same levels of productivity and credentials.
In 2006, Danish psychologist Helmuth Nyborg was asked to vacate his position at Aarhus University after publishing a paper in Personality and Individual Differences that showed an 8 point IQ difference in favour of men.Nyborg, Helmuth (2005). "Sex-related differences in general intelligence g, brain size, and social status". Personality and Individual Differences 39: 497-509.
Summers in an economist. Where did he get this idea about variability? I'm not sure, but apparently this idea has some "mainstream" support in Psychology.
Baumeister's 2007 APA address
Stephen Pinker taking this position in a debate with Liz Spelke
There's a lot of interesting stuff here. But I was really captured by the similarities of the Pinker/Baumeister argument because they have the same flaws that seemed pretty obvious to me. I actually do hesitate when it seems like a smart person is arguing something transparently stupid. A plausible alternate hypothesis is that I'm wrong or misunderstanding something important. So pointing out the core problems here may help us evaluate which hypothesis is more plausible (they're being stupid or I am).
The variance hypothesis
I included the links so that you could check my account of the hypothesis, but it's not really complicated. There are plenty of data that say men are overrpresentated at both tails of the distribution. You could probably argue about the data, or tackle the question of what is being distributed, IQ? intelligence? success? ability? Success is probably the best description (although all those constructs co-load) and I'm happy to stipulate the data are what they are for the purpose of evaluating the rest of the argument.
The problem is the inference that the differential variability is inherent, i.e., it's based on a significant genetic contribution. The obvious alternate hypothesis is that the differential representation in the tails is predominantly cultural or societal based on differential treatment of men/women (boys/girls).
In case you're wondering, a simple way you get differential representation in the tails is via feedback loops. Let's say individuals vary in ability and some small percentage have the ability to become geniuses if provided with effective teaching/instruction/training (note this is a fairly nativist argument itself and not the only one). E.g., exhibiting ability -> attention & more teaching -> greater ability -> more specialized teaching... etc. until you push somebody out into the very upper tail of ability/success. If this trait was equally distributed across genders, but there was even a slightly lower chance that the feedback loop gets started for women then men, you'll end up with women underrepresented in the tail due to cultural differences.
Is that plausible? Maybe. There is an abundant evidence that men and woman (boys and girls) are treated differently at least. What's the evidence that the overrepresentation in the tails is genetically based (inherent)?
There isn't any actually. Baumeister references greater variability in height among men, but not only can I not find any source for that (I looked up average height charts and the distributions look roughly identical) but you might even expect SD to go up with mean (men are taller).
Very weirdly, both Baumeister and Pinker argue that men have been selected for greater risk taking historically as if that was related. Baumeister also argues that men are under greater selection pressure to pass their genes along (fewer men have passed their genes through time than women apparently). But neither argument is remotely relevant to producing greater variability in success. In fact, greater selection should produce less variability (as any remotely clueful evolutionary psychology should know) not more. You could use either of these to argue why you think men should be smarter on average (they aren't) but neither is related to increased variability.
Pinker gives us one pseudo-shred of data: "And biologists since Darwin have noted that for many traits and many species, males are the more variable gender."
So there's your alternative hypotheses to consider: men are the more variable gender or social/cultural affects influence the degree to which women acheive upper-tail success.
Pinker's talk is the far less egregious of the two, but the absence of consideration of social or cultural effects is still staggering. He spends a considerable amount of time documenting existing differences between genders to prove that there are some. I don't believe there are reasonable people who disagree with that, but it still doesn't mean success/ability is genetically determined. It doesn't even mean all those differences are genetic/inherent.
There are a whole lot of examples of this in his talk, but here's my current favorite as a parent of a 13yo girl who is in an accelerated math track in high school:
Fifth, mathematical reasoning. Girls and women get better school grades in mathematics and pretty much everything else these days. And women are better at mathematical calculation. But consistently, men score better on mathematical word problems and on tests of mathematical reasoning, at least statistically. Again, here is a meta analysis, with 254 data sets and 3 million subjects. It shows no significant difference in childhood; this is a difference that emerges around puberty, like many secondary sexual characteristics. But there are sizable differences in adolescence and adulthood, especially in high-end samples. Here is an example of the average SAT mathematical scores, showing a 40-point difference in favor of men that's pretty much consistent from 1972 to 1997. In the Study of Mathematically Precocious Youth (in which 7th graders were given the SAT, which of course ordinarily is administered only to older, college-bound kids), the ratio of those scoring over 700 is 2.8 to 1 male to female. (Admittedly, and interestingly, that's down from 25 years ago, when the ratio was 13-to1, and perhaps we can discuss some of the reasons.) At the 760 cutoff, the ratio nowadays is 7 males to 1 female.
"...this is a difference that emerges around puberty, like many secondary sexual characteristics." Can Dr. Pinker seriously not consider the possibility that the development of secondary sexual characteristics would increase the differential social/cultural treatment of young girls? Does he think the only thing that changes for girls at puberty is the concentration of hormones in the body? wtf?
And in his own data here, he admits the number of girls in the upper tail has changed dramatically over the recent period of greater concern about gender issues and more gender-based affirmative action.
So re-consider the binary policy question of "gender-based affirmitive action, yes or no?" as it is impacted by this very difficult scientific question, "is the observed increased variability in success/ability of men relative to women based on inherent genetic factors or environmental/cultural factors?" And you'll see why I'm so thorougly annoyed about this.
Larry Summers was President of Harvard at the time he suggested this was a major cause of differential representation of men and women on the faculty. Is it still safe to ignore the policy implications of weak science?
P.S. There is one claim in the midst of Pinker's laundry list aimed at proving men and women are inherently different that I should probably go find better data on.
Seventh, a lack of differential treatment by parents and teachers. These conclusions come as a shock to many people. One comes from Lytton and Romney's meta-analysis of sex-specific socialization involving 172 studies and 28,000 children, in which they looked both at parents' reports and at direct observations of how parents treat their sons and daughters — and found few or no differences among contemporary Americans. In particular, there was no difference in the categories "Encouraging Achievement" and "Encouraging Achievement in Mathematics."
There is a widespread myth that teachers (who of course are disproportionately female) are dupes who perpetuate gender inequities by failing to call on girls in class, and who otherwise having low expectations of girls' performance. In fact Jussim and Eccles, in a study of 100 teachers and 1,800 students, concluded that teachers seemed to be basing their perceptions of students on those students' actual performances and motivation.
A few basic points of reality: First, math and science teachers at the middle and high school level aren't predominantly female (elementary school teachers are and they aren't dupes, but they are influenced by cultural expectations). Second, he's apparently never heard of the Implicit Attitudes Test which shows marked discrepancies between intentions and actual bias. People regularly report doing their best to be non-biased, but are not able to consistently hold to it. The fact that parents report they mean to be as encouragng to girls in school does not guarantee that they are. And the fact that teachers believe they are responding to their perceptions of student interest does not mean their perceptions aren't influenced by a bias to see women/girls as less interested in math/science material (which, incidentally, is a feedback loop that would push women away from the tails and back towards the mean of the distribution).
BTW, the best way to see the patently obvious difference in treatment of girls/boys in school is to look at how they treat each other. You get a better picture of the cultural bias bleeding through because they don't have the frontal lobes to effectively inhibit the implicit attitudes they pick up from the world. But I should get some actual data on this.
No comments:
Post a Comment