The Case Against Lawrence Summers (Aaron Swartz: The Weblog)

Updated with fascinating information from another study. Jump to new material.

Just between you and me, shouldn’t the World Bank be encouraging MORE migration of the dirty industries to the LDCs [Less Developed Countries]? … I think the economic logic behind dumping a load of toxic waste in the lowest wage country is impeccable and we should face up to that. … countries in Africa are vastly UNDER-polluted

— Lawrence H. Summers †

On January 14, 2005, Harvard President Lawrence H. Summers offered “some attempts at provocation” at a conference on “Diversifying the Science & Engineering Workforce”, specifically discussing “women’s representation in tenured positions in science and engineering at top universities and research institutions”.†

He begins by suggesting that underrepresentation isn’t always due to discrimination:

Catholics are substantially underrepresented in investment banking, which is an enormously high-paying profession in our society; that white men are very substantially underrepresented in the National Basketball Association; and that Jews are very substantially underrepresented in farming and in agriculture.

So, he says, we have to ask why women are underrepresented and he offers three possibilities. The first is what he calls “the high-powered job hypothesis”, namely that “young women in their mid-twenties make a decision that they don’t want to have a job that they think about eighty hours a week”. (“Is our society right [in these expectations and imbalances]?” He tables the question.) The second is “differential availability of aptitude at the high end” — that there is a difference in the variability of “mathematical ability, scientific ability” that is “not plausibly culturally determined” which, by his rough calculations, means there are five times as many male math/science geniuses as there are women math/science geniuses.

“I would far prefer to believe something else,” Summers says, but “the combination of the high-powered job hypothesis and the differing variances probably explains a fair amount of this problem.”

Could the differing variances be due to socialization? Summers doesn’t think so. He says that “a hundred different kibbutzes” each independently decided to reverse course from a sexual egalitarianism and let “the men … fix the tractors and the women … work in the nurseries”. And furthermore:

my experience with my two and a half year old twin daughters who were not given dolls and who were given trucks, and found themselves saying to each other, look, daddy truck is carrying the baby truck, tells me something.

(Summers does not say whether two-person sample was also raised without TV and books and all the other images of socialization that say girls should play with baby dolls.)

Is it discrimination?

If it was really the case that everybody was discriminating, there would be very substantial opportunities for a limited number of people who were not prepared to discriminate to assemble remarkable departments of high quality people at relatively limited cost simply by the act of their not discriminating … I think one sees relatively little evidence of that.

So, he says, the general problems of universities are those of the “high-powered job”, the specific problems of the sciences are due to natural varying ability. “I would like nothing better than to be proved wrong,” but “empirical psychology” and “the data” say otherwise. And our personal prejudices have to bow before the objectivity of science.

This is a tune that is by no means new. As Stephen Jay Gould points out in his fine book, The Mismeasure of Man, throughout history those who have tried to justify existing inequalities by blaming biological determinism have said the same thing.

Paul Broca, for example, who carefully weighed numerous brains to see which groups were intelligent and which were not, was truly sad to discover that the brains of blacks were smaller than those of whites. But, he argued, there was nothing he could do: “There is no faith, however respectable, no interest, however legitimate, which must not accommodate itself to the progress of human knowledge and bend before truth.”

Despite such lofty principles, Gould shows that, quite aside from the false assumption that brain size is related to intelligence, Broca repeatedly and consistently manipulated his data to reach these conclusions. Gould believes such manipulation was unconscious, even though at times it was quite extreme. (As one example, Broca threw out entire systems of measurement when the inferior races scored too well on them.)

The tone is a theme through Gould’s book, so it is no surprise to see it reappear today. But is it any more true?

Broca’s major error was assuming that the size of someone’s brain could tell you how intelligent they are. This is of course incorrect — people’s brain size is mostly determined by the size and build of the rest of their body — and trouble the assumption seems absurd. Yet we believe in a notion that is just as silly — that IQ tests and math exams measure some sort of innate intelligence.

In the present context, a study by Claude Steele brings some of the problems into sharp relief. (I am working here from Steele’s chapter in Young, Gifted, and Black.) Steele, with Steven Spencer and Diane Quinn, took some of the best and most dedicated math students they could find and gave them an extremely difficult math test. The men performed more than three and a half times as well as the women — an enormous gap. Then they gave students the same test, but told them this was a special test in which women always did as well as men. The gap closed almost entirely, with women’s scores increasing dramatically. (Steele’s research shows similar effects with other victims of stereotypes, like blacks.)

Steele suggests that women’s scores are depressed by “stereotype threat” — a woman comes across a hard problem that they have trouble solving, and they begin to worry that people might think they’re having trouble at math because they’re female, and they begin to worry that this might be true. (Needless to say, comments like Summers’s can’t do much to alleviate these fears.) When they’re told the stereotypes can’t apply, the fears go away and they perform fine.

But the mechanisms involved are unimportant for our purposes. The key point is that the supposedly objective examination measure of intelligence is seriously flawed, even on a subject as supposedly objective as a math test. These tests are not just measuring intelligence; at the very least they’re also measuring something like self-confidence.

As Gould argues, we are tempted to measure things and then we are tempted to assume the numbers that result refer to something real — that tests in math measure something called “mathematical ability”. But this is a logical leap — the case must be carefully proven. There’s no evidence that such a thing as “math ability” even exists, let alone that it can be measured.

Biological determinists like to respond to such arguments by saying that the speaker is denying the influence of biology, when all reasonable people know that both biology and environment have an impact — say 40% biology, 60% environment. But it is the determinists who are missing the point. Skills cannot be divided up in so absurd a manner.

Let us put aside brains for a second and imagine the arm muscle. Some people are born with a naturally skinny body type that doesn’t build much arm muscle. Others naturally build muscle like crazy. Clearly biology plays a role. But it’s absurd to say that it’s 40% biology, or any other number — a muscular person whose arm is paralyzed will not be very muscular at all, while a weak person who works out incessantly will have huge biceps.

It’s not hard to see how the brain could work the same way: people are born with natural tendencies, but work or environment can quickly change this “default” destiny.

In a real twist of irony, it turns out that it is exactly this confusion that causes the gender gap. Further research by Carol Dweck has investigated whether students believe that “mathematical ability” is a learned skill or an innate gift. A simple study shows the shocking effects of this belief. Students were given an obscure non-verbal IQ-style test that was designed to be easy for their age group. Afterwards, half were told “You got a great score. You must have worked really hard.” and half were told “You got a great score. You must be really good at this.”. Then they asked kids if they wanted to try harder questions that might help them learn more. The ones who were praised for effort were happy to — one effort-based kid (in another study) rubbed his hands together, licked his lips, and exclaimed “I always love a challenge!” — but intelligence-based kids tried to avoid it, perhaps fearing they’d look stupid.

They were then given the harder problems, much too hard for them to solve. Then they were given more easy problems again. The gift kids did much worse on the third set of problems. When asked if they wanted to take more problems home, they said they already had them at home (an absurd lie). By contrast, the skill kids not only asked for some to take home, one even asked for the name of the tests “so my mom can buy more when I run out”. The kids were also asked to write a note about the tests to other kids who might take them. The notes were anonymous, but there was a little place to put your score. Nearly 40% of the gift kids lied and exaggerated their scores. All this from just one little sentence — the kids were otherwise identical.

In other words, Dweck says, telling kids they’re smart makes the stupid and liars.

Dweck’s observations of classrooms find that boys are more often chided on the basis of effort (“Johnny, I know you’d do better if you just spent more time on this”), perhaps leading girls to infer that their ability is innate. Her studies find that girls are more likely to believe their ability is innate than boys and that it is these girls who are the cause of the gender gap in ability. Teaching these girls that mathematical ability comes from hard work can eliminate the gap.

[Personal note: Both Dweck and Steele have been recently hired away from Stanford and presented their results to my class.]

Looking at the long history of how even supposedly scientific evidence of the differences intelligence between groups has been false and distorted, one ought to be very careful before reviving such claims. Summers was not only not careful in his evidence, he didn’t even bother to present evidence.

There are few worse things an intellectual can do than present false claims without evidence. If you present true claims, of course, there is no problem. And if you present false claims with evidence, one can evaluate the quality of the evidence. But if you simply state something as true, it has a way of seeping unquestioned into people’s heads. And how much worse, then, to spread these falsehoods on such a subject, where they can do great harm.

I’ll close with a bit from the question period after Summers’s talk:

Q: I noticed [this is] being recorded so I hope that we’ll be able to have a copy of it. That would be nice.

LHS: We’ll see. (LAUGHTER)

Further reading: Sexist Calvinism

posted March 09, 2005 03:40 AM (Politics) (24 comments) #

The Case Against Lawrence Summers

Nearby

Comments

Add Your Comment