Thursday, January 15, 2009

Bad Science in Education

Deborah Meier from Bridging Differences has a good post about painting a picture of what education could be, as a means of motivating society to put forth the necessary resources to attain it.  She references Charles Murray's The Bell Curve, a book claiming that differences in education achievement are function of innate intelligence, or IQ, and that white kids achieve more than black kids because they are innately smarter.
We both know that on the biggest question—of human potential—Murray is dead wrong. It takes only one example to prove that point. It is no longer a matter of hope or faith for me, but experience. Although one example doesn’t demonstrate how it can be done on a larger scale.
But this post - by a pretty big figure in the education community - makes an elementary statistical error.  Murray claims that the white kid bell curve is shifted several points to the right of the black kid bell curve, so that the typical white kid is smarter than the typical black kid; he does not say all white kids are smarter than all black kids.  In fact, the graph below demonstrates that his own argument requires that almost 50% of black kids are smarter than almost 50% of white kids.

Now, I happen to think Murray is wrong too.  But when people make mistakes like Meier, it makes people like Murray appear more credible, and I think all of this underscores the lack of attention the education community places on mathematical or scientific literacy (perhaps because so few of us have math and science backgrounds).

The more compelling argument against Murray is one mentioned by cognitive scientist Stephen Pinker in The Blank Slate: which is that while population sub-group IQs can be different at a given point in time, they tend to converge in the long run.  And in fact, this is what has happened to most American immigrant groups, suggesting that differences in sub-group IQs are environmental rather than innate.


Anonymous said...

The more I think about it, the more I think Kevin Drum's suggestion (sorry, I can't find the original blog post) that basic statistics take the place of algebra makes sense. We encounter statistics almost every day, and most of those statistics are interpreted completely incorrectly. We'd all be better off if the general population actually understood the significance of distributions and means. - Dank

Jack said...

i don't know about that. to be able to comfortably apply basic statistical/logical intuitions to every day life you really have to understand them. you need basic algebra to perform a hypothesis test. non-normal distributions require manipulation of exponent variables. understanding the very idea of a distribution requires some understanding of functions. on the other hand: knowing how to use a p-value table or being able to define basic things like variance, std dev, etc. or struggle through an anova - that's not enough to develop functional literacy. ie its insufficiently understood to intuitively apply to basic tasks. like breaking up meier's bad falsification.