Dada is the sun, Dada is the egg. Dada is the Police of the Police.

1/10/2006

IQ

Once again, there's a discussion going on about IQ tests that is drenched in misunderstanding.

Usually, this debate centers on whether or not IQ tests really measure intelligence or not. The answer is no. IQ tests measure one thing, and one thing only: how well one performs on IQ tests. That's the only thing it possibly could measure.

The question, then, is whether or not the ability to do well on IQ tests correlates with the possession of 'intelligence'. Drum says:
Like it or not, all the PC handwaving in the world won't change the fact that (a) IQ tests are a pretty good measure of the cognitive ability normally referred to as "intelligence" and (b) intelligence is an important trait for a wide variety of modern day tasks.
Now, (b) is almost tautological, or at least true in a completely uninteresting and uncontroversial way. (a) rests on a couple of assumptions:

1. There is an identifiable cognitive ability that 'intelligence' designates
2. It has been established that this ability correlates with IQ scores

No support is offered for either of these assumptions, and they both seem shaky at best. At the very least, before a claim like Drum's can be made, we would need some kind of definition of 'intelligence' and some evidential support for the alleged correlation with IQ. But a definition of intelligence is likely to be highly arbitrary, and it's difficult to see what kind of evidence could possibly be marshaled to back up the claim of a correlation. You would basically need a study that kept track of how intelligent subjects were, and then mapped the results on to their IQ scores. But what would the measure of intelligence be? It's starting to look like 'intelligence' is just not something amenable to scientific inquiry at all, since it seems to be based at least in part on the judgment of human agents ('intelligent' being to some extent a term of approbation). You might be able to define it as something like 'success' relative to various projects, but even that is run through with evaluative notions, and is certainly not easily quantifiable.

This makes it awfully tempting to dismiss the whole IQ business as pseudo-science. At the very least, we have to acknowledge that until these tasks have been completed, there's not much use in invoking IQ.

Blogarama - The Blog Directory Sanity is not statistical.