I don't believe that IQ tests measure "latent general intelligence". However, your perception of the differences between "actual academic work" and IQ tests notwithstanding, the fact remains that IQ scores fairly accurately predict academic outcomes.
[WS:] In the same way, my age 'predicts' the prices of gasoline in the US. You know as well as I do that correlation does not equal causation, and in order to predict something you need a causal model, not just statistical correlations. The IQ and 'aptitude' testing lacks such causal model in a scientific sense of the word, save for a bunch of racist beliefs about 'general intelligence' or g. Since typical academic tasks differ rather substantially from those included in 'aptitude' tests (which you seem to acknowledged in your reply), the existence of g is a necessary assumption to claim that the latter is a good predictor (in a causal sense) - without such an assumption the predictive claim is a nonsequitur. I would be interested to hear how you causally link test scores to academic performance without assuming some kind of "latent intelligence" they are supposed to measure
Miles: Well, important for whom? If you want to predict academic performance or performance in jobs that require extended formal education, IQ tests do a good job. From the perspective of the capitalist class, that's pretty important (personnel selection). I don't follow the last sentence: what do they obscure? Certainly not people's likelihood of succeeding in various social settings that require formal education!
[WS:] Good question, indeed. There are various statistical techniques that 'predict' business risk or success factors, and they are often 'validated' in the sense that they make money for people who use them. But then businessmen and swindlers or various stripes are just into money grubbing, they do not care whether their risk and profit calculation schemes are built on accurate causal models or spurious statistical correlations.
I thought that science is held to somewhat higher standards of causal modeling. Viewing 'aptitude' testing as science, which implies the existence of a causal model, and treating the market success of that testing as the validation of its predictive power is a swindle in my book - it creates a false impression that the charlatans who use such tests are doctors rather than quacks i.e. they can capture the nature of human cognition instead of peddling snake oil of spurious correlations.
The difference between a spurious correlational swindle and a bone fide causal model is that the former gives you a lot of false negatives, whereas the latter does not. A false negative may not matter that much in money grubbing schemes - as overidentification of risk is good for the bottom line - but it matters quite a bit when it wastes actual human lives by baring them access to societal resources. In that sense, 'aptitude' testing not only obscures the true nature of human cognition, but it is akin to fascism that gives access to a good life to a select few and bars it for the "untermenschen" based on pseudo-scientific beliefs of the 'worthiness' of human life.
Miles: Read more psychometrics; it's true that there are hacks like Murray and Herrnstein associated with psychometrics, but there is a lot of important scientific work being done in the area (see Max's recent web links for some interesting, rigorous psychometric stuff). Keep in mind that IQ testing is not equivalent to psychometrics, and many people with psychometric training reject the claim (as I do) that IQ measures general intelligence.
[WS:] Do I also need to read the bible more to know that organized religion is a bunch of crock? :) But more seriously, while you are technically correct that not all psychometrics is about the measurement of 'human worthiness' - its results are often used to the latter end. Likewise, not all economic theory is the exoneration of the US-style capitalism - but the discipline is often used to that end.
It is very difficult to separate "pure" science from its social application, especially when those applications are highly questionable from an ethical or a political point of view. The Frankfurt School, especially Max Horkheimer, made that point by calling such sciences "herrenwissenschaften" or "master race sciences," and Foucault made a similar point by arguing that knowledge is almost always inseparable from power relations ("knowledge-power"). Pierre Bourdieu made similar observations in his analysis of the French academia (_Homo Academicus_). In such cases, the critique of the applications is a valid critique of the science in question itself, at least in my book.
Wojtek