I don't follow your argument here. The predictive validity of IQ tests is higher than the predictive validity of SSRIs for the treatment of depression. (Proportion of variance accounted for is higher in the former case than the latter.) According to your criterion of science, most drug research is not scientific! There is lots of error variance floating around in studies with human beings, so we'll have a lot of false positives and false negatives in "real" scientific research.
[WS:] This only suggests that a much of psychology is high brow sorcery - based on circumstantial evidence at best :). We indeed do not know how SSRIs exactly affect depression, because we do not know what causes depression (unlike, say, etiology of viral infections.) All we know is statistical correlations.
There is, however, a crucial difference between 'aptitude' testing and SSRI testing. "Aptitude' testing is and historically has been used mainly to justify the denial of access to societal resources (mainly education.) SSRI testing, by contrast, is to make access to a resource available.
That is to say, the main purpose of SSRI testing is to provide medicine to people who may (or may not) need it. The use 'aptitude' testing, by contrast, is to deny access to education to those who supposedly do not qualify. Now if, as you say, 'aptitude' testing does not measure any fixed innate quality of human cognition (g), then such use of the test strikes me as sheer idiocy if not outright malice. It basically consists of the practice of finding out who has not learned his/her lesson to bar them from learning in the future. To use your own analogy, this would be like using SSRI testing to deny people further treatment for depression.
This is where the analogy to fascism comes. Fascism uses spurious correlations between human performance on certain tasks deemed as "civilizational" to identify "life unfit for life" and bar those "unfit" life forms from the access to the "lebensraum" available to those who "qualify" by such criteria, or perhaps to exterminate them altogether. In fact, fascists did exterminate the "feeble minded" i.e. those who would score poorly on your cherished IQ tests. They did forced sterilization based on IQ results. They did deny right political asylum based on IQ results. And they keep denying entry to educational institutions based on the same criteria.
In that context, your claim that 'aptitude' testing may be used to a beneficial purpose sound quite hollow. I can certainly envision the possibility of using such tests to such purpose, e.g. to identify people with special needs for the purpose of providing them with additional resources they may need (e.g. to admit them to the institutions of learning in the first place,) and that is why these tests were initially invented by Binet in the first place. But this is all wishful thinking - the reality is that these tests HAVE BEEN USED mainly to deny access and in the process make a boatload of money for those who make those tests.
Going back to your methodological point about predictive validity - what you are saying is technically correct when you remain in the sphere of correlations and explained variances. But as I said before, my vision of science is different from mere maximizing the odds of getting lucky (no pun intended:).) From a marketing or actuary point of view it does not really matter what cause people to buy cruise vacations or contract terminal illness as long as the marketer can make a successful sales pitch and set the price or premium right. But it takes more in my book to practice science, and that "more" is precisely causal models.
Another point, if we assume that every human being (save those with developmental disorders) is capable of learning - as evidenced by the fact that all human are capable of learning a language, by far the most complex system of knowledge (which btw is Chomsky's point) - then any task that involves learning will have a positive correlation with academic performance by definition. For example, if I test a 10 month old children on their ability to walk, some will "pass" other will "fail" - and yet I will find a positive correlation between these results and the ability to walk at the age of 5 years. Where my "walking test" fails is the false negatives, i.e. prediction that those who fail the test will not be able to walk. This is where knowing the causal model comes really handy - a real understanding of the causes behind the ability to walk would stop one from performing such an idiotic test or using its results to decide the child's future.
In sum, I am not disagreeing with your technical points about correlations between IQ/'aptitude' testing and academic performance. I am making two different arguments: (1) that these are statistical correlations only, which either do not explain or create a false impression how human cognition work, and (2) that the actual use of these tests in social practice cannot be separated from the evaluation of their merits, and that actual use involved mainly elitist, if not fascist, goals of denying access to social resources for certain groups of people.
Wojtek