[nugget]

>Of course, if intelligence is only partially hereditary (as you suggested) then Harvard would do better to just apply some uniform standard of intelligence measurements to all applicants, assuming that the goal is to admit the most intellectually gifted students.

Isn’t this the role of the SAT?

I’d be interested to see the data for legacy versus non-legacy SAT scores for applicants, admits, and matriculants.

I’m not defending Harvard’s current admissions policies nor their historical admissions policies, which seem indefensible by modern standards. My interest stems more from the theoretical implications of assortive mating for traits that are both very heriditary and very valued by society (or very deterministic of “success”.)

[betterunix2]

The SAT is not an intelligence test; it is a test of a student's educational background and the degree to which they engaged in college prep in the years leading up to the test. The fact that the 2/3 of the test is devoted to some measure of literacy should tell you that (literacy is an absurdly poor proxy for intelligence -- especially when it is confined to a specific language). Even the math section is in part a test of a student's educational background, as you must at least be familiar with the specific field of math and particular notation that the test uses.

There is a reason parents enroll their children in SAT prep as early as middle school and even elementary school. Again, let's assume intelligence is hereditary; then an ideal intelligence measurement would be impossible to prepare for, because it should measure something that a person cannot change about themselves (their genes). The fact that SAT prep measurably improves SAT scores says at least one of two things must be true: the SAT is not measuring an innate property, or that intelligence is not simply inherited.

(Spoiler alert: both of those are true.)

[nugget]

Of course you’re correct that test prep can skew the results, but SAT scores are still one of the best measures available at scale, hence why they are so widely used despite the obvious flaws. Other than a DNA test for IQ, you can prep for any known test, whether it’s the SAT, the WPPSI, the WISC-IV, a Rorschach, or the New York Times crossword. As user wycs hinted at elsewhere, a polygenetic spit test for raw IQ potential - the Gattica scenario - is likely not as far off as we believe; when that hits, it’s a whole new world (from pre-conception.)

[betterunix2]

Again, SAT scores are an extremely poor measurement of intelligence, because at best they only apply to people who (1) speak American English with complete fluency, (2) have literacy in English, and (3) have been taught the syntax, symbols, and particular fields of math tested in math section. Beyond that, since the modern SAT includes a writing section, there is a large subjective component of the score that further reduces its usefulness as a measure of intelligence. The SAT I took had no essay, but even the old test would have done a poor job of measuring the intelligence of someone who did not receive a high school education.

The actual purpose of the SAT is to screen students for a minimum educational background needed to complete a four-year degree at an American college. It only applies to a typical American education, and only to certain specific aspects of that education (there is no section on music, art, history, etc.).

[SamReidHughes]

They removed the writing section from the modern SAT, it's back to 1600 points.

I take it you're saying the SAT math section is a good measure of intelligence for all the kids that went to high school.

[honkycat]

SAT is great at measuring who spent the most money / who goes to the best grade/high school but not a whole lot more than that.