Hacker News Reader

Comment by whatshisface | Hacker News Reader

That's not as true as one would hope. There's not a lot of "cultural fluency" involved in matching squares, unless you're from the jungle and have grown up never seeing any squares. (The theory that people who grew up not looking at corners would be less able to perceive square shapes has received some attention, but we have to admit that doesn't have much relevance to people living in modern countries like America or India .)

Now, cultural fluency questions have appeared on IQ tests, especially in the past. However that's not so much an attack on the idea of "g" as it is a problem with antiquated methods of measuring it.

That's not to say that IQ tests are that accurate. The biggest confounder I know of is ADHD, which causes the tests to come out five whole points lower on average. I'm sure there are others.

View on HN

> There's not a lot of "cultural fluency" involved in matching squares

There are a lot of cultural factors involved in believing a seemingly pointless exercise like matching squares is worth doing. You have to believe that something good can come out of it in general; and you have to believe that something good can come out of it specifically for yourself.

Accurately measuring the performance of someone who is not interested in demonstrating their performance seems like an impossible task for anyone but GLaDOS from the Portal series. ;)

More seriously, random pattern matching questions are the best you can do to control for the not-caring factor, because they eliminate the advantage of test prep. The SAT has a significant test prep factor, which I agree is highly confounded by "culture" (a byword for how much your parents make you study) and the circumstances and opportunities related to test prep, including the motivation of the student, and the availability of a place to study and study materials. IQ tests are significantly more meritocratic than algorithms interviews or the SAT.

I've never taken an IQ test and never plan to. I took the SAT, but it has significantly changed over the years; my score today would not be translatable to my score then.

OT, but I really love GlaDOS and hope there's another Portal soon.

If you haven't seen it yet, check out "Portal Reloaded" on Steam. It's a fan-made game like "Portal Stories Mel" from a few years ago, but Reloaded goes as far as adding entirely new mechanics. It's about as close as we'll ever get to Portal 3.

That could be true, but it’s still a good measure of IQ. This “cultural” angle is a narrative that started from criticisms of SAT vocab tests and critics of IQ tests latched onto it and hope they can convince everyone that IQ isn’t real and everyone is equally smart in their own way

Colleges don't filter their entrants by matching squares. Even the SAT test could be said to be culturally biased in a way that discounts rural ways of life and their rich legacy of traditional knowledge.

Ironically, college admissions would get a lot more egalitarian if they switched to IQ tests, because the long list of confounding factors (or "privileges" as some people call them) like the availability of free hours to study in, have much less impact on matching abstract symbols in rows than they have on the SAT or other partially knowledge-based tests.

I'm not sure if we want to do that because the capacity for working hard is often more valuable than "g". One's capacity and willingness to work hard for a long time is impossible to untangle from one's circumstances including one's upbringing. Fairly testing for that is much harder than testing for g but it may be too crucial to ignore.

It probably wouldn't make much difference. There's pretty good correlation between SAT and IQ. See e.g. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6963451/

That correlation is not invariant to cultural factors. Your own link mentions some of these - namely, "epistemic belief of learning, performance-avoidance goals, and parental education", for which we can easily measure variation across ethnicity (which of course is a lot more visible and more easily operationalized that a more nuanced understanding of culture, including urban vs. rural).

The paper you linked doesn't report the correlation as far as I could see, but it did cite a paper that said it was 0.86[0]. That's great for a social sciences correlation, but not so great as a way to be fair. 0.86 leaves a lot of room for other processes - and in competitive scenarios at the top of the distribution, the confounders could make all the difference.

[0] https://pubmed.ncbi.nlm.nih.gov/15147489/

Reply on news.ycombinator.com