Comment by serjester | Hacker News Reader

Comment by serjester | Hacker News Reader

Unpopular opinion but this is complete pseudoscience. Starting with the notion that the most complex thing in the universe, someone's intelligence (ie high dimensionality), can be simplified down to a single number.

This study is cursed by circularity. The only thing IQ tests are good for is testing is how you perform on sterile, academic tests which unfortunately underpin our current education system. Even then it explains at best between 2 and 13% of the performance in some tasks. No measure that fails 80–95% of the time should be treated seriously.

But let's assume IQ is an accurate gauge of someone's intelligence - why do some countries have an average IQ of 70? Are we seriously going to assume that entire countries are filled with people that could be classified as having an intellectual disability in the US? Clearly there's more going on here.

Not to mention the effect of IQ they found is smaller than the difference between IQ tests for the same individual. It's a shame that epidemiological psychology get's treated as anything remotely resembling science.

Edit: Small changes for cohesion.

> Unpopular opinion but this is complete pseudoscience.

There was significantly more research put into this paper than thought put into your comment. You dismiss the entire paper using a bunch of hand-waving (with a few instances of emotional manipulation like "psuedoscience" and "it's a shame" and "are we seriously going to assume") and no actual arguments. If you cannot actually point to specific flaws with the experimental methodology, you definitely shouldn't accuse the paper of "psuedoscience".

> Starting with the notion that the most complex thing in the universe, someone's intelligence (high dimensionality), can be simplified down to a single number.

That assumption is never made in the paper. The authors use various tests (the Henmon-Nelson Test of Mental Abilities, the Armed Services Vocational Aptitude Battery, the Edinburgh Reading Test, and the Friendly Maths Test) to provide numbers that they use as a proxy measurement for intelligence, but they never state that the two are equivalent. Moreover, just because our cognitive processes are extremely complex does not mean that you can't capture the majority of variance using a small number of measurements (which is, uh, how a lot of modern statistics works).

> Then throw in the effect of IQ being smaller than the difference between IQ tests for the same individual.

The authors do not use IQ as their sole measure of "intelligence", so this isn't relevant. There isn't even a single "IQ test"!

Moreover, even assuming that what you claim is true (because you didn't even provide a source), it's not really relevant - "proportion that measured IQ contributes to a specific kind of a success" isn't comparable with "variance between different IQ measurements for an individual" - operator<() isn't defined on those two types. Plus, all you're claiming is that it's hard to accurately measure IQ - which isn't relevant if the variance is 10% but we're curious about whether the difference between 90 and 110 IQ is significant.

> why do some countries have an average IQ of 70?

I think of at least two plausible reasons - malnutrition stunting intellectual growth at crucial stages in a human's development, and lack of education. Throwing a "Are we seriously going to assume" out there isn't an argument.

> Clearly there's more going on here.

If there is, you haven't pointed to it.

I'll comment on two of your points. I'd strongly argue you lose most of the salient information when you try to represent a highly dimensional system with a single numbers. That's the curse of dimensionality. I agree, modern statistics does not understand this, causing no shortage of issues when applied to complex systems - eugenics, high carb diets, 08 recession, etc.

Secondly I'd recommend making an effort to travel to these "low IQ countries". I can assure you the average person is not much dumber than the average American.

If IQ tends to correlate with abstract thinking, a person who is smart in daily life may still be < 100 points of IQ.

I had classmates who were quite smart in interpersonal communication, but really struggled with abstract tasks in mathematics or physics. One of them is a fairly successful businessman now, but still shudders when someone says "Pythagorean theorem" aloud.

Given how heavily do IQ tests lean on geometry and abstract pattern matching, he might score rather low.

- strongly argue you lose most of the salient information when you try to represent a highly dimensional system with a single numbers.

First you'd have to demonstrate that intelligence is multi-dimensional, at which point you'd still fail since you can represent any point in any number of dimensions by single numbers. Hilbert curves are a trivial mapping. The entire field of manifold embedding exists as a counterpoint to your claim.

And, as others point out, it's also possible that one or a few values capture increasingly larger and larger amounts of the entire effect.

Claiming high dimensional items aren't describable with a single number is simply mathematically wrong.

> I can assure you the average person is not much dumber than the average American.

If you have two scales and used them to weight the difference in mass of two objects, and one says there is a difference and the other doesn’t, you cannot just conclude that there must not be any difference. Obviously one scale is wrong. So in this case we have to ask which scale is wrong, the collective results of hundreds of well documented scientifically conducted studies or your “gut feeling”.

> malnutrition

Yes, but

> lack of education

of intellectual stimula, really.

Reply on news.ycombinator.com