Hacker News Reader

Comment by jtchang | Hacker News Reader

One of the tweets:

> Hsu’s beliefs in innate biological differences between human populations, especially regarding intelligence --an unscientific position which makes him unsuited to direct a research institution’s funding or its graduate studies

I can see why this belief is controversial and in an ideal world I wish everyone was born exactly the same. However I don't think applying strong scientific research and methods to try and answer this hypothesis is necessarily wrong. Unless it has already been proven otherwise in which case I am unaware of the research.

View on HN

It has been studied pretty extensively. For example, see this survey https://psycnet.apa.org/record/1997-30052-009

In pretty much every study within group differences are far and away more massive that between group differences. Unlike other things like height, eye colour, or facial hair growth, intelligence is highly variable even between identical twins. The genetic link seems weak at best.

It’s also more or less impossible to study with great accuracy - you’re trying to study how race affects intelligence, but how do you correct for non-biological factors? Race is deeply intertwined with economics, education, nutrition, healthcare, and every other aspect of society and culture. Any study you do, regardless of results, will have a mountain or potentially over corrected or under corrected factors.

Finally, what are we even studying? Defining intelligence is a nightmare, and a quick glance at history shows how the definition of intelligence has changed over time and between cultures. Trying to look at how race’s biological factors affect intelligence irrespective of sociological factors is in many ways a fool’s errand - the very definition of intelligence is wrapped right up in those sociological factors!

> In pretty much every study within group differences are far and away more massive that between group differences.

You are perpetuating what’s known as Lewontin’s fallacy.

This claim is extremely narrowly true: clustering on a 1-dimensional fixation index is insufficient to reliably mechanically classify race. If you don’t know what that means, stop trying to use this claim in your arguments.

Here’s what it doesn’t mean: it doesn’t mean that there aren’t significant population-level phenotypic differences between ethnic groups. These differences can include intelligence (and 30 years of twin studies suggests they probably do).

> Here’s what it doesn’t mean: it doesn’t mean that there aren’t significant population-level phenotypic differences between ethnic group

But it does mean that we don't have any evidence for this claim.

> (and 30 years of twin studies suggests they probably do).

This would be a great place for a citation. I'm wondering how a twin study is going to show the effects of genetics on intelligence. As GP suggests, twin studies show differences in intelligence between even identical twins raised in the same household, so Amy conclusions from whatever study you're referencing would need to show statistically significant variation controlling for the noise.

> This would be a great place for a citation.

If you actually care, just google it. There are dozens of great large-n metastudies. Chances are you don’t care (mostly based on the fact that you clearly haven’t googled it already), so I have to be jealous of my time - sorry.

> would need to show statistically significant variation controlling for the noise.

Obviously g-factor researchers have thought of this. Current best estimates place 60-80% of human intelligence variance as being genetically heritable.

I've been able to find evidence that disagrees with you, but not evidence that agrees with you. Since you've made very specific claims that most contrary evidence is invalid because it is not multvariate.

It would have taken you less time to post a single link than to write your response, so I don't buy the time excuse.

> Obviously g-factor researchers have thought of this. Current best estimates place 60-80%

I'm aware. You claimed that twin studies (which control for race by comparing between people of the same race and are usually limited to small sample sizes) have revealed racial ethnic differences in g controlling for the environmental factors.

This claim is questionable on its face. Your heritability claim is much less contentious, but heritability of g and race correlation of g, controlling for environmental factors, are very different claims.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4270739/

This took me literally 10 seconds to find.

Adult intelligence is ~80% heritable. There is >1stddev difference in mean intelligence between many ethnic groups. (Do you need me to provide a citation for this as well, or can you google it this time?) Do the math. What should your bayes factor be on this being environmental?

You can also find studies that specifically try to find environmental causal factors on intelligence - the effect sizes for identifiable factors (like poverty, or other common scapegoats) are laughably small. Once again, feel free to google this.

I’m sorry if I come across as hostile by asking you to look things up, but usually when people demand citations for things that A) they should already know if they’re debating about this and B) are easy to find, I have to assume they’re trying to waste my time and/or arguing from bad faith.

I never asked you to cite intelligence heritability. Like I said, I agree that intelligence is at least partially heritable.

I asked specifically for you to justify the original claim, that there are phenotypic differences between ethnic groups that include intelligence, as supported by twin studies. That's a very specific and as far as I know very unfounded claim.

As I explained - look at the percent of variance that’s heritable. Now look at the measured difference in population intelligence. Now tell me - what’s the probability of that being environmental, based on the fraction of intelligence variability that can’t be attributed to genetics? A fermi estimate is fine - you’ll end up close to zero anyway. You shouldn’t need a document to spell this out for you in excruciating detail.

Edit: to humor the question: https://www1.udel.edu/educ/gottfredson/30years/Rushton-Jense...

The conclusions in this paper should have been expected give the explanation I already gave.

Rushton is perhaps not the person you want to cite if you want to disassociate yourself from scientific racism.

And that paper, which I'm familiar with, doesn't address many of the modern criticisms of such theories that are based on the idea of "g" being socially constructed/influenced in the first place. (nor does it actually do much, if anything, to support the original claim you made: not all of their conclusions are phenotypic in nature, at a minimum, sections 3, 7, 8, 9, and 11 can be explained culturally despite what the paper says).

I'd suggest "Reconciling IQ Gains and Heritability", or "Race, IQ, and the search for statistical signals associated with so-called “X”-factors", which notes as a semi-conclusion that we know of environmental factors that have clear and demonstrable causal effect on IQ. We know of no specific genes that have this effect. So concluding that a bunch of individual genes have small effects is actually less supported than the similar claim that a bunch of individual societal/environmental factors have impact. Both of these are more recent than the paper you cite, and could be seen as academic responses to perceived flaws.

> So concluding that a bunch of individual genes have small effects is actually less supported than the similar claim that a bunch of individual societal/environmental factors have impact.

What an absurd claim! This is specifically the hypothesis that the last several decades of twin studies is designed to test. You are either ignorant of this body of research or you are being disingenuous. We can distinguish between these two cases because adopted twins and siblings have highly correlated genetic factors and weakly correlated environmental factors.

I notice that you keep avoiding the absurd improbability of environmental factors mediating bulk racial differences given the magnitude of population-level differences and the relatively small fraction of variance attributable to environment - but you seem desperate not to acknowledge the latter.

> We can distinguish between these two cases because adopted twins and siblings have highly correlated genetic factors and weakly correlated environmental factors.

We can't, and the papers I cite give a highly coherent explanation for this, a phenomena called cultural filtering.

Like I said, I suggest you go and actually read recent papers instead of relying on a survey paper from fifteen years ago that relies on actual research older than many of the professors publishing this stuff today.

It's not worth having this conversation is you're unrolling to engage with modern research, and it'll just devolve into an ideological flamewar, which dang prefers us not to have. So again please read some modern research. Or don't, but don't claim this stuff is settled when your are a decade out of date.

I looked at the things you "cited" (i.e. name dropped without linking).

"Race, IQ, and the search for statistical signals associated with so-called “X”-factors":

A) no non-paywalled sources, so I'm suspicious if you even read it

B) written by a philosophy professor, not a scientist, with no scientific background

C) the argument of this paper, as far as I can tell from the abstract, is "the pure-environmental hypothesis is nonfalsifiable", i.e. completely useless

I would critique it more if I could actually read it.

"Reconciling IQ Gains and Heritability" - this isn't actually a paper, it's just a section in some article. Where did you even find this? It barely shows up on the internet - your comment is the third result. In any case, it A) doesn't seem to support your claim, and B) doesn't strike me as especially compelling research - they're just summarizing some thoughts about the Flynn effect.

This further increases my confidence that you are arguing in bad faith.

> the papers I cite give a highly coherent explanation for this, a phenomena called cultural filtering.

I can't find this term anywhere in these papers (or elsewhere).

> no non-paywalled sources, so I'm suspicious if you even read it

I have access to many paywalled articles.

> "Reconciling IQ Gains and Heritability" - this isn't actually a paper, it's just a section in some article.

Ah you're right, I meant to cite that section of the larger article, but the wider article is rather well cited.

> phenomena called cultural filtering

tl;dr: The idea is that culture magnifies any underlying genetics. The example given is two siblings/twins separated at birth, both are taller than average. As a result, they both end up playing basketball and being better than average as children, and as a result attend basketball camps. They both get good training and end up as college basketball players.

The question is to what extent culture serves to magnify any underlying differences. And whether different culture(s) can magnify these differences in different ways.

> The question is to what extent culture serves to magnify any underlying differences.

This sounds like irrelevant bikeshedding between genotypes and phenotypes. It also doesn't have any bearing against my claims, c.f. my original comment about "significant population-level phenotypic differences between ethnic groups".

> This sounds like irrelevant bikeshedding between genotypes and phenotypes. It also doesn't have any bearing against my claims, c.f. my original comment about "significant population-level phenotypic differences between ethnic groups".

Of course it does: if the only concrete support for phenotypic differences is based on twin data (or similar), and that data is subject to magnification due to cultural factors, the underlying phenotypic impact is smaller.

That's not what "phenotypic" means. Phenotypic environment includes any "cultural amplification" or similar factors. Almost all genetic competitions are subject to some sort of king effect, e.g. hypergamy and concomitant GMV in non-monogamous species. We still consider the results of these effects to be phenotypic.

Ok then I'm really unclear what you're point is. Yes, there are observable performance differences that align across observable ethnic boundaries.

No one in this thread questioned this (keep in mind stereotype threat is a phenotypic difference). If that's the only claim you're making, I'm not sure why you felt the need to cite Rushton of all people, who was specifically arguing that these differences were genotypic in nature.

That's a terrible argument that's often used to deflect from its crux. Of course within group differences are larger than racial differences. You're comparing a single (admittedly aggregated) factor to the sum total of human variation. That it has a lowered effect isn't surprising.

Furthermore, the paper you cite doesn't even deny that different groups test consistently differently. Furthermore, it also explicitly states studies of test bias have consistently shown negatives.

I also strongly oppose the conflation of population level metrics with individual level metrics.

It's better to just let the advertisers and political manipulators gather the same data and do essentially the same research, but in secret and for their own aims, and benefit from the resulting understaning, than for the research, and it's possibly unpleasant resulting knowledge, be in public and used to defend against those who would otherwise get to capitalize on it, at the expense of the very people you presumably would like to protect from abuse?

Ignorance is never the correct answer no matter how unpleasant some bit of knowledge is.

Whether the answers come back in your favor or not, having the answer is better than not having the answer, or worse, only some people having the answer and not everyone else.

This guy might indeed harbor odious intentions, but some of these arguments are not holding water on their own.

I notice that people much more often twist themselves into knots to make comments like yours when it's black people being talked about, but the idea of Ashkenazi Jews being up to a standard deviation smarter [1] is much less controversial. It's still controversial, but it's nowhere near the nuclear topic that black-v-white intelligence is.

[1] https://en.wikipedia.org/wiki/Ashkenazi_Jewish_intelligence

Your claim is false. It is not highly variable even between identical twins. https://en.wikipedia.org/wiki/Heritability_of_IQ

"The mean correlation of IQ scores between monozygotic twins was 0.86, between siblings 0.47, between half-siblings 0.31, and between cousins 0.15."

I don't think it would be an ideal if everyone were born exactly the same. Seems like a lot of beauty in the world around us comes from our different faces, thoughts, beliefs, culture, etc, all born of how we're all different.

I think they mean "born the same" as in, born with the same potential and capabilities, not literal clones of each other.

Reply on news.ycombinator.com