tailcalled
1yr ago
Not understanding the phenotypic null hypothesis is not a mere "bias in favor of their pre-existing beliefs". It's a fundamental failure to comprehend what is even being talked about. A failure to understand/accept it is just as bad as if a leftist considers inequality of outcome to be a proof of racism, and claims that inequality of outcome disproves genetic explanations.
tailcalled
1yr ago
As someone who believes that the racial IQ gaps are innate but who is wondering just how deep the rabbit hole of HBD incompetence goes, I would be interested in hearing more. What's the best case that it is due to environmental factors?
tailcalled
1yr ago
That's a rather lazy strawman. Whenever this comes up here, people say that their primary motivation is to show that the progressive argument "outcome disparities, therefore discrimination, therefore reverse discrimination is necessary" does not hold and leads to unfair treatment. You can very much come at it from an egalitarian point of view.
I'm basing my experience more on the Twitter debates that hover around Erik Turkheimer than I am on the discussions that happen on The Motte, as I don't have much experience with The Motte. It's quite possible that The Motte is different.
You lost me here. Could you elaborate for the innumerate?
I think if you're innumerate you should just not rely on quantitative studies such as twin studies because they require numeracy to interpret.
tailcalled
1yr ago
Given how high stakes the LSATs and SATs are, who doesn't put maximal effort?
Neither I nor the person who did the study were making a claim about the SAT, we were considering a setting where people fill out surveys for pay on the internet.
Yet some people get way higher LSAT scores than others even though everyone is trying hard, suggesting innate differences still matter even if incentives also matter. Probably 100k+ people annually apply to top law & med schools, top colleges; I am sure most of them are putting in maximum effort on high-stakes tests such as the LSAT, GRE, ACT, SAT, MCAT, etc.
I didn't say effort was the determining factor, I said that it mattered, with natural variation in effort covering maybe slightly less than 10% of the variance (so if you go from slacking to max effort, maybe that gets you a bit more than a standard deviation of score).
Part of the problem with a lot of HBD-dunking, such as by Taleb, is they fail to control for individual preferences. So in aggregate, correlations between innate qualities and outcomes are low, but you are not accounting for individual preferences, not that genes do not matter.
Much of individual preferences are heritable too, but yes I agree that group-level differences are much less "noisy" than individual-level differences.
tailcalled
1yr ago
Maybe the cause of the differences is on an environmental factor that is hard to intervene on.
tailcalled
1yr ago
🤷 I also want to know the truth; what triggered this discussion was mainly that the other post had been placed in the best-of section of this website. Take it up with the moderators if you think the conduct of each side is irrelevant.
Though I don't think the conduct is irrelevant, because of Aumann's agreement theorem. If there was a side that conducted itself well, you could just copy your opinion from what they say. Because they're not behaving well, you can't do that. That seems like important information to know.
Also I wouldn't say I don't have any space for people who want to know the truth:
There's some honest people on either side who have been swept up in the drama, but in terms of the direction of the energy which drives the whole debate, this is what lies underneath it.
tailcalled
1yr ago
Just for context, is there some specific intervention Erik Turkheimer has endorsed that you are objecting to/which you think has been shown to be ineffective?
tailcalled
1yr ago
I mean I can try, but I already wrote an explanation in my blog post, so I'm not sure how much more I can say about it. Is there some specific part of my linked post that you find confusing?
tailcalled
1yr ago
Regarding the gender gap in math ability, has any country tried to just make girls take twice as many math classes as boys do for a generation to see if that would reverse it?
tailcalled
1yr ago
Racial differences in sports accomplishment sound likely due to bodily differences in sports abilities. Next question please.
tailcalled
1yr ago
I've seen HBDers point out that there is a genetic correlation between homosexuality and mental illness, and use this as an argument that homosexuality and mental illness are innately related, as a counterargument against e.g. homosexuals being bullied and becoming mentally ill as a result of that. (I've also seen lots of other examples, but this was one of the key examples that made me decide to write the post.)
If a genetic correlation referred to a genetic confounding aka horizontal pleiotropy aka "one gene has two unrelated effects", then that counterargument would make sense. The bullying theory of gay mental illness doesn't predict that there is a biological connection between homosexuality and mental illness.
However, genetic correlation actually refers to something more subtle. Two variables are genetically correlated if genetic factors that contribute to one variable also contribute to the other variable. So for instance, intelligence and education are genetically correlated, because genes that contribute intelligence makes people better able to pass exams etc., which unlocks better educational opportunities. In this case, intelligence and education are not genetically confounded; rather they are just ordinarily causally related, and this makes them genetically correlated, just as it makes them ordinarily correlated.
First -- your writing style doesn't work well for me. It's too abstract, and you don't clearly state your point. For example, your point #1 "Heritability simply does not mean what a lot of HBDers want it to mean - because of the phenotypic null hypothesis." What is your point here? What do you think "HBDs want it to mean"? What is the "phenotypic null hypothesis"? It's not good writing to make me chase down you thoughts, especially on other sites, especially before you've proven you're worth the work. I went to that site, and am not much more enlightened. You seem to somewhere make the point "Things influenced by genes often go though non-biological channels". Or maybe "things that look inherited aren't always". Sure, I'd say both are fairly non-controversial. A classic example of the second is, e.g. "speaking French" which looks inherited on the surface, but is clearly not biological. And yes, our environment and society mediate all kinds of things, we live in a complex interconnected world.
Do you have more of a point? I couldn't really tell (of course, that can be on me, but ... I've read and understood a fair number of others on this topic, but not your writing...) I really don't know what your code and diagrams at the end are supposed to show. Summarize your cool conclusion! E.g. "Even though X is not directly responsible, in a naive analysis it looks like it is, exactly like QQQ, which actually is directly responsible. Here's how that can play out ...". I think you're saying something like that, but you don't bother actually saying it (or I missed it).
Here are some examples of places where I'd bring it up on twitter:
The post comes from frustration with these sorts of situations. Importantly, it's not the specific interactions but instead that they are repeated and that often not much update happens. This seems like a point that should get signal-boosted more in a healthy community.
In any case you sort of seem to be saying "we can't figure anything out" which both seems wrong, and kind of useless. Do you apply this to all such studies? Maybe we should -- I admit, I tend to write off almost of all psychological and sociological studies these days, because they seem so ideologically captured. On the other hand, between statistics, twin studies (and separated twin studies), and sibling studies, we seem to be able to do a pretty good job on some things.
Yes, frequently scroll through my twitter and just fire of random quick tweets that debunk random social science studies. And I don't trust any social science studies until I've checked the methodology myself.
Second -- you seem to be coming at this from a place of significant bias. "Rightist inclined people want to preserve racial inequality of outcomes," is an incredibly weak straw-man, it's basically "Everyone I disagree with is a racist". Is that really the best you can do, in terms of extending charity to the people you disagree with? I personally, like most of the others here, see the acknowledgement of group differences (and for what it's worth, I don't really care much if it's culture or biology, and both seem taboo anyway) as primarily an alternative for differing outcomes, without discrimination being the ONLY explanation.
I'm in tech. There aren't many women, nor many black people. This is ascribed to sexism and racism, which doesn't match what I've seen, experienced, or heard from the affected people (from women at least; I haven't asked many black colleagues about racism). I see my company following policy to massively privilege both groups, and to blame white cis-men for all the problems, and those both seem wrong, and even damaging to me (and to a number of people in the targeted groups, e.g. women who just want to be SWEs, and not feel they got their role because of their sex, and no, I'm not concern trolling, the suspicion around the privileging is real). I see differing interests (and maybe ability at the margins) and degrees as the main reasons for the differing representation, but we're not allowed to notice that, as "It's not the pipeline". James Damore got fired for trying to make this point.
You also see this censorship of blasphemy in the US, especially around crime, where apparently pointing out some choice statistics around violent crime is considered a hate crime. (Again, FWIW, I'd consider those stats more a cultural issue, but it's a pretty important one, upstream of the 'getting shot by police' issue).
So anyway, what I'd like from you, and I think would benefit you, is to tighten up your writing -- make your point first, then provide an explanation of it (it's a classic academic / systemic thinker error to do it the other way around). Make things more concrete. Work from a specific example and tie your points back to it. People are reluctant to trust generic models, as they are often used to lie (see Abigail Thompson's dissection of Hong-Page's "mathematical proof that diversity trumps ability". There's a nice discussion of it here
Also, try to be more charitable to your outgroup.
You're making assumptions about my ingroups/outgroups that aren't necessarily true. E.g. I've spent tons of money on HBD-aligned research, partly for motivations similar to what you describe here.
No, this doesn't "obviously massively contradict" my common sense, and I think many would disagree. In fact, I thought one of the main points of IQ tests, rather than "effort tests", is that neither effort nor prep makes much of a difference to them. Otherwise, for example, they wouldn't stay very stable over time (which I understand they do).
I was under the impression that the test-retest reliability of IQ tests is 0.85? If we assume that the test-retest reliability of effort is 50% (I suspect it's higher, but let's be generous to your side), and we go by the estimate that 8% of variance in IQ is down to effort, then that would mean that effort would introduce an unreliability of 50%*8%=4% to IQ tests. This is considerably less than the 15% unreliability that I've commonly seen, so I don't see how your argument is even halfway close to working out here.
Prep courses would also have more value, which I don't think do. Do you think when people can't make intellectual leaps others do, they just aren't trying hard enough in that particular moment?
Isn't the item level of IQ tests (which seems like the closest analogy to intellectual leaps to me) dominated by noise? So I'd say noise, not g or effort.
Also even if we ignore the item noise point, by the numbers I'm giving, test effort would only be 8% of the explanation for test underperformance.
I think most would agree effort plays some role -- if I don't care or try at all on an IQ tests, and answer at random, I'll have a low score. If I try to be fast and disciplined, and use all my test-taking savvy, I'll probably (?) do better than if I just breeze through (although I wonder). But basically, once you're trying to do well, it's not really clear what "trying harder" even means on IQ test. It's not like pushing on a bar (and honestly, even for that the range where trying, vs training and genes and drugs, makes a difference, is pretty small in that moment. If I can barely do one pull-up, trying really hard might mean I do one, or two, but I'm generally not going to be able to do 10). So anyway, stop claiming consensus on things people will disagree on (especially things where your "consensus" seems to go against standard definitions).
It sounds to me that you are granting the basic point (effort plays some role), but then acting like you are disagreeing anyway? Is it a quantitative disagreement? If so, what is your estimate for the quantitative effect that effort plays?
tailcalled
1yr ago
Well so here are two models:
-
People start with meritocracy as a value and then try to figure out the factual questions about how abilities work.
-
People start with some policy opinions or alliances or something and then work to rationalize those.
Model 1 is debunked because if people were trying to figure out the factual questions, they would take basic concepts like the phenotypic null hypothesis into account, and they would be upset about having signal-boosted terrible studies like the IQ/effort one. It's possible that there is some alternative to model 2 that people follow, in which case you should feel encouraged to share what that alternative is.
tailcalled
1yr ago
This is mostly it, with one major point:
What about "genetically correlated"? We know tall people are better at basketball - the variables are correlated - but is this effect genetic, or environmental? Did tall people inherit their basketball skill directly, or was it just an effect of their height? If you get leg-extension surgery, do you get better at basketball?
If the basketball skill is an effect of height, then height and basketball skill will be genetically correlated. (For basically the same reason that if basketball skill is an effect of height, then basketball skill and height will be ordinarily correlated.) Not "will be misestimated to be genetically correlated when they are actually environmentally correlated" - but instead "will genuinely be genetically correlated". This is because genetic correlation is not a measure of whether height is horizontally pleiotropic with good slam-dunk technique, but instead just a measure of whether height genes and basketball genes are overlapping, for whatever reason it might be.
tailcalled
1yr ago
I agree that it must be all the way on or all the way off, and I would prefer if [it was all the way on and people cared about figuring out the truth].
tailcalled
1yr ago
Disagreements where Aumann's agreement theorem's assumptions are satisfied rapidly disappear, so whenever you talk about persistent controversies, they will not be satisfied. However, it is still relevant for me to know whether they fail to be satisfied because one side if obstinate, or because both sides are bad, as if it's only one side that's bad then I can just copy my views from the other side.
To begin with: because for most would-be listeners, supporting the opposite camp equals approval of one's family and one's people getting discriminated against (the nigh-inevitable alternative to HBD being acceptance of unredeemed racial guilt). Some are turned off by this; others are quite happy.
I think the honest way of addressing this is "fuck your anti-white racism", not coming up with elaborate justifications that were never epistemically serious to begin with.
HBD/Blank Slate is a political question more than it's a scientific one, a question of whether a civic religion founded on this stated promise of equal innate potential (or at least absence of substantial between-group differences) is needed for the prosperity of a polity and reduction in individual suffering; it's a question of oughts. Turkheimer et al. wear their oughts on their sleeves, adversarially, so it's a bit of a double standard to dunk on HBDers for failing or refusing to understand a peripheral aspect of «is».
This sounds like a scientific metapolitical question as much as it sounds like a political one. See the end of this book review for more discussion.
I have read a decent amount of your writing on the topic today, debate and adjacent content; the more I was reading, the stronger was the feeling that this'll be a fruitless debate. Your writing over the last year displays a trend towards isolated demands for rigor against HBD outlook (for unknown reasons that I assume, for now, are self-preservation and career strategy, a la Abdel The Backstabber, but might be just autistic perfectionism and purity spiraling; take your pick at which is less charitable).
I don't think I have an isolated demand for rigor towards HBD. In other places than the linked ones I regularly challenge anti-HBD. The issue is that neither side wants to be rigorous aka both sides just want to make things up with studies that support their general vibe without checking whether those studies are actually of any evidentiary value.
Now to be clear, the faults of HBDers are not big. [...] Your two examples aren't damning.
Your thread on effort and application of instrumental variables seems to sidestep the problem of scores influencing perception of effort, and anyway it doesn't affect the debate in absence of good evidence for differences in incentives&effort
If scores influence perception of effort, then it would lead to a downwards bias in the estimated effect, not an upwards bias. Also it is not really mathematically plausible that the direction of causality is reversed because that would require an effect size of around 2, whereas unless there are shenanigans going on they should only go up to 1.
And yes I agree that it would be ridiculous to say that the racial differences in IQ are due to differences in effort. The effect of effort is fairly small, so you would need a humongous difference in effort to cause it. The problem is that HBDers do not separate this question from the question of whether effort affects IQ scores, and so they say stupid things like that effort doesn't matter for IQ in order to protect the notion of race differences, when really the obvious answer should be "yes, effort matters but it doesn't explain race differences".
(and, well, whatever happens with low-stakes online IQ testing, people try «as hard as they can» in contexts relevant for long-term life outcomes or even on serious IQ tests).
I don't believe that. I often didn't try as hard as I physically could on exams. For instance once I was done with exams I might leave earlier if allowed instead of checking over my answers for errors, where of course if I had checked over my answers I would inevitably have found some errors and thereby increased my scores. I also never really prepared for exams ahead of time.
Now to be clear, the faults of HBDers are not big. [...] Your two examples aren't damning.
Your triumphant post on the environmental null hypothesis is… inconclusive, since it doesn't bother comparing models following from theories on their predictive merits (and also other twins and pedigree schemes). You can't just «call» ENH like you so often do on Twitter, and certainly can't just assume that nobody accounts for measurement error, that all causal effects will generate genetic correlations, and thus that the usual HBDer touting of gen correlations/gen confounding is invalid.
First, it's phenotypic null hypothesis, not environmental null hypothesis, i.e. if someone is smarter due to having a bigger brain and this bigger brain makes them do better in exams and therefore get more education, then that would be weird to call "environmental"
Secondly, while I can't assume that nobody accounts for measurement error, I can assume that the overwhelming proportion of studies don't account for measurement error, because I've seen that they don't. For an example in the case of personality, see how this shows almost all twin studies on the heritability of personality to be wrong: https://emilkirkegaard.dk/en/2017/02/getting-personality-right/
And I can totally assume that causal effects between individual difference variables generate genetic correlations between them (as long as they are heritable, but remember, everything is heritable). I gave you the theory and a simulation for why; if you disagree then show me a hypothetical counterexample where there is a causal effect but no induced genetic correlation. This is where I really consider it a big, damning fault, because this is really basic stuff.
A failure to understand the phenotypic null hypothesis is the same tier of error as an environmentalist who assumes that any difference in outcomes must prove that there is racial discrimination. No! Racial differences in abilities also predict racial differences in outcomes. Similarly, nonbiologically mediated/moderated causation predicts heritability and genetic correlations too, not as any sort of bias or measurement error or anything, but simply from the definition of what heritability is. If you don't accept this then you are not seriously engaging in the debate and no better than the environmentalists.
Now it's perfectly reasonable to raise questions about genetic confounding. It's just idiotic/dishonest/??? to say that genetic correlations prove genetic confounding.
Take any actually published model emphasizing environment (i.e. an X factor creating race differences – I mean, okay, the gay-bullying-mental illness stuff at least proposes a half-legible causal pathway), and subject it to scrutiny next to a genetics-first one – most likely, it'll fall apart first.
I am not bringing up the phenotypic null hypothesis as being a debunking of the HBD view on racial differences in intelligence. I agree that it doesn't debunk the HBD view on racial differences in intelligence.
I am bringing it up as a basic fact about heritability how heritability works that HBDers fail to understand/fail to engage with.
tailcalled
1yr ago
Phenotypic null hypothesis isn't an obscure methodological limitation, it is a fundamental property of concepts like "heritability" and "genetic correlation" which can easily be derived in simulation studies. If you don't understand how it works then you shouldn't use concepts like heritability and genetic correlation at all in the first place.
tailcalled
1yr ago
Yes, this is called environmental correlations, it is correct that the phenotypic null hypothesis also predicts the environmental correlation to be high whereas the genetic confounding hypothesis doesn't predict that. (Specifically, the phenotypic null hypothesis predicts every variance component to be correlated, whereas confounding hypotheses only predicts confounded variance components to be correlated.)
Some of the studies on homosexuality and mental illness finds the environmental correlation to be zero, which supports the genetic confounding view. I have at times acknowledged that/pointed that out. I also address the concept in my linked post.
The thing that bothers me is not the conclusion but instead the argument: WHY would HBDers make the argument with genetic correlations in the first place, when clearly it is the environmental correlations that are the key question? Because they don't know the phenotypic null hypothesis. But WHY would HBDers not know the phenotypic null hypothesis when it is such a basic concept for heritability? Because the phenotypic null hypothesis is anmoying and sounds like an outgroup thing, is my hypothesis.
tailcalled
1yr ago
Don't most girls not participate in such programs? You can only really expect to have an effect on the ones who participate, so if most don't participate, you are closer to not having the program at all than you are to having it for all the girls.
tailcalled
1yr ago
Yes, but that probably says something about female interest in STEM, which is probably somewhat correlated with aptitude. You could of course argue that this, too, is caused by an all-pervasive patriarchy, but given that the presence of such programs in a society has a negative correlation with female interest in STEM (i.e. the so-called "gender equality paradox"), I find that harder and harder to believe.
AFAIK just about all sex difference correlate with gender equality, including obviously-societal ones such as gender differences in names:
https://twitter.com/DegenRolf/status/1432940616653152259
(And of course it's been shown to apply to e.g. gender stereotypes too, though that could very well be due to stereotype accuracy.)
I don't know why they correlate like this, but I feel like this gives you something equivalent to the phenotypic null hypothesis for the gender equality paradox: if the paradox applies to some variable X, and X is causally upstream of some variable Y, then a priori you'd expect the paradox in X to create a paradox in Y.
There are more people with an interest in, and a knack for, STEM to be found within the male than within the female population. Given that environmental interventions have rather spectactularly failed to reverse course in this regard, Occam's razor would suggest that biology plays a factor here.
The thing is, as kids, boys will look at what men do and mimic that, whereas girls will look at what women do and mimic that. I don't know whether it is the sex difference in programming etc. that is biological, or if it is the sex difference in mimicry that is biological. I wouldn't expect it to be both, because what would that lead to if you took all the world's female programmers and male elementary school teachers and had them create a society where they raised a generation of ordinary children? Would the boys in this society do programming (and thus be mimicking women), or mimic men (and thus do teaching)?
Since I don't know the answer to this question, I can't tell if the sex difference in programming is innate or not.
tailcalled
1yr ago
AFAIK the mainstream hypothesis among child developmental psychologists is a self-socialization theory, where children mimic adults in their culture. The whole patriarchy thing is more of a feminist activist thing that doesn't necessarily generalize to the expert's views. Of course the experts might be wrong too, but I'm just saying, as far as I know nobody has tried whether raising children in a culture where math and technical subjects is more of a women's thing works.
tailcalled
1yr ago
Hm, that seems interesting but unfalsifiable. Even if parents try to pretend really hard that everyone is living in such a culture, can't you always say that the children are picking up on their deep-seated true social beliefs?
For just about any hypothesis, you can of course cook up versions that are unfalsifiable, but I don't think the basic version of the hypothesis is unfalsifiable. Of course direct interventions would be way expensive; but the child psychologists have various experiments they claim supports it that I don't think I've seen a strong HBD response to. (Not to say that I agree with the environmentalists here, I am currently agnostic I think.)
tailcalled
1yr ago
edit: I wonder to what degree growing up in a progressive family and school environment already approximates this.
edit: ! Can you test to which degree girls who grow up in prog environments go into programming vs trad? If this thesis is true, shouldn't it have some effect?
The theory is that children mimic same-sex adults, not that children do as same-sex adults tell them to. Doing what adults tell them to would be highly exploitable, so children often contradict instructions from adults.
Masculine/feminine work interests are fairly independent of progressivism/conservativism, so children would not be exposed to more gender-atypical work interests from their parents that way. However I've seen feminists provide a study showing that girls who had more masculine mothers grew up to be more masculine in lifestyle as adults.
But that study seems obviously genetically confounded. However you could adjust for the genetic confounding because the genetic effect should apply to both parents whereas the self-socialization effect should be sex-specific. I considered setting up a study for that once, but I got distracted.
I just created an account on The Motte to PM someone a question, and afterwards I started browsing through some links and found this post in the vault: Belief Against an Intelligence Gap / Why the Woke Won't Argue: A look at Turkheimer and HBD research.
Now, a year ago or so, I would probably have strongly agreed with this post. But recently, I've come to a totally different conclusion: HBDers tend to totally refuse to engage with basic principles of the debate. I say that as an HBDer who has started interacting with other HBDers, and correcting when they make mistakes. Two core examples I have in mind:
Heritability simply does not mean what a lot of HBDers want it to mean - because of the phenotypic null hypothesis. You often see HBDers declare success when yet another twin study shows that yet another variable is highly heritable, or that there is a genetic correlation between two variables which are usually suggested to be causally linked to each other. In the latter case, I often see HBDers act as if the genetic correlation proves that there is genetic confounding between the two variables, which is a ridiculous suggestion if you think through the actual math. It's perfectly reasonable to say that the debate struggles with progressing because anti-HBDers aren't properly engaging with HBDers, but it would be a lie to also pretend that HBDers aren't also guilty of lack of thought and engagement.
HBDers often signal-boost nonserious or dishonest studies. My go-to example of this is this study on effort and IQ, which claimed to find that effort does not matter for IQ scores. This obviously massively contradicts common sense, and indeed when I took a quick look at the study, its data actually totally supported the notion that effort matters for IQ, and it's merely that the researcher (who is a well-respected leading IQ researcher!) analyzed it wrong (see my analysis in the thread, or perform the analysis for yourself). The researcher still has not changed his mind on the flaws of it, and I regularly see the study pop up on my timeline. If HBDers are going to boost these kinds of studies and ignore critique of them, then why should anyone listen to HBDers?
So, what view would I suggest? A far more symmetric view: Leftist inclined people want to create racial equality of outcomes, and they therefore boost whichever kinds of rationalizations they can come up with for the achievability and justification of such equality. Rightist inclined people want to preserve racial inequality of outcomes, and they therefore boost whichever kinds of rationalizations they can come up with for the unachievability of equality and justification of inequality. There's some honest people on either side who have been swept up in the drama, but in terms of the direction of the energy which drives the whole debate, this is what lies underneath it.
More options
Context Copy link