4 ms·

From a didactic perspective, I think he goes off the rails by immediately focusing on something as abstract and hard to measure as intelligence, and where there

by abeppu 5y ago

From a didactic perspective, I think he goes off the rails by immediately focusing on something as abstract and hard to measure as intelligence, and where there's a long history of people talking about whether it's even a single trait. In terms of just explaining the breeder's equation, I think he could have been more effective talking about, IDK, tall Dutch people. The other thing that's annoying here is that although this is an 'equation', as I understand it it's typically not predictive because that heritability term is so fuzzy. It's not that you know the heritability term in advance and predict the response to selection; measuring the response to selection also updates your information about the heritability. His examples have a cherry-picking feel for this reason. Take the "national merit finalists" thing he mentions multiple times. Note that he doesn't mention the IQ distribution of the college professors (surely it's logistically impractical to get any large number of busy faculty to spend time taking such tests) or of the college towns (which might not be the same as the overall population mean), or the degree to which being a professor is correlated with IQ or the degree to which being a national merit finalists is correlated to IQ. Had the data come out differently (and I suspect he mentions _finalist_ multiple times because there wasn't as distinct a signal among _winners_), this perspective leaves plenty of room to declare that (a) maybe heritability of intelligence is lower than initially thought or assortative mating along IQ lines is weaker than expected (b) maybe professor-ness is less correlated to IQ than assumed or (c) maybe national merit finalist status isn't especially correlated to IQ (semifinalists have to submit a bunch of info other than test results).

mynameishere 5y ago
This kind of quibbling always comes up. He's just using professors and national merit finalists as examples. The truth is, you could probably take the top 10 percent highest-earning car salesmen (EDIT: and women), drop them on a desert island and, after two generations, get the highest-IQ society the world has ever known. But that's less obvious, so why use such an example?
- ImaCake 5y ago
  But it’s not clear at all that intelligence is a single trait. It is incredibly messy and this means any analysis of it has to be very careful of confounders. In contrast, height is by many definitions a single trait, we know exactly what confounds it’s measurement (and those things are easy to measure!) and we have a pretty good estimate of it’s heritability.
  - bordercases 5y ago
    Remapping intelligence to certain cognitive performance parameters, many of which can be measured in a very metrical and objective way, still tends to produce correlations. See Figure 3 of this article: [0] - there is a mix of simple perceptual tasks and more complex reasoning ones, which on the surface appear independent. The paper admits at the end that intelligence isn't one kind of task, but instead many kinds of systems used together. That still doesn't prevent their ability to construct a version of g-factor. And very simple metrics, taken alone, seem to inter-correlate well with IQ scores, like reaction time.[1] This correlation increases with age, suggesting a biological factor in preserving brain integrity that would nullify educational effects of intelligence. If your version of intelligence has nothing to do with cognitive activity I don't know what to tell you. It's true that effectiveness in society might not be correlated with intelligence. That says more about what society rewards and the kind of labor required to get it. And I don't mean to say that society should reward people based off of test scores, that's dumb. Nor should we assortate people based off of pure test scores, also dumb. But I see no real conceptual problems regarding intelligence that doesn't try to conflate intelligence entirely with general effectiveness. There are IQ realists like Steven Hsu that don't do this, placing it on equal importance with conscientiousness [4]. As with height, there are innate and behavioral/developmental components of general effectiveness (I could dope my child with growth hormone to make him taller, or mal-nourish them to go the other way). As with emotional intelligence, being a bastion of full-nurture theorists for behavior that look for non-academic modes of measuring effectiveness, we can get heritability correlates of 0.2-0.4[2], and for psychopathy up to 0.5[3], mediated by biological factors like inherent oxytocin concentrations. It's hard to say you can succeed without without any ability to learn, but there are high paying jobs that involve mainly physical labor (while taking on some mortality risk which can be managed). And there are people who are smart assholes, and other people who are dumb yet compassionate. So I have no problem with accepting the cognitive version of intelligence as with the right attitude there is no anti-social conclusion to risk. Society takes all sorts of people to exist, even if to some extent, the more well-rounded the better. [0] "Fractionating human intelligence", www.owenlab.uwo.ca/pdf/2012%20-%20Hampshire%20-%20Neuron.pdf [1] "The relationship between intelligence and reaction time varies with age", https://www.sciencedirect.com/science/article/pii/S0160289617300636 https://www.sciencedirect.com/science/article/pii/S016028961... [2] https://www.researchgate.net/publication/23302097_A_Behavioral_Genetic_Study_of_Trait_Emotional_Intelligence https://www.researchgate.net/publication/23302097_A_Behavior... [3] https://www.geneticlifehacks.com/psychopath-genes-born-not-made/ https://www.geneticlifehacks.com/psychopath-genes-born-not-m... [4] "Genetics and Intelligence", https://youtube.com/watch?v=62jZENi1ed8 https://youtube.com/watch?v=62jZENi1ed8
pfortuny 5y ago
Even more so: every time IQ and regression to the mean is explained, it is always in the “more-to-less direction”, whereas it works in both and both are relevant.
pjscott 5y ago
> (surely it's logistically impractical to get any large number of busy faculty to spend time taking such tests) You can estimate it decently by looking at SAT scores, which most professors will have taken at some point. Your error bars will be larger than if you administered IQ tests to everyone, but it's a lot easier.

by abeppu 5y ago

mynameishere 5y ago

This kind of quibbling always comes up. He's just using professors and national merit finalists as examples. The truth is, you could probably take the top 10 percent highest-earning car salesmen (EDIT: and women), drop them on a desert island and, after two generations, get the highest-IQ society the world has ever known. But that's less obvious, so why use such an example?

ImaCake 5y ago
But it’s not clear at all that intelligence is a single trait. It is incredibly messy and this means any analysis of it has to be very careful of confounders. In contrast, height is by many definitions a single trait, we know exactly what confounds it’s measurement (and those things are easy to measure!) and we have a pretty good estimate of it’s heritability.
- bordercases 5y ago
  Remapping intelligence to certain cognitive performance parameters, many of which can be measured in a very metrical and objective way, still tends to produce correlations. See Figure 3 of this article: [0] - there is a mix of simple perceptual tasks and more complex reasoning ones, which on the surface appear independent. The paper admits at the end that intelligence isn't one kind of task, but instead many kinds of systems used together. That still doesn't prevent their ability to construct a version of g-factor. And very simple metrics, taken alone, seem to inter-correlate well with IQ scores, like reaction time.[1] This correlation increases with age, suggesting a biological factor in preserving brain integrity that would nullify educational effects of intelligence. If your version of intelligence has nothing to do with cognitive activity I don't know what to tell you. It's true that effectiveness in society might not be correlated with intelligence. That says more about what society rewards and the kind of labor required to get it. And I don't mean to say that society should reward people based off of test scores, that's dumb. Nor should we assortate people based off of pure test scores, also dumb. But I see no real conceptual problems regarding intelligence that doesn't try to conflate intelligence entirely with general effectiveness. There are IQ realists like Steven Hsu that don't do this, placing it on equal importance with conscientiousness [4]. As with height, there are innate and behavioral/developmental components of general effectiveness (I could dope my child with growth hormone to make him taller, or mal-nourish them to go the other way). As with emotional intelligence, being a bastion of full-nurture theorists for behavior that look for non-academic modes of measuring effectiveness, we can get heritability correlates of 0.2-0.4[2], and for psychopathy up to 0.5[3], mediated by biological factors like inherent oxytocin concentrations. It's hard to say you can succeed without without any ability to learn, but there are high paying jobs that involve mainly physical labor (while taking on some mortality risk which can be managed). And there are people who are smart assholes, and other people who are dumb yet compassionate. So I have no problem with accepting the cognitive version of intelligence as with the right attitude there is no anti-social conclusion to risk. Society takes all sorts of people to exist, even if to some extent, the more well-rounded the better. [0] "Fractionating human intelligence", www.owenlab.uwo.ca/pdf/2012%20-%20Hampshire%20-%20Neuron.pdf [1] "The relationship between intelligence and reaction time varies with age", https://www.sciencedirect.com/science/article/pii/S0160289617300636 https://www.sciencedirect.com/science/article/pii/S016028961... [2] https://www.researchgate.net/publication/23302097_A_Behavioral_Genetic_Study_of_Trait_Emotional_Intelligence https://www.researchgate.net/publication/23302097_A_Behavior... [3] https://www.geneticlifehacks.com/psychopath-genes-born-not-made/ https://www.geneticlifehacks.com/psychopath-genes-born-not-m... [4] "Genetics and Intelligence", https://youtube.com/watch?v=62jZENi1ed8 https://youtube.com/watch?v=62jZENi1ed8

pfortuny 5y ago

Even more so: every time IQ and regression to the mean is explained, it is always in the “more-to-less direction”, whereas it works in both and both are relevant.

pjscott 5y ago

> (surely it's logistically impractical to get any large number of busy faculty to spend time taking such tests) You can estimate it decently by looking at SAT scores, which most professors will have taken at some point. Your error bars will be larger than if you administered IQ tests to everyone, but it's a lot easier.