> Things are correlated. No one seems to seriously dispute this. So why all the debates?
I know many very outstanding programmers - but very few are also outstanding at writing poetry, or learning foreign languages, or engaging in impactful political action.
I've no doubt that 'general science' performance is correlated with 'assembling objects' performance. But I wouldn't expect Bobby Fischer to have great insights into world politics just because he was a chess grandmaster.
They wouldn't be outstanding - they don't have specialized training.
But how do your outstanding programmers compare at writing poetry... when compared to your not-outstanding programmers who also have no specialized training?
If they applied themselves to those things the way they did to programming, there's a good chance they'll be outstanding or at least remarkably good in those as well.
Intelligence is a necessary but not sufficient condition for competence at intellectual tasks.
I think this is more because of how people choose to spend their time. I know several outstanding programmers that are also great at learning foreign languages and writing poetry, and in fact, it seems that learning human languages and computer languages have some strong correlations in research about linguistics learning.
But, no matter how good you may be at any one of these things, to become good at the others you must spend time focused on learning and doing them. Most of the great programmers I know prefer to spend their time in pursuing other things that they become good at instead, primarily stuff to do with physical engineering. (Aside, I've always thought the propensity for using physical engineering as a non-work hobby is guided by frustrations from the intangible nature of programming/software.) But, I do in fact know several that travel and learn foreign languages, or that write poetry and short stories as a form of outlet.
article is quite misleading, I can't believe author wasted 5 mins of my life explaining data science and visualisation while all I cared for are answers.
Another way to look at this would be to look at the correlation between scores of different standardized mental tests. ASVAB, IQ, and SAT scores are strongly correlated to one another, even though they use slightly different testing methodologies and focus in some ways on different things.
All of the evidence we've gathered over decades seems to indicate that there is a general factor for intelligence. But as he points out in his conclusion, this is /descriptive/, meaning we can arguably accurately determine someone's intelligence via testing and that single general factor accurately describes their intelligence in relation to others. It does not in any way imply we understand the causative factors that lead that score to go up or down in any meaningful way.
For reasons I don't fully understand, there's a lot of folks in the social sciences these day that seem to be waging war on the very concept that you can describe intelligence generally (and IQ tests specifically). The mere fact that the results of these types of tests on large population groups tend to follow a normal distribution is alone good evidence that they're indicative, as in most things diverse large population groups tend to fall on a normal distribution.
It's because they don't want to deal with the implications and the massive unfairness of the fact that some people are just born smarter than others. Easier to just pretend it's not true.
In general, I have seen a tendency from liberal, humanity disciplines that are trying to get rid of empirical studies to justify their political ideologies: "equality for all vs. equality of outcomes", "truth is relative", etc.
I think it's basically coming down to a battle of the mathematically literate vs. not, and unfortunately the non-literate have a lot of stickiness in academia because they can generate ideas people want to believe with no empirical "where the rubber meets the road".
> For reasons I don't fully understand, there's a lot of folks in the social sciences these day that seem to be waging war on the very concept that you can describe intelligence generally (and IQ tests specifically).
The unfortunate reality is that social sciences are largely corrupted by political activism. Their activities are often prescriptive social engineering that attempt to use their perceived scientific authority rather than actual descriptive science. Put simply, they are saying what they want to be so, rather than what is so.
I've heard the argument that even if there is a reliable measure of IQ and there is a measurable difference between groups of people, that knowledge only harms people and doesn't do any good, so we shouldn't pursue it anyways.
So I think to get everyone on the same page, you have to address that concern in some form, or there will be some people who are always against the idea of measurable intelligence.
I'm not sure that's the same argument. If a chemical engineer wants to explain why she won't research chemical weapon agents, she's not going to say "chemical weapons don't exist" or "everything's made of chemicals so chemical weapons aren't a coherent category". She'll say chemical weapons are a clear, coherent, and real idea, but the risks of understanding them better outweigh any potential benefits.
It's a reasonable argument, and one I personally tend to lean towards, but you can't meaningfully debate it if you don't have a shared understanding that a general factor does indeed exist.
"Corrupted by political activism" is probably going too far. It's not controversial to acknowledge that lots of historical attempts to measure intelligence have been pretty biased in favor of English speaking white men. Many of those attempts were biased just because of unexamined biases (the kind everyone has). Unfortunately some were also very deliberate spins (one notable example being the book The Bell Curve).
I think it's reasonable to conclude that measuring intelligence isn't valuable enough for all of the negative externalities attached to it given how the world currently is. But then from there it's easy to make the easy jump to saying it's impossible to measure intelligence without those negative externalities, which is probably not true. But if you can't measure intelligence without significant bias, you aren't really measuring it at all, you're just measuring your biases.
> It's not controversial to acknowledge that lots of historical attempts to measure intelligence have been pretty biased in favor of English speaking white men.
Do you have some specific examples of this purported bias?
> Unfortunately some were also very deliberate spins (one notable example being the book The Bell Curve).
Specifically what spin are you referring to?
> I think it's reasonable to conclude that measuring intelligence isn't valuable enough for all of the negative externalities attached to it given how the world currently is.
It is because we have a more nuanced take on what intelligence is. We don’t think that intelligence is something that you can measure with a pen and paper test. Being smart also includes stuff like “reading the room”, being empathetic, knowing when to quit, etc.
If there is a general intelligence—which I very much doubt—our current tests don’t even come close to measuring it.
Maybe because they are designed to and because of selection bias. These tests are expected to correlate, if they don’t they are considered a bad test and is thrown into obscurity (but more likely fixed until it does).
But if they didn't even come close to measuring it then there would be no correlation ? It's not like these tests are applied retroactively to explain past performances - they offer actual predictive value. Selection bias can account for some of it but there are ways to test if this and from what I've seen it still holds up.
I’m not disputing that these tests measure something, they obviously do. However what they are measuring is not what most people consider a satisfactory definition of intelligence. In fact most psychometricians simply give up and define intelligence based on what these tests measure (i.e. operation definition). Without justification that is very poor science.
SAT scores would be positively correlated with success even if they were entirely random as long as universities continued to use them as a factor in admissions. It's sort of a corollary of Goodhart's law that it is difficult to assess the quality of a measure while it is being used as a target.
IIRC, the strongest predictive factor for SAT scores is... parental income. So if what these tests are measuring is not a "general factor of intelligence" but "quality of home environment", you can readily see how that could correlate well with outcomes like "success in later life."
SAT mixes in intelligence with education so it's bound to skew more towards people with access, but it's also correlated with IQ, and IQ has significant heritability as well (things like separated twin studies have shown the genetic component quite clearly). So it's probably a double whammy there - good environment and good genetics.
What we can measure though is useful since it correlates to so many things. Controlling for every factor we can think of, parental income, school quality, etc etc, people who have high IQs when they're young have better outcomes as adults on average - in income, in happiness surveys, lower divorce rate, etc.
I think people aren't aware that controlling for every factor is what distills the idea of general intelligence. That even after you account for everything you can think of, there is some quality that allows people to perform better mentally on many different types of tasks.
The existence of 'general intelligence' is a pretty different question from "can we measure 'general intelligence'"? The analysis demonstrated does not account for the possibility of confounding by some external factor unrelated to intelligence (income, tutoring, etc), which is what many will argue. At best you can claim a notion of 'general test-taking ability' with no ambiguity. Once you recognize the problem-solving patterns in these tests, doing well becomes significantly easier, which is an almost inescapable measurement problem.
And re: normal distribution implying some fundamental truth, the same can be said of an iterated guessing game (eg: take the sum of your correct guesses over 100 coin-flips - the population's scores on that game will be normally distributed).
As pointed out by Stephen J. Gould, intelligence is not normally distributed. The IQ test results are. But IQ test results are standardized to be normally distributed. It is an assumption that intelligence is normally distributed, and evidence suggests it is not. More likely—if it exists at all—it is log-normally distributed.
Psychometricians believe that there is some fundimental truth in the normal distributions, and then standardize their test scores so that it follows this distribution in order to gain validity. It is a pseudo-scientific self fulfilling prophesy.
This viewpoint is generally acknowledged as wrong by the scientific community, and Gould is generally acknowledged as a hack who sold out by telling people what they wanted to hear in exchange for book sales.
> Gould is generally acknowledged as a hack who sold out by telling people what they wanted to hear in exchange for book sales.
It’s funny you should say that. Because I generally regard the authors Gould is addressing in his book Mismeasure of Man, including S.G. Morton, C. Murray, A. Jensen, etc. as hacks that used questionable scientific means to tell people what they wanted to hear—namely the superiority of white men—in exchange for book sales.
Stephen J Gould was a scientific fraud, with an ideological axe to grind, and repeatedly lied and misrepresented reality to smear what he saw as his ideological opponents. His most famous work on the topic, The Mismeasure of Man, for example, discussed Morton’s skull measuring experiments, who found that people of various races have different average cranial capacities. This is now a well established scientific fact, nevertheless Gould found this so offensive, that he claimed Morton fudged results by putting finger on the scale, based on no evidence whatsoever, simply disbelieving this sort of human variation between populations. Morton’s skulls have been in fact remeasured, and found to match Morton’s measurements within error margin appropriate to the method. This has been communicated to Gould himself, but he did not issue any correction or indeed mention any of it for the second edition of his book, showing clear evidence for his ideologically driven lack of scientific integrity.
It's great that you found some super specific thing where his wrong answer isn't provably the result of fudging of numbers, but he's still wrong for various reasons. And someone else measured other skulls at roughly same time and came to the correct conclusion that you couldn't really make any such claim based on the data they had.
Morton was wrong on many counts, to be sure, just as any scientist worth of salt must be. However, he was not wrong on the very thing Gould accused him of, which was ideologically based fudging of data, something that Gould was in fact guilty himself.
He did ideologically fudge his data, though, to the point that he could, incorrectly, make the claim that he already believed, but that wasn't actually there in the data.
Gould's point was about inadvertant bias in science and he was correct in this case.
You are doing the very thing you are accusing Gould of doing. Rebutting a claim based on the fact that the person making that claim was wrong in an unrelated claim. This is called an ad-hominen fallacy.
You might be right that Gould was a fraud, however you still have to show that his claim about the fact that intelligence isn’t normally distributed is wrong. In fact the burden of proof is kind of weird here, since assuming a normal distribution is a pretty strong assumption, and rejecting that assumption based on a lack of evidence should be enough. However Gould goes further and cites evidence that intelligence has a pretty skewed distribution.
If you want to reject Gould’s claim here, then do so. Don’t just say he’s a fraud and then assume his wrong about everything.
I'm not sure I understand the objection. I agree that standardizing to a normal distribution might be arbitrary, but how could one argue what the true transform of the distribution is? After all, a log-normally distributed variable is just exp() of a normally distributed one. Would anything be different if instead of IQ, we called it log-IQ?
I don’t think there is any research into finding “the true distribution” of intelligence. But intuitively intelligence should have a pretty skewed distribution with a long tail on the lower end where people with varying levels of mental handicap are skewing it, most people above average and a handful of people slightly more intelligent then the mode. This aligns with observation of non-IQ measures of intelligence like school grade or the SAT. This also works intuitively how we think of IQ. If you have an IQ of 70 there is probably something wrong, if you have an IQ of 130, you are probably just a normal person that might be slightly smarter then me.
The problem is assuming standard normal distribution as the true distribution for intelligence is simply bad science. Whatever the distribution truly is (if there even is one) it is not this one. And then psychometricians (including OP) use this as a point in favor of the validity of IQ as a good measure.
I don't understand what one could mean by the true marginal distribution of a latent quantity that's only defined in terms of its joint distribution with measurements. What difference would it make if we transformed IQ scores to a skewed distribution?
To give an uncontroversial example, we use logarithmic scales for earthquakes precisely because the underlying phenomenon is heavy-tailed. It's just for convenience, it doesn't imply anything about the phenomenon being quantified.
> The mere fact that the results of these types of tests on large population groups tend to follow a normal distribution is alone good evidence that they're indicative, as in most things diverse large population groups tend to fall on a normal distribution.
The claim here is that OP is wrong. Firstly OP is wrong because normal distribution is not a sufficient indicator for some underlying phenomenon. Or as yxwvut puts it:
> [...] normal distribution implying some fundamental truth, the same can be said of an iterated guessing game (eg: take the sum of your correct guesses over 100 coin-flips - the population's scores on that game will be normally distributed).
And then secondly because intelligence doesn’t even follow a normal curve (only IQ does because it is transformed to).
What matter here is not the true distribution of intelligence. I’m not even sure there is one. What matters is that proponents of general intelligence use the normal distribution of IQ as a justification. By arguing that intelligence does not fall on normal distribution but rather on a skewed distribution, I’m simply pulling the rug away from this argument.
I admit this is not a strong argument against the existence of general intelligence, since if there exists such a thing, then there is nothing that says such a thing couldn’t exist on a skewed distribution (except psychometricians and proponents of IQ which only accept standard bell curve [I wonder why]). However in this subthread we’re only trying to correct OP’s claim that a normal distribution is indicative for general intelligence.
Thanks for explaining. It sounds like we agree that "OP is wrong because normal distribution is not a sufficient indicator for some underlying phenomenon". And that we're not sure that there's a "true distribution" of intelligence. But I don't know how you can then argue that "intelligence does not fall on normal distribution but rather on a skewed distribution". Again, how could we choose between calling skewed thing vs its log transform "intelligence"?
I actually don’t believe in general intelligence, and I don’t believe intelligence is this one thing that you can measure. So I see how it can be confusing when I go ahead and say: “Intelligence has a skewed distribution.”
But when I’m talking about “intelligence” in this context I’m using working within the assumptions of the psychometric. In other words, I forgo my priors and argue on the terms of my opponents. I.e. given that they are right about the existence of general intelligence and that IQ tests are a good and accurate way to measure intelligence, they are still not correct about it being normally distributed.
To be precise. When I use the term “intelligence” in this context I’m basically using the same operationally defined term as psychometricians. That is: Intelligence is whatever the IQ tests are measuring.
Now as for which distribution to pick for this thing that IQ tests are measuring. Why pick a skewed distribution instead of a normal one? Psychometricians pick the bell curve because they believe it gives this term some validity. I would pick a skewed distribution because that fits our intuition and fits the underlying data. I would call the former justification pseudo-science as you have no reason other then your prior assumptions to do that. Then why do the latter? I would say the only reason to do that is to prove the people doing the former wrong.
Thanks for elaborating. But I still don't understand in what sense a skewed distribution "fits the underlying data". I presume that for any given task, we would have to write down a model or likelihood that relates intelligence to the outcome of the task. No matter what transformation of intelligence we use when discussing population statistics, as long as it's monotonic then we can always undo that transformation when we write down a model.
Again, regarding earthquakes: Does a logarithmic scale "fit the underlying data better"? I don't think the question even makes sense. It's like asking which font fits the underlying data better. It's an arbitrary choice that doesn't need to affect our predictions or models at all. But you also said that "the only reason to do that is to prove the people doing the former wrong." so maybe we agree on this, too.
I think you might be misunderstanding psychometrics if your objection to IQ is the supposed implication that "intelligence is this one thing that you can measure". As far as I understand, psychometricians would agree that intelligence has many facets, and that most of them are hard to measure. I think the claim that they're making is simply that one can summarize estimates of these facets into a single scalar that will be correlated with all of them. Just like GDP correlates with many hard-to-measure and complicated facets of a country's wealth. Would you say the fact that GDP can't measure some important things about a country's wealth, or that it's a complicated thing, implies that there's no such thing as "general wealth"? The same analogy works for e.g. health, strength, or pretty much any property that we have a word for.
The way I understand psychometrics—in a simplified matter—is as a misunderstanding of this philosophical statement: “If it can’t be measured, it doesn’t exists”. I see it as a bastardization of logical positivism in many ways, including by totally misusing operational definition. Resulting in the believe that: “We measured it, it must exist!”.
Measuring wealth is an excellent example. GDP might correlated with other measures of wealth, but it hardly explains the whole story. It can be useful in models but you can’t summarize nation’s wealth by only looking at GDP. Similarly IQ can be a useful tool. IQ tests can be used to find people with mental disabilities so that appropriate accommodations can be provided (although most of the time, there are easier, cheaper and better tools to diagnose mental disabilities). However by looking at the IQ alone (especially around and above the mean) you really can’t say much about about the mental capabilities of a person, only correlations of varying usefulness..
There is an important distinction though. I would argue that GDP (arguably) proves it’s usefulness in different economic models. However the same cannot be extended to IQs. I don’t know of many psychological models which use IQ (or any notion of general intelligence for that matter) to explain behavior (other then a hacks trying to prove the superiority of the white race, but let’s ignore those, what they are doing, it is not science).
It sounds like you're just saying that IQ is overhyped. You agree that IQ can be a useful tool. And everyone agrees that having more data than just IQ lets you make better predictions.
How is IQ not "explaining behavior" a problem? We also wouldn't say that GDP "explains the economy", but that's not necessary for it to be useful. Both of these statistics are relatively strongly correlated with many outcomes we care about, compared to other things that are about as easy to measure. I think that's the main reason people use them, and it's a good one.
It sounds like you've seen people over-use aggregate statistics as a way to downplay nuance. That must be annoying. But that doesn't make their use philosophically or scientifically invalid.
> It sounds like you're just saying that IQ is overhyped
Yes. That pretty much summarizes it. If IQ doesn’t further our knowledge of human behavior nor the human mind it is at best a useless construct. We have other more targeted measures (such as SAT) which are easier and cheaper to administer, and are far more useful in both models and evaluations/diagnostics. GDP has the benefit here of being extremely easy to measure and is therefor useful in more general models. IQ doesn’t have this quality, so that is not why most people use it.
Now if that was the only problem with IQ I wouldn’t be so bothered by it. Scientist are allowed to use an inefficient means of measuring their constructs. However IQ advocates often have a vested interest in an IQ that is immutable and (to some degree) heritable. This is often assumed to be true and IQ tests are later fixed in order to make them more robust in score against subsequent trails. This—off course—is circular reasoning (or a self fulfilling prophecy) and is not good science. But it is still used to explain group differences. This is where using IQ becomes scientifically invalid.
Now without the claim of immutability and heritability this wouldn’t matter. We would see a group difference and then we could then go on and try to explain it (e.g. within the context of sociology). However with immutability and heritability, you don’t have to explain it further. The answer is in psychology and genetics. (This is by the way the basis for scientific racism mentioned before).
> If IQ doesn’t further our knowledge of human behavior nor the human mind it is at best a useless construct.
I mean, I think it does further our knowledge of these things, but even if it didn't, that wouldn't make it useless, because it still predicts things we care about pretty well.
> GDP has the benefit here of being extremely easy to measure... IQ doesn’t have this quality
Which measures of intelligence do you think are easier to measure than IQ? You mention the SAT positively below, which is very similar to an IQ test.
> IQ advocates often have a vested interest in an IQ that is immutable and (to some degree) heritable. This is often assumed to be true...
I mean, every researcher would say that it's only immutable and heritable to a degree, and there's tons of research quantifying the degree of mutability and heritability (e.g. twin studies, admixture studies).
> But it is still used to explain group differences. This is where using IQ becomes scientifically invalid.
Is there a specific claim you have in mind that you consider invalid? Can you give me an example of an explanation of group differences that you would consider scientifically valid?
> with immutability and heritability, you don’t have to explain it further. The answer is in psychology and genetics.
Are you saying these claims happen to be false, or that they're categorically impossible / bad explanations? Is there any aspect of human behavior that you think can be explained by psychology or genetics?
> Is there a specific claim you have in mind that you consider invalid? Can you give me an example of an explanation of group differences that you would consider scientifically valid?
Yes. The Bell Curve by Richard J. Herrnstein, Charles Murray (1994). I’m currently having a debate about it in another thread. It uses these assumptions about IQ and the demonstrates group difference with flawed studies, which it then goes on to justify conservative policies.
An example of valid—albeit trivial—group difference on the is height difference across genders. A more interesting group difference is how a mock jury is less likely to convict a minority group if a member of that minority is present in the jury. Then of course the field of medicine has many examples of interesting group difference based on genetics. However as a species we humans are remarkably homogeneous so if you find a group difference in behavior based on genetics it will most likely have an extremely small effect size, and as such be of limited interest. At best this is where IQ research falls in.
> There's tons of research quantifying the degree of mutability and heritability (e.g. twin studies, admixture studies).
These are still up for debate and not without bias. The original twin studies by C. Burt were likely falsified (https://en.wikipedia.org/wiki/Cyril_Burt#%22The_Burt_Affair%...) later efforts of reproducing the results are not without controversies either. And even if you look beyond these controversies there is an inherent bias as adopted twins are way more likely to be adopted within the same socio-economic group, adjusting for that is really tough. I’m not aware of any admixture study which explains a significant group difference in behavior, so I can’t comment on this. But I’ll leave this quote from Wikipedia instead (https://en.wikipedia.org/wiki/Race_and_intelligence#Racial_a...):
> Reviewing the evidence from admixture studies Hunt (2010) considers it to be inconclusive because of too many uncontrolled variables. Mackintosh (2011, p. 338) quotes a statement by Nisbett (2009) to the effect that admixture studies have not provided a shred of evidence in favor of a genetic basis for the IQ gap.
Virtually everything even slightly abstracted from directly quantifiable tasks: sporting skill, driving skill, intelligence, etc. None of those are measurable without significant confounding or a 'dumbing down' to some adjacent task (that still probably suffers from some degree of confounding via 'teaching to the test'). Sport performance depends on your cohort, driving accident rates depend on your environment, intelligence tests have a learned component, etc. Trying to develop a system that allows rank-ordering of ability along an axis that isn't subject to influence from an irrelevant external factor is impossible.
There is some evidence that bodily-kinesthetic and musical ability does not correlate that well will g as other things, but they're still very correlated.
See for instance:
Castejon, J. L., Perez, A. M., & Gilar, R. (2010). Confirmatory factor analysis of Project Spectrum activities. A second-order g factor or multiple intelligences?. Intelligence, 38(5), 481-496.
Visser, B. A., Ashton, M. C., & Vernon, P. A. (2006). Beyond g: Putting multiple intelligences theory to the test. Intelligence, 34(5), 487-502.
Looking just at SATs, it is very common for people to take a practice test early on, then do lots of work/ revision, then eventually recieve a much higher score.
If SAT score were measuring some immutable value, then it wouldn't be so easy to improve one's score with revision, and tutors.
The degree to which the SAT is coachable is a subject of debate. The biggest studies have suggested that even private tutoring has an effect of less than 40 points.
Kasparov had a nice line about this in "How life imitates chess". I'm paraphrasing but it was something along the lines of "there are lots of things that cannot be tought to you which you can nontheless teach yourself".
I think that thinking processes is very much in that category, and I think that the investment in time is considerable. I think its typical to expect gradual improvement -- a constant response to training. Anecdotally, it doesn't work like that.
I'm not really sure anyone above has implied that intelligence is immutable - just that you can describe it generally.
Consider the physical fitness correlations in the article - I don't think anyone would argue that your level of physical fitness is immutable. But just because the value is mutable is no reason to disregard the current value - it can still be a meaningful and useful indicator.
That's part of why the r for SAT-IQ is somewhere between 0.5-0.8[0] rather than higher. Also the average SAT score increase by retaking seems to be 40[1] points which doesn't skew things that much.
Well I think there is a separate question of "does this work on a population level" versus "does this work for any given individual". I think we are used to "vaccine mostly works for everyone" type thinking. However, I think that there are many things that will work for individuals but not for general populations.
So SATs are not generally trainable by a universal approach -- including personal tuition. I think that is believable. However, that for any given individual there is no intervention which will substantially improve their SAT results is a much stronger statement which I find less believable.
IQ doesn't behave like you think it might in fairly uncontroversial situations is part of the problem. For example, have a read about the "Rationality Quotient" and the surprising conclusion that rationality is poorly correlated with IQ. I'd certainly expect smarter people to be more rational, but it is not so.
I think that the treatment of things like time factors and tool use is also somewhat trivialised. Those that play chess will note that excellent Rapid or standard players can be terrible Bullet or Blitz players.
Roger Penrose -- nobel prize winning mathematical physicist -- was held back a year on count of his arithmetic being that bad. Yet he is a shockingly good mathematician mostly qua geometrical intuition to which he can convert other things. Who cares, Penrose can just use a calculator. Similarly, I'd say that many IQ tests are just symbolic regression. What if I was brilliant at stating symbolic regression problems, but poor at solving them? Then me + solver would make all the difference, unlike if I was poor at stating them ...
Showing outliers just proves there are outliers not that the averages are uselesss.
>Those that play chess will note that excellent Rapid or standard players can be terrible Bullet or Blitz players.
I play chess and those are definitely highly correlated, I suspect at least r of 0.7 especially for those who have put decent time into the different controls.
I didn't mention Penrose because he is an outlier. I mention him as an example where tool use disproportionally effects resultant performance, or conversely as an example of where a specific link in a reasoning chain effects the outcome. I think that is more close to normal than not. I.e. you can often patch people, and they become better at something.
> For reasons I don't fully understand, there's a lot of folks in the social sciences these day that seem to be waging war on the very concept that you can describe intelligence generally (and IQ tests specifically).
I think part of the issue is that it's very hard to measure intelligence accurately (whatever that means) across cultural divides. If you gave an American IQ test (e.g. an SAT) to Europeans, Asians, and Africans you'll likely get different results, not (necessarily) because of innate differences but because of educational/cultural backgrounds. Poverty levels can also change scores.
This makes it tricky to know what to do with these scores and we need to acknowledge that. This leads some to conclude that we should get rid of the notion of a measurable IQ altogether. That goes too far in my opinion but I understand their concern.
Unclear if it is hard to measure intelligence or if we just don’t like the results.
Of course poverty and education can change the results. This is only a problem with the tests if we know for a fact intelligence is independent of material circumstances. Nobody would be surprised that good nutrition and regular exercise make it possible for someone to run faster or lift heavier weights.
Of course we as a society are also failing to provide the basics to large segments of the population, but getting rid of testing is a motivated effort to hide failures rather than try to improve. “Can’t get a fever if you don’t check temperatures…”
Some of the difference in scores will be driven by differences in g, some by differences in motivation, some by education, etc. There are tests that, controlling for motivation, do a pretty good job of getting around this (like Raven's progressive matrices, etc.) They don't even require language.
I think you're on the right track, though. But the key sticking point is this: many people are very uncomfortable with the idea that there are differences in intelligence among people, both within and between populations. Unfortunately, that is undeniably true. And so we get this ridiculous, unproductive dance.
Also, just as an aside, the latest evidence around the "poverty actually makes you literally stupider" claim is that it's not supported--studies aren't replicating. Which makes sense because the claim is prima facie crazy, but I digress...
>For reasons I don't fully understand, there's a lot of folks in the social sciences these day that seem to be waging war on the very concept that you can describe intelligence generally (and IQ tests specifically).
The reason is that general intelligence tests show differences between racial and ethnic groups. Which isn't shocking, as intelligence is certainly driven by genetics and compounded by culture. But because it is, it upsets people.
IQ testing is a pretty valuable filtering mechanism. Smart people are really important to advancing society. Replacing it with some other, more subjective metric just moves the filter into the realm of political control.
Which seems like a bad idea to me, but what do I know? I'm not a genderqueer left-handed ginger aboriginal, so I'm not very smart.
Like I said, it upsets people, and since we value not upsetting people, that's the metric we optimize for. This has costs and consequences, like anything else.
It's the social sciences version of technical debt.
> Eschew flamebait. Avoid unrelated controversies and generic tangents.
I’m a prime target for this flamebate, and you’ve managed to upset me quite a bit. I’m actually surprised I didn’t bite already. Indeed I’ll actually just leave it at that and pat my self on the back for resisting.
I think that the "whole story" matters, even if the bits you highlight are a large part of it. A bit like the relationship between epigenetics and the rest of genetics -- it rather changes how you interpret what you already know.
Given that we cannot say what intelligence is better than "its what intelligent people do", I'd say we were far from the "whole story".
I think pretending that there is no "whole story" and using IQ to make policy decisions about populations is political. I don't think anyone cared that a bunch of researchers were tinkering with some psychometrics. They cared that a whole racial group could be shut out of certain jobs on the basis of a half baked abstract reasoning test.
>Given that we cannot say what intelligence is better than "its what intelligent people do", I'd say we were far from the "whole story".
I'm not opposed to this, and I doubt most people would be either. Take college admissions as an example. A low-SAT, low-GPA student in a one-on-one interview with an admissions officer could dazzle the interviewer to such a level that admission with a scholarship could be guaranteed.
We don't do that often for a variety of reasons, but one would certainly be because of the quasi-legal requirement to tick certain boxes to avoid trouble. Okay, but that has costs. We've been able to paper over a lot of this with economic good times, but that's unlikely to last forever. Hard reality always wins in the end.
>The mere fact that the results of these types of tests on large population groups tend to follow a normal distribution.
The results do not follow a normal distribution though. IQ is forced into a normal distribution with a standard deviation of 15 by definition, it is constructed in such a way that it's forced into that distribution. The actual measurements, ie. raw test scores, are not normally distributed.
I think you're conflating IQ and all the other things parent talks about. Almost by definition with these sorts of processes we'll see Gaussian curves.
The raw test scores are normally distributed. This much is pretty much forced by the central limit theorem. Now, of course, the raw score distribution does not have mean of 100 and standard deviation of 15, but that’s just a norming issue, to make it useful to compare across the tests.
The fact that various test scores are correlated to each other only means, could mean that the tests themselves are correlated or that test-taking abilities are correlated. The resulting normal distribution also suggest tests are indicative of test-taking abilities. But... how do you show that test-taking ability correlates to intelligence?
The test taking ability is highly correlated to real world ability on practical tasks. See, for example, Hunter and Schmidt, “The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings.” (1998), which summarizes a mountain of research over almost a century, and in fact finds that test taking ability is strongly predictive of real world performance.
> ASVAB, IQ, and SAT scores are strongly correlated to one another, even though they use slightly different testing methodologies and focus in some ways on different things.
It's also possible that single cultural factors could be so pervasive that they overwhelmingly swamp any signal by forcing a correlation just on its own.
eg: Before the 1960s or so, you could look at every society that had ever existed and make the observation that, except for a countable list of exceptions, men received more education that women. This might lead you to a natural conclusion that men have some natural innate advantage in receiving education that women lack.
Except that's not the truth at all, the truth is that patriarchy was such a universal of the human condition that it affected every single corner of the planet to such an extent that there was no control group until very, very recently. So much so that even the slightest letting off of patriarchy shows women are so much more innately suited to education that colleges have had to quietly implement affirmative action programs for men simply to keep their gender balances even remotely in check.
Many people would argue that classism and racism so thoroughly pervades our understanding of intelligence that we will simply never get our control group until some future date when they recede enough for us to truly see their effects.
>For reasons I don't fully understand, there's a lot of folks in the social sciences these day that seem to be waging war on the very concept that you can describe intelligence generally
The fallout from eugenics gave social science a complex. Innate difference and especially innate superiority of one person over another is taboo. The field is poisoned by motivated reasoning.
>At the height of the controversy, during the 1970s to 1980s, the debate was highly ideologised. In Not in Our Genes: Biology, Ideology and Human Nature (1984), Richard Lewontin, Steven Rose and Leon Kamin criticise "genetic determinism" from a Marxist framework, arguing that "Science is the ultimate legitimator of bourgeois ideology ... If biological determinism is a weapon in the struggle between classes, then the universities are weapons factories, and their teaching and research faculties are the engineers, designers, and production workers." The debate thus shifted away from whether heritable traits exist to whether it was politically or ethically permissible to admit their existence.
Interesting article but it neglects a very important factor in its comparison of physical test results and mental test results, namely the extent to which physical and mental capabilities are trainable. It's a simple exercise to, just take the lower half of the cohort in each categorty (mental/physical) in a given study and give them intensive one-on-one training in all areas for a year, then compare their performance on said mental and physical tasks after that training. Doing this for different age groups might be particularly interesting.
There's often a bias towards assuming that physical skills are trainable while mental skills are not, because many people have been raised to assume that the latter are due to mainly to genetic factors. This 'genetic superiority' is typically then used to justify current class structures, racial makeup of the wealthiest 1% etc. Such views have little if any supporting scientific evidence - mental skills are just as trainable as physical skills (yes there's a genetic component in both cases, which might place an upper bound on one's performance, but such genetic factors are not correlated with race/gender/etc.).
Perhaps the best physical test for innate athletic ability is the standing vertical jump. There isn't an awful lot you can do to improve it. You can lose weight and train explosiveness, but someone with a 10" vertical will never be even average. Compare that to deadlifts, where any individual can triple their max through training.
I'm guessing there is an equivalent with mental capabilities. Somethings you can train, some are innate.
Something like the SAT is a game that people learn to win.
This is oft claimed, but data doesn't back it up. SAT is marginally affected by coaching and training.
"Coached students are only slightly more likely to have large score gains than uncoached students. In addition, about 1/3 of students experience no score gain or score loss following coaching. The typical gain associated with coaching is 8 points for verbal and 18 points for math"
You can easily “win” an IQ test: just memorize the answers (ignoring the fact that not everyone, in fact, can easily memorize a lot of random stuff like that). Of course, no reasonable person would say that a person who memorized all answers in IQ test to get a score of 150 has really improved her mental ability.
The situation is somewhat the same with SAT, but it’s not so bad in actual reality: there is no single SAT test, there are multiple instantiatons, and you can’t just memorize all answers for all SAT tests. At the same time, the actual problems on the test come from a rather limited pool of patterns, and if one learns to tackle each pattern, one can really improve their SAT score. This is, in fact, perfectly commonplace: everyone significantly improves their SAT score between grades 8 and 12 (at least among people who do take SAT at both ages).
Does it make SAT useless, as everyone learn all the patterns and gets perfect score? Empirically, no: there is large variation in actual obtained scores, and the results are mostly in expected directions: “better”, “smarter” students get better scores and vice versa.
Why doesn’t everyone get perfect score? Quite simply, because variation in mental ability also makes it so that learning the pattern is easier to some people than it is to others. This is by far the source of most of the variation: there is some residual component of conscientiousness, but it’s much less important: spending more time cramming has lower returns the less able you are, and for everyone, there are significant diminishing returns to extra hour of cramming, and ultimately high scorers don’t differ that much in terms of how much time they spend prepping to SAT compared to mid-scorers.
You ever watch the NFL scouting Combine? It's a stat every recruiter looks at for most sports. As a competitive volleyball player I (and my coach) obsessed over a 5% increase that came from years of training.
The evidence that you can train for the SAT is pretty scarce.
- Some people are so disinterested in standardized tests that they never did a practice test or anything like it (this is me!)
- These people get a pretty significant bump from taking the test a second time, from sheer familiarity effects (I didn't bother, but maybe I should have!)
- It is unclear that extensive SAT tutoring does anything for anyone more than "force you to look at an SAT practice test at least once, and alert you to the fact that you are allowed to take the test many times, thus giving you a chance to choose the max of a naturally varying set rather than a single measurement"
>Perhaps the best physical test for innate athletic ability is the standing vertical jump. There isn't an awful lot you can do to improve it.
This isn't a very good example, because the reason it gets harder is that increasing the body's muscle mass also increases the strength needed to propel it a certain distance. Hence gaining muscle mass both helps and hurts the vertical jump. Is there any reason to believe intelligence should be like this?
No, it's not a measure of strength. It's a measure of explosiveness, or rate of force production. They are actually very different physical traits. Strength can be trained to a large extent. Explosiveness is mainly genetics. You can improve it by 15 or 20%. That's it.
"bias towards assuming that physical skills are trainable while mental skills are not,"
For the layman, probably. Actual IQ tests try to ensure that subjects don't have experience with the test tasks. If they do, then the test is not as effective at measuring innate ability to learn or problem solve since it has been trained and the strategy to solve is already known. So it shouldn't be a misperception in the fields that deal with it.
We do this experiment all the time. It’s called “education”. It’s results are somewhat disappointing: people who do well in grades 1-3 also mostly do well in grades 13-16, and the converse is also true. Abilities are trainable, but trainability is also varying across people, and we don’t know how to train people to be better at learning to any substantial degree.
Anyone who has actual and nontrivial experience in teaching (say, over 50 students over at least few months) will tell you this simple empirical fact: some people just “get it” much easier and quicker than others. The only way to escape this reality is to sort people by mental ability, so that you only see people in the same ability bracket, but that’s just it, an escape.
We collectively have enormous amount of experience in educational interventions, and, in short, none of it really works. It’s easy to complain about lack of resources, about low teacher quality, about wrong methods, and all of these are in fact real problems, and pretty complex and hard to tackle too. At the same time, even if we pull each of those to the state of the art, we can only expect a marginal improvement.
Have you noticed how the educational establishments seem to try a new thing every 5-10 years or so? That’s because none of it really pushes the needle, and these 5-10 years are what it takes to realize that. In reality, the “new thing” is in fact usually an old thing that has already been tried a generation or two ago, but which has not been seen by the current generation of education workers. That’s because everything has already been tried, and none of it really works.
On a more optimistic side, that means that even if the education system is suboptimal, it doesn’t really matter, as outcomes are more or less the same anyway.
>We do this experiment all the time. It’s called “education”.
The poster you're replying to said "just take the lower half of the cohort in each categorty (mental/physical) in a given study and give them intensive one-on-one training in all areas for a year". Pretty much every study has show that one-to-one tutoring is far, far more effective than the standard "education" system.
>We collectively have enormous amount of experience in educational interventions, and, in short, none of it really works.
What are you talking about? There's plenty of evidence that one-to-one tutoring can be quite effective at increasing test scores. Yes this approach can't be deployed at scale due to the sheer cost of it, but that's completely irrelevant to the question of whether intelligence is trainable.
No, one on one tutoring is only marginally more effective than standard education system. As I said, the education establishment has expended enormous amount of effort on figuring out how to teach better, and collected mountain of evidence. You won’t find a lot of d > 1 outcomes in literature, and not many more of d > 0.5. Many of those you find did not replicate, and you should expect at least some bogus outcomes because of how p-values work. Look at federal Head Start assessment, for example: gains are minimal, and they dissipate over time. This is the pattern that repeats over and over in educational studies: either gains are minimal and dissipate, or the study fails to replicate.
> No, one on one tutoring is only marginally more effective than standard education system.
Two sigma https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem is not a "marginal" effect. And a chunk of that can be explained by mastery learning, which while still hard to scale up is far easier than individual tutoring as a whole. The educational system is not institutionally incented to effectively pursue these empirically workable approaches. It's a matter of bureaucracy-driven failure, not some immutable fact about the world.
There's even specifically developed approaches to effectively train and educate those who might be often described as 'low I.Q. individuals', such as "direct instruction". Guess what, they get no takeup from the educational establishment, simply because they're (quite wrongly!) perceived as being demeaning for teachers.
The average teacher would rather pretend to educate those who are already highly functionally literate and familiar with educational dynamics (including test taking, of course) and can simply "learn by themselves". Education schools consistently cater to these attitudes and push the same approach.
> "Anyone who has actual and nontrivial experience in teaching (say, over 50 students over at least few months) will tell you this simple empirical fact: some people just “get it” much easier and quicker than others. The only way to escape this reality is to sort people by mental ability, so that you only see people in the same ability bracket, but that’s just it, an escape."
Consider learning how to throw a ball. Do some new students 'just get it' - or did their parents spend hours tossing a ball with them previously? Point being, those anecdotes about students who just get it don't factor in variables like earlier childhood education etc. Few people would argue that such physical skills are not eminently trainable - and clearly, mental skills are also trainable.
The question is, is there something like a clearly delineated genetic factor than makes it easier for someone to, for example, learn a new language, relative to someone lacking that factor? No real clear evidence seems to exist for that claim.
I actually don't know of any studies of the nature I suggested. Note such a study would be fairly expensive as the cost of standardized high-quality one-on-one tutoring (with both mental and physical coaches) over a period of one year for the study cohort would be rather high, so such studies may simply not exist.
It's not soapboxing; in particular it would be interesting to see if the 'ability to learn and improve' changed with age. It's commonly assumed that it's harder to learn new things as we age, but I've never seen that data quantified personally... Hopefully it's less true than we imagine it to be.
There’s nothing “all-inclusive” - but we have really robust research for each piece of this in a vacuum, because frankly it would be incredibly difficult to test the generalizability of any single tutoring methodology when applied long-term.
Plenty of exercise physiology research on coaching format (personal trainer vs. group classes), and strength training adaptations by age that shouldn’t be hard to find at all to cover the physical side.
There are physical skills with genetic factors correlated with race or rather with groups of people with shared genetics.
Here is a short and incomplete list: ability to handle sun or lack thereof, ability to handle heat or lack thereof, ability to live in high altitudes, ability to process lactose, ability to tolerate alcohol, resistance to malaria, height.
There are also diseases correlated with groups of genetically related people.
However, these groups most often don't fit so cleanly into the US's constructs of race.
There could be mental skills correlated with groups of people with shared genetics.
The question is how we handle that if that is the case. One approach is to deny it. Another is to force or encourage people to have children with others to smudge out differences. Another is to let it be.
I suspect that the unspoken answer is denial while encouraging the smudging out
of visible differences. I suspect that people are morally against the last and that it would lead to massive conflict.
I think a similar calculus has happened many times before. I think that is partially the origin of "whiteness".
This is the focus of the highly-controversial 1990s book “The Bell Curve”. If you want to follow the academic history of the topic, I highly recommend reading it along with responses to it by critics.
The problem with this approach is that the correlations only work for average and below average cases. If someone is illiterate, chances are he's not going to be great at physics or medicine either. If someone is overweight, chances are he's not very good at running or ice skating.
Correlations stop working as we move above average, though. Paul Morphy was good at chess, but not so good as a lawyer. Michael Jordan is great at basketball, but not at golf.
Indeed. And that what these tests were all originally designed to do before they were coopted by the eugenics movement. IQ tests are really good at detecting if someone needs more help. They are not good at detecting if someone is smarter then another.
I'm not sure that's anything more than an encoding issue. The apparent bifurcation at "above average" is probably reducible to "above average in at least one technical domain" with an additional encoding for "above average in more than one technical domain". Your concern is resolved by the first, with information loss of polyglots captured by the second.
There are 450 NBA players and 250 golfers on the PGA tour. So if you take this out of a sample of just men in the USA (massive underestimate), then to be a NBA player is about 3 / million and pro golfer is about 2 / million. So that means if there was no correlation, to be both would be 6 / trillion. So if they were correlated such that the combination is 100x more likely than average, the odds would be 0.6 / billion, or 2 men on Earth. But, this doesn't take into account the tradeoff in training time between the 2 sports. Realistically, that would be zero people. So one would expect even with a strong correlation, nobody would be able to be a pro athlete in multiple sports. But wait, some people have! https://en.wikipedia.org/wiki/Deion_Sanders, and your example of Michael Jordan was able to play in the minor leagues of MLB, something that most people could never come close to if they trained for it their whole life. So, yes, there is absolutely a general athleticism.
edit: so would be higher numbers because would have to take into account number of people who have been pro athletes at some point in their life, not just right now. But not enough higher that the point doesn't stand.
Also, Michael Jordan's golf handicap was 1.3, which is not PGA level, but extremely good (top 3% of all gofers). On any given day on a random golf course, good chance he'd be the best golfer there.
The argument seems to rely on measurement data. That is, all the intelligence measures we have are correlated, therefore there is a single intelligence factor. What this argument misses is the fact that not observing something doesn’t mean that it doesn’t exist. For example, we weren’t able to observe atoms for a long time time but they were there all along. Gardener tried to make this argument in the 70s regarding multiple intelligences but I don’t know about the measurement evidence backing that theory up.
On one hand, g is such a flamebait topic that’s been covered to death. As mentioned in the piece, there’s a hard scientific consensus, even if people don’t bother to look into it before spouting their hot takes.
But that being said, the visualization on this are top notch, and the effort put into them shows - fantastic article in terms of presentation, and hats off to the author for putting it together.
As each axis measures something different, there is some sense of unit dependency (e.g. measuring height in one axis (cm) and measuring weight in another axis (kg)). It is silly to say one number in one axis with a different unit is comparable to another one in another axis.
A different way to say it is that each axis can be (should be) normalized before any kind of factor analysis. The cigar shape is as much a function of per-axis normalization than whatever the underlying pattern is.
One problem is that g is not very well defined. It's just some vague factor.
I think it would make sense to define it as something like: the extent someone has well calibrated Bayesian reasoning that exploits information to its maximum to do accurate predictions.
It's hard to test though because there is a dearth of synthetic benchmarks or algorithms to compare to. Humans seem to still be better at performing these types of Bayesian "common sense" tasks than any synthetic procedure, so we can only compare using tests that we subjectively agree other "smart" humans seem to be good at.
> the extent someone has well calibrated Bayesian reasoning that exploits information to its maximum to do accurate predictions.
The concept of g actually completely rules this interpretation out based on nothing. If g was a matter of learning reasoning skills and applying them, you'd be able to teach it. It wouldn't be some magical race/sex/class number.
Using calibrated Bayesian reasoning can certainly be an innate trait. Some people seem good at making good predictions and assessing uncertainty of the data and knowledge they have whether they've been thought or not and it seems like an unusually difficult thing to teach. Will you make that much better predictions even if you read all of the lesswrong.com cannon?
I dont think your definition is accurate though. IQ tests like progressive matrices dont test only if you have good probabilistic models but how well you can identify the pattern in the matrices. If the patterns follow a minimum entropy distribution, there is only so much that falls to technique, and the technique is intuitive (prefer simple over complex); there is still the problem of raw computation that is guessing and rejecting.
> Suppose you’re hiring someone for a job that requires a few different mental tasks. (Arithmetic, sequential memory, whatever.) If you knew someone’s g, you could guess how well they’d do at each task. But it would only be a guess! To really know, you still need to test the skills individually.
I think this part in the conclusion somewhat contradicts a lot of the analysis. In fact, the results have shown that if you knew someone's g, your guess at their performance on each task would be a good guess.
Yeah, it seems like political correctness to me. The author obviously understands your point, and what the author writes isn't exactly false--but it is misleading enough to keep some likely screamers off his/her back.
> This question is a trap. If you try to answer it, you’ll find yourself beset by semantic questions. What’s intelligence? What’s a factor? And if you get past those, you’ll then find a bleak valley of statistical arcana.
Thing is, building the ontology is the most important part of the scientific process. Otherwise you're just praying to a blind idiot god and listen into the void for answers. Or as the author puts it, "statistical arcana."
To be concrete, I'm pretty confident I overtook half of humanity in that fitness test in the last year. Not because I was bitten by a radioactive spider, but because I started to work out. Now to claim the analogy to intelligence, that would need an answer to what the analogy of working out is in intelligence.
Or to look at it on the level of people who compete in a sport, I'm quite confident that Tom Brady is not very fit by NFL player standards. Similar for Xavi in soccer, but Christiano Ronaldo is probably near the top, and I don't have a good intuition in case of Messi, because he is a quite small player and that will have an effect.
The preceding criticism is enabled, by having a good intuition of what the fitness test is, I have a pretty good intuition about the mechanics of a push up. That is not there for intelligence, and I strongly suspect that this whishi-washiness is whats actually holding up all arguments about iq, since it prevents concrete criticism.
I think that might be on purpose. Proponents of general IQ really want IQ to be immutable (i.e. mostly hereditary) and really don’t want it to be diverse and nuanced (hence general intelligence). I.e. they don’t want IQ test to resemble fitness tests. So they try their best to design their tests in a way that it stays consistent across time and exercise does not affect test scores that much. A test that actually measures a specific thing and increases with exercise (e.g. the SAT) is discounted as a factor.
In fact the foundations of IQ tests lies heavily on the eugenics movement which wanted to prove the superiority of able bodied white men. They actually started by assuming brain size correlates with intelligence (it doesn’t) and then set out and proved that on average white men had the biggest brain of them all (which is also false). The nuances of fitness does not help in this goals as one can easily just pick a type of intelligence/fitness which white men perform poorly at in general so their claim was that there exists a general intelligence to remove this nuance. Their current IQ tests are all set up to assume this is true, and hence you can see it in the results. If you can’t then you are not really measuring intelligence.
Yeah some racists thought black people are stupid because of brain size. Has less than nothing to do with modern IQ tests. I’m not following your argument that the racist behavior of people in the general study of intelligence invalidate all study of it forever.
I drew the parallels because we know the brain size studies were pseudo science with an agenda to prove a ludicrous statement. IQs are still up for debate, but we do know that modern IQ tests have their foundations in the eugenics movement. And they certainly had the same agenda as the brain size hacks, so I don’t think the parallels are unfair. There has been a high profile literature as early as the 1990s (The Bell Curve) which tries to use the IQ to prove this nonsense. So even if the brain size hacks are all gone, and the eugenics movement died with WW2, there are still people with an agenda where a nuanced free immutable “intelligence” is their goal.
Saying the “foundations” isn’t an argument. Many geneticists used to be hard core eugenicists doesn’t invalidate the whole field or it’s developments. The validity of modern iq exams has been studied by hundreds or thousands of modern non racist psychologists and the evidence is public info
The evidence is still disputed, and the validity is encoded in the tests. Modern IQ tests are adjusted to be robust across multiple trials, that doesn’t prove that intelligence is immutable by a long shot, it is merely a self fulfilling prophecy.
Books such as The Bell Curve are not in and of them self racist. This book in particular makes zero racist claims. However it relies on deeply racist studies as its premise, so if you want to conclude anything from it, that conclusion is racist.
I don’t think every proponent of IQ is racist, however there is a lot of racism embedded in the research, sometimes directly, but most often by association.
It seems like most of the skills measured in the studies referenced relate to spacial skills or memory/recall.
That's not surprising, since those are fairly easy to measure.
However, I bet if you could measure
artistic, musical, social, and emotional skills,
the data would look less cigarre-shaped.
IMVHO that's a classic fallacy of a managerial-driven thinking: numbers are not the essence of anything, they can be used in many things to summarize to the extreme something but not everywhere for everything.
It's like to say "how many degree of temperature and humidity there are in this room"? Most just look at nearest thermo-hygrometer and say out load a number, let's say 19℃ considering that number "the truth" at maximum with the relative precision error of the meter. Actually that's a "somewhat realistic approximation", a room is not a well insulated calorimeter, so I can say that's 32℃ just because behind the windows with direct sunlight that's the temperature locally measured with 30% humidity, but also on the other extreme of the room, one of an old masonry house placed direct on ground soil, where I measure 14℃ and 60% humidity.
For intelligence it's not much different, only even harder to measure. We can say someone is very skilled in math so "he/she is intelligent" and craft some kind of test (like uni admissions tests common in various faculties around the world) but can also be absolutely not much skilled in math but very smart in talking and comprehend peoples so we can craft a "social ability skill" intelligence scale etc and finally combine all those test in a giant one and establish a scale, but again that scale is a kind of generic approximation of something we try to "standardize" after a long process of observation on many peoples abilities. It's not a real measure out of something tangible and really meaningless, for instance if you need a developer how it's intelligence score matter? Someone might have a high score but still being far less able to sole certain kind of programming tasks than someone else.
Long story short: there is no real and meaningful measure for intelligence, we just try to pretend there is something to craft studies on that produced like an industrial product because in a managerial-driven society anything must be like that, it's not possible to think about immeasurable, not programmable Gantt diagrams style things.
135 comments
[ 0.21 ms ] story [ 180 ms ] threadI know many very outstanding programmers - but very few are also outstanding at writing poetry, or learning foreign languages, or engaging in impactful political action.
I've no doubt that 'general science' performance is correlated with 'assembling objects' performance. But I wouldn't expect Bobby Fischer to have great insights into world politics just because he was a chess grandmaster.
But how do your outstanding programmers compare at writing poetry... when compared to your not-outstanding programmers who also have no specialized training?
Intelligence is a necessary but not sufficient condition for competence at intellectual tasks.
But, no matter how good you may be at any one of these things, to become good at the others you must spend time focused on learning and doing them. Most of the great programmers I know prefer to spend their time in pursuing other things that they become good at instead, primarily stuff to do with physical engineering. (Aside, I've always thought the propensity for using physical engineering as a non-work hobby is guided by frustrations from the intangible nature of programming/software.) But, I do in fact know several that travel and learn foreign languages, or that write poetry and short stories as a form of outlet.
All of the evidence we've gathered over decades seems to indicate that there is a general factor for intelligence. But as he points out in his conclusion, this is /descriptive/, meaning we can arguably accurately determine someone's intelligence via testing and that single general factor accurately describes their intelligence in relation to others. It does not in any way imply we understand the causative factors that lead that score to go up or down in any meaningful way.
For reasons I don't fully understand, there's a lot of folks in the social sciences these day that seem to be waging war on the very concept that you can describe intelligence generally (and IQ tests specifically). The mere fact that the results of these types of tests on large population groups tend to follow a normal distribution is alone good evidence that they're indicative, as in most things diverse large population groups tend to fall on a normal distribution.
I think it's basically coming down to a battle of the mathematically literate vs. not, and unfortunately the non-literate have a lot of stickiness in academia because they can generate ideas people want to believe with no empirical "where the rubber meets the road".
> For reasons I don't fully understand, there's a lot of folks in the social sciences these day that seem to be waging war on the very concept that you can describe intelligence generally (and IQ tests specifically).
The unfortunate reality is that social sciences are largely corrupted by political activism. Their activities are often prescriptive social engineering that attempt to use their perceived scientific authority rather than actual descriptive science. Put simply, they are saying what they want to be so, rather than what is so.
So I think to get everyone on the same page, you have to address that concern in some form, or there will be some people who are always against the idea of measurable intelligence.
It's a reasonable argument, and one I personally tend to lean towards, but you can't meaningfully debate it if you don't have a shared understanding that a general factor does indeed exist.
I think it's reasonable to conclude that measuring intelligence isn't valuable enough for all of the negative externalities attached to it given how the world currently is. But then from there it's easy to make the easy jump to saying it's impossible to measure intelligence without those negative externalities, which is probably not true. But if you can't measure intelligence without significant bias, you aren't really measuring it at all, you're just measuring your biases.
Do you have some specific examples of this purported bias?
> Unfortunately some were also very deliberate spins (one notable example being the book The Bell Curve).
Specifically what spin are you referring to?
> I think it's reasonable to conclude that measuring intelligence isn't valuable enough for all of the negative externalities attached to it given how the world currently is.
Insufficiently valuable to whom?
If there is a general intelligence—which I very much doubt—our current tests don’t even come close to measuring it.
Then why do they correlate with outcomes so well ?
And re: normal distribution implying some fundamental truth, the same can be said of an iterated guessing game (eg: take the sum of your correct guesses over 100 coin-flips - the population's scores on that game will be normally distributed).
As pointed out by Stephen J. Gould, intelligence is not normally distributed. The IQ test results are. But IQ test results are standardized to be normally distributed. It is an assumption that intelligence is normally distributed, and evidence suggests it is not. More likely—if it exists at all—it is log-normally distributed.
Psychometricians believe that there is some fundimental truth in the normal distributions, and then standardize their test scores so that it follows this distribution in order to gain validity. It is a pseudo-scientific self fulfilling prophesy.
It’s funny you should say that. Because I generally regard the authors Gould is addressing in his book Mismeasure of Man, including S.G. Morton, C. Murray, A. Jensen, etc. as hacks that used questionable scientific means to tell people what they wanted to hear—namely the superiority of white men—in exchange for book sales.
Morton was wrong.
https://en.wikipedia.org/wiki/Samuel_George_Morton
It's great that you found some super specific thing where his wrong answer isn't provably the result of fudging of numbers, but he's still wrong for various reasons. And someone else measured other skulls at roughly same time and came to the correct conclusion that you couldn't really make any such claim based on the data they had.
Gould's point was about inadvertant bias in science and he was correct in this case.
You might be right that Gould was a fraud, however you still have to show that his claim about the fact that intelligence isn’t normally distributed is wrong. In fact the burden of proof is kind of weird here, since assuming a normal distribution is a pretty strong assumption, and rejecting that assumption based on a lack of evidence should be enough. However Gould goes further and cites evidence that intelligence has a pretty skewed distribution.
If you want to reject Gould’s claim here, then do so. Don’t just say he’s a fraud and then assume his wrong about everything.
The problem is assuming standard normal distribution as the true distribution for intelligence is simply bad science. Whatever the distribution truly is (if there even is one) it is not this one. And then psychometricians (including OP) use this as a point in favor of the validity of IQ as a good measure.
To give an uncontroversial example, we use logarithmic scales for earthquakes precisely because the underlying phenomenon is heavy-tailed. It's just for convenience, it doesn't imply anything about the phenomenon being quantified.
> The mere fact that the results of these types of tests on large population groups tend to follow a normal distribution is alone good evidence that they're indicative, as in most things diverse large population groups tend to fall on a normal distribution.
The claim here is that OP is wrong. Firstly OP is wrong because normal distribution is not a sufficient indicator for some underlying phenomenon. Or as yxwvut puts it:
> [...] normal distribution implying some fundamental truth, the same can be said of an iterated guessing game (eg: take the sum of your correct guesses over 100 coin-flips - the population's scores on that game will be normally distributed).
And then secondly because intelligence doesn’t even follow a normal curve (only IQ does because it is transformed to).
What matter here is not the true distribution of intelligence. I’m not even sure there is one. What matters is that proponents of general intelligence use the normal distribution of IQ as a justification. By arguing that intelligence does not fall on normal distribution but rather on a skewed distribution, I’m simply pulling the rug away from this argument.
I admit this is not a strong argument against the existence of general intelligence, since if there exists such a thing, then there is nothing that says such a thing couldn’t exist on a skewed distribution (except psychometricians and proponents of IQ which only accept standard bell curve [I wonder why]). However in this subthread we’re only trying to correct OP’s claim that a normal distribution is indicative for general intelligence.
I actually don’t believe in general intelligence, and I don’t believe intelligence is this one thing that you can measure. So I see how it can be confusing when I go ahead and say: “Intelligence has a skewed distribution.”
But when I’m talking about “intelligence” in this context I’m using working within the assumptions of the psychometric. In other words, I forgo my priors and argue on the terms of my opponents. I.e. given that they are right about the existence of general intelligence and that IQ tests are a good and accurate way to measure intelligence, they are still not correct about it being normally distributed.
To be precise. When I use the term “intelligence” in this context I’m basically using the same operationally defined term as psychometricians. That is: Intelligence is whatever the IQ tests are measuring.
Now as for which distribution to pick for this thing that IQ tests are measuring. Why pick a skewed distribution instead of a normal one? Psychometricians pick the bell curve because they believe it gives this term some validity. I would pick a skewed distribution because that fits our intuition and fits the underlying data. I would call the former justification pseudo-science as you have no reason other then your prior assumptions to do that. Then why do the latter? I would say the only reason to do that is to prove the people doing the former wrong.
Again, regarding earthquakes: Does a logarithmic scale "fit the underlying data better"? I don't think the question even makes sense. It's like asking which font fits the underlying data better. It's an arbitrary choice that doesn't need to affect our predictions or models at all. But you also said that "the only reason to do that is to prove the people doing the former wrong." so maybe we agree on this, too.
I think you might be misunderstanding psychometrics if your objection to IQ is the supposed implication that "intelligence is this one thing that you can measure". As far as I understand, psychometricians would agree that intelligence has many facets, and that most of them are hard to measure. I think the claim that they're making is simply that one can summarize estimates of these facets into a single scalar that will be correlated with all of them. Just like GDP correlates with many hard-to-measure and complicated facets of a country's wealth. Would you say the fact that GDP can't measure some important things about a country's wealth, or that it's a complicated thing, implies that there's no such thing as "general wealth"? The same analogy works for e.g. health, strength, or pretty much any property that we have a word for.
Measuring wealth is an excellent example. GDP might correlated with other measures of wealth, but it hardly explains the whole story. It can be useful in models but you can’t summarize nation’s wealth by only looking at GDP. Similarly IQ can be a useful tool. IQ tests can be used to find people with mental disabilities so that appropriate accommodations can be provided (although most of the time, there are easier, cheaper and better tools to diagnose mental disabilities). However by looking at the IQ alone (especially around and above the mean) you really can’t say much about about the mental capabilities of a person, only correlations of varying usefulness..
There is an important distinction though. I would argue that GDP (arguably) proves it’s usefulness in different economic models. However the same cannot be extended to IQs. I don’t know of many psychological models which use IQ (or any notion of general intelligence for that matter) to explain behavior (other then a hacks trying to prove the superiority of the white race, but let’s ignore those, what they are doing, it is not science).
How is IQ not "explaining behavior" a problem? We also wouldn't say that GDP "explains the economy", but that's not necessary for it to be useful. Both of these statistics are relatively strongly correlated with many outcomes we care about, compared to other things that are about as easy to measure. I think that's the main reason people use them, and it's a good one.
It sounds like you've seen people over-use aggregate statistics as a way to downplay nuance. That must be annoying. But that doesn't make their use philosophically or scientifically invalid.
Yes. That pretty much summarizes it. If IQ doesn’t further our knowledge of human behavior nor the human mind it is at best a useless construct. We have other more targeted measures (such as SAT) which are easier and cheaper to administer, and are far more useful in both models and evaluations/diagnostics. GDP has the benefit here of being extremely easy to measure and is therefor useful in more general models. IQ doesn’t have this quality, so that is not why most people use it.
Now if that was the only problem with IQ I wouldn’t be so bothered by it. Scientist are allowed to use an inefficient means of measuring their constructs. However IQ advocates often have a vested interest in an IQ that is immutable and (to some degree) heritable. This is often assumed to be true and IQ tests are later fixed in order to make them more robust in score against subsequent trails. This—off course—is circular reasoning (or a self fulfilling prophecy) and is not good science. But it is still used to explain group differences. This is where using IQ becomes scientifically invalid.
Now without the claim of immutability and heritability this wouldn’t matter. We would see a group difference and then we could then go on and try to explain it (e.g. within the context of sociology). However with immutability and heritability, you don’t have to explain it further. The answer is in psychology and genetics. (This is by the way the basis for scientific racism mentioned before).
I mean, I think it does further our knowledge of these things, but even if it didn't, that wouldn't make it useless, because it still predicts things we care about pretty well.
> GDP has the benefit here of being extremely easy to measure... IQ doesn’t have this quality
Which measures of intelligence do you think are easier to measure than IQ? You mention the SAT positively below, which is very similar to an IQ test.
> IQ advocates often have a vested interest in an IQ that is immutable and (to some degree) heritable. This is often assumed to be true...
I mean, every researcher would say that it's only immutable and heritable to a degree, and there's tons of research quantifying the degree of mutability and heritability (e.g. twin studies, admixture studies).
> But it is still used to explain group differences. This is where using IQ becomes scientifically invalid.
Is there a specific claim you have in mind that you consider invalid? Can you give me an example of an explanation of group differences that you would consider scientifically valid?
> with immutability and heritability, you don’t have to explain it further. The answer is in psychology and genetics.
Are you saying these claims happen to be false, or that they're categorically impossible / bad explanations? Is there any aspect of human behavior that you think can be explained by psychology or genetics?
Yes. The Bell Curve by Richard J. Herrnstein, Charles Murray (1994). I’m currently having a debate about it in another thread. It uses these assumptions about IQ and the demonstrates group difference with flawed studies, which it then goes on to justify conservative policies.
An example of valid—albeit trivial—group difference on the is height difference across genders. A more interesting group difference is how a mock jury is less likely to convict a minority group if a member of that minority is present in the jury. Then of course the field of medicine has many examples of interesting group difference based on genetics. However as a species we humans are remarkably homogeneous so if you find a group difference in behavior based on genetics it will most likely have an extremely small effect size, and as such be of limited interest. At best this is where IQ research falls in.
> There's tons of research quantifying the degree of mutability and heritability (e.g. twin studies, admixture studies).
These are still up for debate and not without bias. The original twin studies by C. Burt were likely falsified (https://en.wikipedia.org/wiki/Cyril_Burt#%22The_Burt_Affair%...) later efforts of reproducing the results are not without controversies either. And even if you look beyond these controversies there is an inherent bias as adopted twins are way more likely to be adopted within the same socio-economic group, adjusting for that is really tough. I’m not aware of any admixture study which explains a significant group difference in behavior, so I can’t comment on this. But I’ll leave this quote from Wikipedia instead (https://en.wikipedia.org/wiki/Race_and_intelligence#Racial_a...):
> Reviewing the evidence from admixture studies Hunt (2010) considers it to be inconclusive because of too many uncontrolled variables. Mackintosh (2011, p. 338) quotes a statement by Nisbett (2009) to the effect that admixture studies have not provided a shred of evidence in favor of a genetic basis for the IQ gap.
(It doesn't exist)
See for instance:
Castejon, J. L., Perez, A. M., & Gilar, R. (2010). Confirmatory factor analysis of Project Spectrum activities. A second-order g factor or multiple intelligences?. Intelligence, 38(5), 481-496.
Visser, B. A., Ashton, M. C., & Vernon, P. A. (2006). Beyond g: Putting multiple intelligences theory to the test. Intelligence, 34(5), 487-502.
If SAT score were measuring some immutable value, then it wouldn't be so easy to improve one's score with revision, and tutors.
I think that thinking processes is very much in that category, and I think that the investment in time is considerable. I think its typical to expect gradual improvement -- a constant response to training. Anecdotally, it doesn't work like that.
Consider the physical fitness correlations in the article - I don't think anyone would argue that your level of physical fitness is immutable. But just because the value is mutable is no reason to disregard the current value - it can still be a meaningful and useful indicator.
Most money spent on test prep is just wasted.
0. https://en.m.wikipedia.org/wiki/Intelligence_quotient
1. https://www.manhattanreview.com/sat-retaking/#:~:text=The%20....
So SATs are not generally trainable by a universal approach -- including personal tuition. I think that is believable. However, that for any given individual there is no intervention which will substantially improve their SAT results is a much stronger statement which I find less believable.
I think that the treatment of things like time factors and tool use is also somewhat trivialised. Those that play chess will note that excellent Rapid or standard players can be terrible Bullet or Blitz players.
Roger Penrose -- nobel prize winning mathematical physicist -- was held back a year on count of his arithmetic being that bad. Yet he is a shockingly good mathematician mostly qua geometrical intuition to which he can convert other things. Who cares, Penrose can just use a calculator. Similarly, I'd say that many IQ tests are just symbolic regression. What if I was brilliant at stating symbolic regression problems, but poor at solving them? Then me + solver would make all the difference, unlike if I was poor at stating them ...
And so on.
Showing outliers just proves there are outliers not that the averages are uselesss.
>Those that play chess will note that excellent Rapid or standard players can be terrible Bullet or Blitz players.
I play chess and those are definitely highly correlated, I suspect at least r of 0.7 especially for those who have put decent time into the different controls.
Regarding chess, research suggests you're wrong:
https://www.sciencedirect.com/science/article/abs/pii/S14690...
https://hvandermaas.socsci.uva.nl/Homepage_Han_van_der_Maas/...
I think part of the issue is that it's very hard to measure intelligence accurately (whatever that means) across cultural divides. If you gave an American IQ test (e.g. an SAT) to Europeans, Asians, and Africans you'll likely get different results, not (necessarily) because of innate differences but because of educational/cultural backgrounds. Poverty levels can also change scores.
This makes it tricky to know what to do with these scores and we need to acknowledge that. This leads some to conclude that we should get rid of the notion of a measurable IQ altogether. That goes too far in my opinion but I understand their concern.
I think you're on the right track, though. But the key sticking point is this: many people are very uncomfortable with the idea that there are differences in intelligence among people, both within and between populations. Unfortunately, that is undeniably true. And so we get this ridiculous, unproductive dance.
Also, just as an aside, the latest evidence around the "poverty actually makes you literally stupider" claim is that it's not supported--studies aren't replicating. Which makes sense because the claim is prima facie crazy, but I digress...
The reason is that general intelligence tests show differences between racial and ethnic groups. Which isn't shocking, as intelligence is certainly driven by genetics and compounded by culture. But because it is, it upsets people.
IQ testing is a pretty valuable filtering mechanism. Smart people are really important to advancing society. Replacing it with some other, more subjective metric just moves the filter into the realm of political control.
Which seems like a bad idea to me, but what do I know? I'm not a genderqueer left-handed ginger aboriginal, so I'm not very smart.
It's the social sciences version of technical debt.
> we value not upsetting people, that's the metric we optimize for.
From the HN guidelines (https://news.ycombinator.com/newsguidelines.html):
> Eschew flamebait. Avoid unrelated controversies and generic tangents.
I’m a prime target for this flamebate, and you’ve managed to upset me quite a bit. I’m actually surprised I didn’t bite already. Indeed I’ll actually just leave it at that and pat my self on the back for resisting.
Given that we cannot say what intelligence is better than "its what intelligent people do", I'd say we were far from the "whole story".
I think pretending that there is no "whole story" and using IQ to make policy decisions about populations is political. I don't think anyone cared that a bunch of researchers were tinkering with some psychometrics. They cared that a whole racial group could be shut out of certain jobs on the basis of a half baked abstract reasoning test.
I'm not opposed to this, and I doubt most people would be either. Take college admissions as an example. A low-SAT, low-GPA student in a one-on-one interview with an admissions officer could dazzle the interviewer to such a level that admission with a scholarship could be guaranteed.
We don't do that often for a variety of reasons, but one would certainly be because of the quasi-legal requirement to tick certain boxes to avoid trouble. Okay, but that has costs. We've been able to paper over a lot of this with economic good times, but that's unlikely to last forever. Hard reality always wins in the end.
The results do not follow a normal distribution though. IQ is forced into a normal distribution with a standard deviation of 15 by definition, it is constructed in such a way that it's forced into that distribution. The actual measurements, ie. raw test scores, are not normally distributed.
It's also possible that single cultural factors could be so pervasive that they overwhelmingly swamp any signal by forcing a correlation just on its own.
eg: Before the 1960s or so, you could look at every society that had ever existed and make the observation that, except for a countable list of exceptions, men received more education that women. This might lead you to a natural conclusion that men have some natural innate advantage in receiving education that women lack.
Except that's not the truth at all, the truth is that patriarchy was such a universal of the human condition that it affected every single corner of the planet to such an extent that there was no control group until very, very recently. So much so that even the slightest letting off of patriarchy shows women are so much more innately suited to education that colleges have had to quietly implement affirmative action programs for men simply to keep their gender balances even remotely in check.
Many people would argue that classism and racism so thoroughly pervades our understanding of intelligence that we will simply never get our control group until some future date when they recede enough for us to truly see their effects.
The fallout from eugenics gave social science a complex. Innate difference and especially innate superiority of one person over another is taboo. The field is poisoned by motivated reasoning.
>At the height of the controversy, during the 1970s to 1980s, the debate was highly ideologised. In Not in Our Genes: Biology, Ideology and Human Nature (1984), Richard Lewontin, Steven Rose and Leon Kamin criticise "genetic determinism" from a Marxist framework, arguing that "Science is the ultimate legitimator of bourgeois ideology ... If biological determinism is a weapon in the struggle between classes, then the universities are weapons factories, and their teaching and research faculties are the engineers, designers, and production workers." The debate thus shifted away from whether heritable traits exist to whether it was politically or ethically permissible to admit their existence.
https://en.wikipedia.org/wiki/Nature_versus_nurture#Determin...
There's often a bias towards assuming that physical skills are trainable while mental skills are not, because many people have been raised to assume that the latter are due to mainly to genetic factors. This 'genetic superiority' is typically then used to justify current class structures, racial makeup of the wealthiest 1% etc. Such views have little if any supporting scientific evidence - mental skills are just as trainable as physical skills (yes there's a genetic component in both cases, which might place an upper bound on one's performance, but such genetic factors are not correlated with race/gender/etc.).
I'm guessing there is an equivalent with mental capabilities. Somethings you can train, some are innate.
Something like the SAT is a game that people learn to win.
This is oft claimed, but data doesn't back it up. SAT is marginally affected by coaching and training.
"Coached students are only slightly more likely to have large score gains than uncoached students. In addition, about 1/3 of students experience no score gain or score loss following coaching. The typical gain associated with coaching is 8 points for verbal and 18 points for math"
https://www.washingtonpost.com/news/answer-sheet/wp/2017/05/...
The situation is somewhat the same with SAT, but it’s not so bad in actual reality: there is no single SAT test, there are multiple instantiatons, and you can’t just memorize all answers for all SAT tests. At the same time, the actual problems on the test come from a rather limited pool of patterns, and if one learns to tackle each pattern, one can really improve their SAT score. This is, in fact, perfectly commonplace: everyone significantly improves their SAT score between grades 8 and 12 (at least among people who do take SAT at both ages).
Does it make SAT useless, as everyone learn all the patterns and gets perfect score? Empirically, no: there is large variation in actual obtained scores, and the results are mostly in expected directions: “better”, “smarter” students get better scores and vice versa.
Why doesn’t everyone get perfect score? Quite simply, because variation in mental ability also makes it so that learning the pattern is easier to some people than it is to others. This is by far the source of most of the variation: there is some residual component of conscientiousness, but it’s much less important: spending more time cramming has lower returns the less able you are, and for everyone, there are significant diminishing returns to extra hour of cramming, and ultimately high scorers don’t differ that much in terms of how much time they spend prepping to SAT compared to mid-scorers.
The evidence that you can train for the SAT is pretty scarce.
https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1745-3984.... https://nepc.colorado.edu/sites/default/files/Briggs_Theeffe... https://slate.com/technology/2019/04/sat-prep-courses-do-the...
Certainly I don't think I've ever not done better the second time I looked at a paper, so maybe my experience is just special.
This isn't a very good example, because the reason it gets harder is that increasing the body's muscle mass also increases the strength needed to propel it a certain distance. Hence gaining muscle mass both helps and hurts the vertical jump. Is there any reason to believe intelligence should be like this?
For the layman, probably. Actual IQ tests try to ensure that subjects don't have experience with the test tasks. If they do, then the test is not as effective at measuring innate ability to learn or problem solve since it has been trained and the strategy to solve is already known. So it shouldn't be a misperception in the fields that deal with it.
Anyone who has actual and nontrivial experience in teaching (say, over 50 students over at least few months) will tell you this simple empirical fact: some people just “get it” much easier and quicker than others. The only way to escape this reality is to sort people by mental ability, so that you only see people in the same ability bracket, but that’s just it, an escape.
We collectively have enormous amount of experience in educational interventions, and, in short, none of it really works. It’s easy to complain about lack of resources, about low teacher quality, about wrong methods, and all of these are in fact real problems, and pretty complex and hard to tackle too. At the same time, even if we pull each of those to the state of the art, we can only expect a marginal improvement.
Have you noticed how the educational establishments seem to try a new thing every 5-10 years or so? That’s because none of it really pushes the needle, and these 5-10 years are what it takes to realize that. In reality, the “new thing” is in fact usually an old thing that has already been tried a generation or two ago, but which has not been seen by the current generation of education workers. That’s because everything has already been tried, and none of it really works.
On a more optimistic side, that means that even if the education system is suboptimal, it doesn’t really matter, as outcomes are more or less the same anyway.
The poster you're replying to said "just take the lower half of the cohort in each categorty (mental/physical) in a given study and give them intensive one-on-one training in all areas for a year". Pretty much every study has show that one-to-one tutoring is far, far more effective than the standard "education" system.
>We collectively have enormous amount of experience in educational interventions, and, in short, none of it really works.
What are you talking about? There's plenty of evidence that one-to-one tutoring can be quite effective at increasing test scores. Yes this approach can't be deployed at scale due to the sheer cost of it, but that's completely irrelevant to the question of whether intelligence is trainable.
You should watch some https://en.wikipedia.org/wiki/Memory_sport competitions, would blow your mind what people are capable of training themselves to do.
> You should watch some https://en.wikipedia.org/wiki/Memory_sport competitions, would blow your mind what people are capable of training themselves to do.
What some people are capable of, yes. Does it translate to those people being better at other mental tasks of mental ability? No.
Two sigma https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem is not a "marginal" effect. And a chunk of that can be explained by mastery learning, which while still hard to scale up is far easier than individual tutoring as a whole. The educational system is not institutionally incented to effectively pursue these empirically workable approaches. It's a matter of bureaucracy-driven failure, not some immutable fact about the world.
There's even specifically developed approaches to effectively train and educate those who might be often described as 'low I.Q. individuals', such as "direct instruction". Guess what, they get no takeup from the educational establishment, simply because they're (quite wrongly!) perceived as being demeaning for teachers.
The average teacher would rather pretend to educate those who are already highly functionally literate and familiar with educational dynamics (including test taking, of course) and can simply "learn by themselves". Education schools consistently cater to these attitudes and push the same approach.
Consider learning how to throw a ball. Do some new students 'just get it' - or did their parents spend hours tossing a ball with them previously? Point being, those anecdotes about students who just get it don't factor in variables like earlier childhood education etc. Few people would argue that such physical skills are not eminently trainable - and clearly, mental skills are also trainable.
The question is, is there something like a clearly delineated genetic factor than makes it easier for someone to, for example, learn a new language, relative to someone lacking that factor? No real clear evidence seems to exist for that claim.
There’s an absolute avalanche of data on this topic - do you want pointers to it, or are you looking to soapbox?
It's not soapboxing; in particular it would be interesting to see if the 'ability to learn and improve' changed with age. It's commonly assumed that it's harder to learn new things as we age, but I've never seen that data quantified personally... Hopefully it's less true than we imagine it to be.
This is the seminal paper on one-on-one academic tutoring from 1984 that’s spawned hundreds of follow-up papers: http://web.mit.edu/5.95/readings/bloom-two-sigma.pdf.
The biggest corpus of raw stats on age-related learning AND the generalizability of that knowledge come from the fad of brain-training games from the 2000s (e.g. https://scholar.google.com/scholar?hl=en&as_sdt=0%2C44&q=bra...).
But the crux of what you’re talking about, i.e. “physical skills are trainable while mental skills are not”, fall under the banners of “transfer of learning” for cross-domain academic performance, and “malleability of intelligence” for age-related differences in cognitive improvement: e.g. https://files.eric.ed.gov/fulltext/EJ903882.pdf, https://psycnet.apa.org/doi/10.1037/0033-295X.109.1.116.
Plenty of exercise physiology research on coaching format (personal trainer vs. group classes), and strength training adaptations by age that shouldn’t be hard to find at all to cover the physical side.
Here is a short and incomplete list: ability to handle sun or lack thereof, ability to handle heat or lack thereof, ability to live in high altitudes, ability to process lactose, ability to tolerate alcohol, resistance to malaria, height.
There are also diseases correlated with groups of genetically related people.
However, these groups most often don't fit so cleanly into the US's constructs of race.
There could be mental skills correlated with groups of people with shared genetics.
The question is how we handle that if that is the case. One approach is to deny it. Another is to force or encourage people to have children with others to smudge out differences. Another is to let it be.
I suspect that the unspoken answer is denial while encouraging the smudging out of visible differences. I suspect that people are morally against the last and that it would lead to massive conflict.
I think a similar calculus has happened many times before. I think that is partially the origin of "whiteness".
Correlations stop working as we move above average, though. Paul Morphy was good at chess, but not so good as a lawyer. Michael Jordan is great at basketball, but not at golf.
edit: so would be higher numbers because would have to take into account number of people who have been pro athletes at some point in their life, not just right now. But not enough higher that the point doesn't stand.
But that being said, the visualization on this are top notch, and the effort put into them shows - fantastic article in terms of presentation, and hats off to the author for putting it together.
As each axis measures something different, there is some sense of unit dependency (e.g. measuring height in one axis (cm) and measuring weight in another axis (kg)). It is silly to say one number in one axis with a different unit is comparable to another one in another axis.
A different way to say it is that each axis can be (should be) normalized before any kind of factor analysis. The cigar shape is as much a function of per-axis normalization than whatever the underlying pattern is.
I think it would make sense to define it as something like: the extent someone has well calibrated Bayesian reasoning that exploits information to its maximum to do accurate predictions.
It's hard to test though because there is a dearth of synthetic benchmarks or algorithms to compare to. Humans seem to still be better at performing these types of Bayesian "common sense" tasks than any synthetic procedure, so we can only compare using tests that we subjectively agree other "smart" humans seem to be good at.
The concept of g actually completely rules this interpretation out based on nothing. If g was a matter of learning reasoning skills and applying them, you'd be able to teach it. It wouldn't be some magical race/sex/class number.
everything else is a useless metric to sell yourself
I think this part in the conclusion somewhat contradicts a lot of the analysis. In fact, the results have shown that if you knew someone's g, your guess at their performance on each task would be a good guess.
Thing is, building the ontology is the most important part of the scientific process. Otherwise you're just praying to a blind idiot god and listen into the void for answers. Or as the author puts it, "statistical arcana."
To be concrete, I'm pretty confident I overtook half of humanity in that fitness test in the last year. Not because I was bitten by a radioactive spider, but because I started to work out. Now to claim the analogy to intelligence, that would need an answer to what the analogy of working out is in intelligence.
Or to look at it on the level of people who compete in a sport, I'm quite confident that Tom Brady is not very fit by NFL player standards. Similar for Xavi in soccer, but Christiano Ronaldo is probably near the top, and I don't have a good intuition in case of Messi, because he is a quite small player and that will have an effect.
The preceding criticism is enabled, by having a good intuition of what the fitness test is, I have a pretty good intuition about the mechanics of a push up. That is not there for intelligence, and I strongly suspect that this whishi-washiness is whats actually holding up all arguments about iq, since it prevents concrete criticism.
In fact the foundations of IQ tests lies heavily on the eugenics movement which wanted to prove the superiority of able bodied white men. They actually started by assuming brain size correlates with intelligence (it doesn’t) and then set out and proved that on average white men had the biggest brain of them all (which is also false). The nuances of fitness does not help in this goals as one can easily just pick a type of intelligence/fitness which white men perform poorly at in general so their claim was that there exists a general intelligence to remove this nuance. Their current IQ tests are all set up to assume this is true, and hence you can see it in the results. If you can’t then you are not really measuring intelligence.
Books such as The Bell Curve are not in and of them self racist. This book in particular makes zero racist claims. However it relies on deeply racist studies as its premise, so if you want to conclude anything from it, that conclusion is racist.
I don’t think every proponent of IQ is racist, however there is a lot of racism embedded in the research, sometimes directly, but most often by association.
The article mentioned testing physical strength of 20 and 80 year olds that reminded me of this.
Which age does best on raw test scores? Seems surprisingly hard to Google this info but I assume it's out there?
edit: a renewed effort to google suggests the things IQ tests test peak in late teens.
https://journals.sagepub.com/doi/pdf/10.1177/095679761456733...
It's like to say "how many degree of temperature and humidity there are in this room"? Most just look at nearest thermo-hygrometer and say out load a number, let's say 19℃ considering that number "the truth" at maximum with the relative precision error of the meter. Actually that's a "somewhat realistic approximation", a room is not a well insulated calorimeter, so I can say that's 32℃ just because behind the windows with direct sunlight that's the temperature locally measured with 30% humidity, but also on the other extreme of the room, one of an old masonry house placed direct on ground soil, where I measure 14℃ and 60% humidity.
For intelligence it's not much different, only even harder to measure. We can say someone is very skilled in math so "he/she is intelligent" and craft some kind of test (like uni admissions tests common in various faculties around the world) but can also be absolutely not much skilled in math but very smart in talking and comprehend peoples so we can craft a "social ability skill" intelligence scale etc and finally combine all those test in a giant one and establish a scale, but again that scale is a kind of generic approximation of something we try to "standardize" after a long process of observation on many peoples abilities. It's not a real measure out of something tangible and really meaningless, for instance if you need a developer how it's intelligence score matter? Someone might have a high score but still being far less able to sole certain kind of programming tasks than someone else.
Long story short: there is no real and meaningful measure for intelligence, we just try to pretend there is something to craft studies on that produced like an industrial product because in a managerial-driven society anything must be like that, it's not possible to think about immeasurable, not programmable Gantt diagrams style things.