5 ms·

all of its multi faceted glory with a single number Think of this number as the first principal component, generated from all these different things. Talking a

by iguy 9y ago

all of its multi faceted glory with a single number Think of this number as the first principal component, generated from all these different things. Talking about it doesn't imply that other directions don't exist. But it's not so strange to work on the leading effect before worrying about sub-leading ones.

pasabagi 9y ago
As a sort of reductio-ad-absurdum, imagine if somebody came up with 'GQ', or 'goodness quotient'. It's absolutely the case you could use a battery of tests to check how likely somebody is to do anti-social behaviour. 'Goodness' is about as conceptually well-defined as intelligence.
- iguy 9y ago
  Isn't the chinese government working on that? :) But more seriously, yes, to figure out how useful the first component is, you want to know something about how much bigger the first eigenvalue is than the second. For the case of IQ, my understanding is that the answer is "quite a lot". The best studies are from the army testing people and assigning them to roles (and seeing how they do). And it seems pretty hard to come up with tests which are better predictors than IQ. I think the one perpendicular component worth measuring for some jobs is manual dexterity? And eyesight, which is a rare thing anti-correlated with IQ, because of teenage bookworms we think. (That said, note that this isn't the whole population, they set a floor by only accepting reasonably healthy young men etc.)
  - pasabagi 9y ago
    The chinese government is working on it - but they are, at the very least, not convinced it is science. Not to mention, most people are horrified. I don't doubt that IQ, like Social Credit, is a useful administrative tool. However, something being useful from an administrative standpoint does not make it scientific. It also doesn't make it desirable. There are very many things that are absolute inconveniences from an administrative standpoint, that are absolutely necessary from a human one. Rights, basic freedoms, and so on.
    - iguy 9y ago
      It doesn't have to be junk science to be morally repugnant. Lots of things would be administratively useful, and aligned with everything we know scientifically, yet evil. This is a great idea, morally: "We hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights". But as you know if you've ever watched the Olympics on TV, "created equal" here isn't a scientific fact.
      - pasabagi 9y ago
        Absolutely. But equally, something being useful does not make it scientific. The thing about running is, it's very easy to explain what we mean by 'good at running'. It's very hard to describe what we mean by 'intelligence'. When you see a very large number of people convinced of the objectivity of something they can't describe, you normally call it a religion. Being convinced of the objectivity of IQ has many of the same flaws.
        iguy 9y ago
        No it's not religion. It's statistical knowledge, in a way that people seem to struggle to grasp. (And there are thorns! What use it is put to is another issue.) If you look at all sports you'll find the same thing. We call the common factor physical fitness, or athleticism, or let's say "AQ". It's exactly the same idea. AQ predicts running speed, although not nearly as well as getting the kids out on the track. Ditto weightlifting. People know this: the kids in the chess club won't agree to arm-wrestle the kids on the track team, because despite having no direct information of their biceps strength, they can guess based on overall impressions. This strategy for keeping their lunch money is scientifically valid. And their profiling of the jocks is a safe distance from being morally repugnant.
        pasabagi 9y ago
        The problem with statistics is, if you put garbage in, you get garbage out. Especially in the case of garbage categories, and garbage questions. Before you do any statistical work, you have to ascertain whether or not the groups you are considering, or the questions you are asking, are actually existing groups and scientific questions. IQ fails this test. Race fails this test. Running does not. If make statistics about ill-defined, politicized questions, at best, you get nonsense. At worst, you get a reinforcement of your muddy, politicized thinking. The problem with science is, if you're not very careful, it can become a wrapper around stuff that isn't scientific. The only way to avoid this is to be absolutely allergic to any concepts that are unclear, uncertain imports from the normal world. And, while I'd love it if intelligence wasn't like that - let's face it, it is. If it wasn't, and we had a clear image of it, we'd probably be capable of doing stuff like building general AI.

by iguy 9y ago

pasabagi 9y ago

As a sort of reductio-ad-absurdum, imagine if somebody came up with 'GQ', or 'goodness quotient'. It's absolutely the case you could use a battery of tests to check how likely somebody is to do anti-social behaviour. 'Goodness' is about as conceptually well-defined as intelligence.

iguy 9y ago
Isn't the chinese government working on that? :) But more seriously, yes, to figure out how useful the first component is, you want to know something about how much bigger the first eigenvalue is than the second. For the case of IQ, my understanding is that the answer is "quite a lot". The best studies are from the army testing people and assigning them to roles (and seeing how they do). And it seems pretty hard to come up with tests which are better predictors than IQ. I think the one perpendicular component worth measuring for some jobs is manual dexterity? And eyesight, which is a rare thing anti-correlated with IQ, because of teenage bookworms we think. (That said, note that this isn't the whole population, they set a floor by only accepting reasonably healthy young men etc.)
- pasabagi 9y ago
  The chinese government is working on it - but they are, at the very least, not convinced it is science. Not to mention, most people are horrified. I don't doubt that IQ, like Social Credit, is a useful administrative tool. However, something being useful from an administrative standpoint does not make it scientific. It also doesn't make it desirable. There are very many things that are absolute inconveniences from an administrative standpoint, that are absolutely necessary from a human one. Rights, basic freedoms, and so on.
  - iguy 9y ago
    It doesn't have to be junk science to be morally repugnant. Lots of things would be administratively useful, and aligned with everything we know scientifically, yet evil. This is a great idea, morally: "We hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights". But as you know if you've ever watched the Olympics on TV, "created equal" here isn't a scientific fact.
    - pasabagi 9y ago
      Absolutely. But equally, something being useful does not make it scientific. The thing about running is, it's very easy to explain what we mean by 'good at running'. It's very hard to describe what we mean by 'intelligence'. When you see a very large number of people convinced of the objectivity of something they can't describe, you normally call it a religion. Being convinced of the objectivity of IQ has many of the same flaws.
      - iguy 9y ago
        No it's not religion. It's statistical knowledge, in a way that people seem to struggle to grasp. (And there are thorns! What use it is put to is another issue.) If you look at all sports you'll find the same thing. We call the common factor physical fitness, or athleticism, or let's say "AQ". It's exactly the same idea. AQ predicts running speed, although not nearly as well as getting the kids out on the track. Ditto weightlifting. People know this: the kids in the chess club won't agree to arm-wrestle the kids on the track team, because despite having no direct information of their biceps strength, they can guess based on overall impressions. This strategy for keeping their lunch money is scientifically valid. And their profiling of the jocks is a safe distance from being morally repugnant.
        pasabagi 9y ago
        The problem with statistics is, if you put garbage in, you get garbage out. Especially in the case of garbage categories, and garbage questions. Before you do any statistical work, you have to ascertain whether or not the groups you are considering, or the questions you are asking, are actually existing groups and scientific questions. IQ fails this test. Race fails this test. Running does not. If make statistics about ill-defined, politicized questions, at best, you get nonsense. At worst, you get a reinforcement of your muddy, politicized thinking. The problem with science is, if you're not very careful, it can become a wrapper around stuff that isn't scientific. The only way to avoid this is to be absolutely allergic to any concepts that are unclear, uncertain imports from the normal world. And, while I'd love it if intelligence wasn't like that - let's face it, it is. If it wasn't, and we had a clear image of it, we'd probably be capable of doing stuff like building general AI.