Someone who performs Bayesian inference that somehow involves probabilities conditioned on the race of a person or a group of people
The evidence someone’s race constitutes about that person’s qualities is usually very easily screened off, as I mentioned here. And given that we’re running on corrupted hardware, I suspect that someone who does try to “performs Bayesian inference that somehow involves probabilities conditioned on the race of a person” ends up subconsciously double-counting evidence and therefore end up with less accurate results than somebody who doesn’t. (As for cases when the evidence from race is not so easy to screen off… well, I’ve never heard anybody being accused of racism for pointing out that Africans have longer penises than Asians.)
I’m not talking about forensic evidence. Even if white people are smarter in average than black people, I think just talking with somebody for ten minutes would give me evidence about their intelligence which would nearly completely screen off that from skin colour. Heck, even just knowing what their job is would screen off much of it.
Even if white people are smarter in average than black people, I think just talking with somebody for ten minutes would give me evidence about their intelligence which would nearly completely screen off that from skin colour.
Also, as Eric Raymond discusses here, especially in the comments, you sometimes need to make judgements without spending ten minutes talking to everyone you see.
Heck, even just knowing what their job is would screen off much of it.
There’s this thing called Affirmative Action, as I mentioned elsewhere in this thread.
Also, as Eric Raymond discusses here, especially in the comments, you sometimes need to make judgements without spending ten minutes talking to everyone you see.
...
I do not require any “moral justification” for acting on the truth as it it really is; truth is its own warrant. (A comment by him).
I facepalmed. Really, Eric? Sorry, I don’t think that a moral realist is perceptive enough to the nuances and ethical knots involved to be a judge on this issue. I don’t know, he might be an excellent scientist, but it’s extremely stupid to be so rash when you’re attempting serious contrarianism.
But you reveal a confusion in your own thinking. It is not “treating other human beings as less-than-equal” to make rational decisions in risk situations; it is only that if you make decisions which are irrationally biased.
Yep, let’s all try to overcome bias really really hard; there’s only one solution, one desirable state, there’s a straight road ahead of us; Kingdom of Rationality, here we come!
(Yvain, thank you a million times for that sobering post!)
Also, as Eric Raymond discusses here, especially in the comments, you sometimes need to make judgements without spending ten minutes talking to everyone you see.
You know, there are countries where the intentional homicide rate is smaller than in John Derbyshire’s country by nearly an order of magnitude.
Heck, even just knowing what their job is would screen off much of it.
There’s this thing called Affirmative Action, as I mentioned elsewhere in this thread.
That thing doesn’t exist in all countries. Plus, I think the reason why you don’t see that many two-digit-IQ people among (say) physics professors is not that they don’t make it, it’s that they don’t even consider doing that, so even if some governmental policy somehow made it easier for black people with an IQ of 90 to succeed than for Jewish people with the same IQ, I would still expect a black physics professor to be smarter than (say) a Jewish truck driver.
so even if some governmental policy somehow made it easier for black people with an IQ of 90 to succeed than for Jewish people with the same IQ, I would still expect a black physics professor to be smarter than (say) a Jewish truck driver.
That’s not the point. The point is that the black physics professor is less smart than the Jewish physics professor.
But the difference is smaller than for the median black person and the median Jewish person. (I said “even just knowing what their job is would screen off much of it”, not “all of it”.)
The bell curve has both the mean and the deviation, you can have a ‘race’ with lower mean and larger standard deviation, and then you can e.g. filter by reliable accomplishment of some kind, such as solving some problem that smartest people in the world attempted and failed, you may end up with situation that the population with lower mean and larger standard deviation will have fewer people whom attain this, but those whom do, are on average smarter. Set bar even higher, and the population with lower mean and larger standard deviation has more people attaining it.
Also, the Gaussian distribution can stop being good approximation very far away from the mean.
edit: and to reply to grand grand parents: I bet i can divide the world into category that includes you, and a category that does not include you, in such a way that the category including you has substantially higher crime rate, or is otherwise bad. Actually if you are from US, I have a pretty natural ‘cultural’ category where your murder rate is about 5..10x of normal for such average income. Other category is the ‘racists’, i.e. the people whom use skin colour as evidence. Those people also have substantially bad behaviour. You of course want to use skin colour as evidence, and don’t want me to use your qualities as evidence. See if I care. If you want to use the skin colour as evidence, lumping together everyone that’s black, I want to use ‘use of skin colour as evidence’, lumping you together with all the nasty racists.
The bell curve has both the mean and the deviation, you can have a ‘race’ with lower mean and larger standard deviation
IIRC, no substantial difference was found in the standard deviations among races. (Whereas for genders, they have the same mean but males have larger sigma, so there are both more male idiots than female idiots and more male geniuses than female geniuses.)
Also, the Gaussian distribution can stop being good approximation very far away from the mean.
Isn’t IQ defined to be a Gaussian (e.g. IQ 160 just means ‘99.99683rd percentile among people your age’), rather than ‘whatever IQ tests measure’? If so, a better statement of that phenomenon would be “IQ tests are inaccurate for extreme values.”
See if I care. If you want to use the skin colour as evidence, lumping together everyone that’s black, I want to use ‘use of skin colour as evidence’, lumping you together with all the nasty racists.
I want to use ‘use of “use of skin colour as evidence” as evidence’ as evidence, but I’m not sure what that’s evidence for. :-)
IRC, no substantial difference was found in the standard deviations among races
Even a small difference translates into enormous ratio between numbers of people, several standard deviations from the mean...
Isn’t IQ defined to be a Gaussian
Yes, and it is defined to have specific standard deviation as well. That definition makes it unsuitable measure. The Gaussian distribution also arises from sum of multiple independent variables. The statement was about intelligence though, which is different thing from both “what IQ tests measure” and “how IQ is defined”.
Another huge failing of IQ is the non-measure of ability to build and use a huge search-able database of methods and facts. Building such database is a long-term memory task and can not be tested in short time span; the existing knowledge can’t be tested without massive influence by the background. Likewise, the IQ test lacks any problems that are actually difficult enough to have some solution methods that some people would know before the test, and some won’t.
Effectively, the IQ tests do not test for heavily parallel processing capability.
For example, I do believe that it would be possible to build ‘superhuman AI’ that runs on a cellphone and aces IQ tests, and could perhaps deceive a human in brief conversation. The same AI would never be able to invent a stone axe from scratch, let alone anything more complicated; it’d be nothing but a glorified calculator.
I want to use ‘use of “use of skin colour as evidence” as evidence’ as evidence, but I’m not sure what that’s evidence for. :-)
Well, the people who use skin colour as evidence, i would guess, are on average less well behaved than rest of society… so you can use it to guess someone’s criminality or other untrustworthiness.
Likewise, the IQ test lacks any problems that are actually difficult enough to have some solution methods that some people would know before the test, and some won’t.
Indeed, when I last took a few IQ tests I felt like I was being tested tested more for familiarity with concepts such as exclusiveOR, cyclical permutations, and similar basic discrete maths stuff than for processing power. (Of course, it does take insight to realize that such concepts are relevant to the questions and processing power to figure out the answer within the time frame of the test, but I think that if I had never heard about XOR or used Sarrus’ rule I would have scored much worse.)
ETA: This is also why I suspect that the correlations between race and IQ aren’t entirely genetic. If Einstein’s twin brother had grown up in a very poor region with no education...
Even a small difference translates into enormous ratio between numbers of people, several standard deviations from the mean...
A distribution with mean 100 and st. dev. 14 will exceed one with mean 90 and st. dev. 16 for all x between about 93 and about 170, and there aren’t that many people with IQs over 170 anyway.
But can we detect such a tiny difference as between std dev 14 and std dev 16 ? After we have to control for really many factors that are different between groups in question?
Also, that was my point, at the level of very high (one in million) intelligence, i.e. actual geniuses, the people you’d call genius without having to detect them using some test. I have a pet hypothesis about the last biological change which caused our technological progress. Little mixing with Neanderthals, raising the standard deviation somewhat.
The IQ test I think get useless past some point, when the IQ test savants that solve it at such level (but can’t learn very well for example, or can’t do problems well that require more of parallel processing), start to outnumber geniuses.
Well, the people who use skin colour as evidence, i would guess, are on average less well behaved than rest of society… so you can use it to guess someone’s criminality or other untrustworthiness.
You have the neonazis among those who use skin colour as evidence of criminality, but not among those who don’t. I don’t know of other differences that were demonstrated, my expectation for other effects is zero. I should expect the overall effect on order of at least the proportion of race motivated violence to overall violence; my expectation is somewhat higher than this though because I would guess that the near-neonazis are likewise more violent, including within-race crime.
I want to use ‘use of “use of skin colour as evidence” as evidence’ as evidence, but I’m not sure what that’s evidence for. :-)
Doh, missed the extra nesting. I doubt it’ll be evidence for much… both neonazis and liberal types use that as evidence, the former as evidence of ingroup-ness and the latter as evidence of badness, so I don’t see for what it would be discriminating.
I can’t remember whether I read this from someone else or came up with it on my own, but when people ask “do you oppose homosexual marriage” in questionnaires to find out political orientations, people answering “yes” will include both those who oppose homosexual marriage but are OK with heterosexual marriage, and those who oppose all marriage, and those groups are very different clusters in political space (paleo-conservatives the former, radical anarchists the latter). (Of course, the latter group is so much smaller than the former than if you’re doing statistics with large numbers of people this shouldn’t be such an issue.)
Even if white people are smarter in average than black people, I think just talking with somebody for ten minutes would give me evidence about their intelligence which would nearly completely screen off that from skin colour.
What if verbal ability and quantitative ability are often decoupled?
I wasn’t talking about “verbal ability” (which, to the extent that can be found out in ten minutes, correlates more with where someone grew up than with IQ), but about what they say, e.g. their reaction to finding out that I’m a physics student (though for this particular example there are lots of confounding factors), or what kinds of activities they enjoy.
If you’re able to drive the conversation like that, you can get information about IQ, and that information may have a larger impact than race. But to “screen off” evidence means making that evidence conditionally independent- once you knew their level of interest in physics, race would give you no information about their IQ. That isn’t the case.
Imagine that all races have Gaussian IQ distributions with the same standard deviation, but different means, and consider just the population of people whose IQs are above 132 (‘geniuses’ for this comment). In such a model, the mean IQ of black geniuses will be smaller than the mean IQ of white geniuses which will be smaller than the mean IQ of Jewish geniuses- so even knowing a lower bound for IQ won’t screen off the evidence provided by race!
Huh, sure, if the likelihood is a reversed Heaviside step. If the likelihood is itself a Gaussian, then the posterior is a Gaussian whose mean is the weighed average of that of the prior and that of the likelihood, weighed by the inverse squared standard deviations. So even if the st.dev. of the likelihood was half that of the prior for each race, the difference in posterior means would shrink by five times.
Right- there’s lots of information out there that will narrow your IQ estimate of someone else more than their race will, like that they’re a professional physicist or member of MENSA, but evidence only becomes worthless when it’s independent of the quantity you’re interested in given the other things you know.
You have a theory that a certain kind of building is highly prone to fire. You see a news report that mentions that a building of that kind has burnt down on Main Street. The news report supports your theory—unless you were a witness to the fire the previous night.
I’m talking about how valuable the evidence is to you, the theory-promoter. If you were there, then the news report tells you nothing you didn’t already know.
In this case, if the news report is consistent with my recollections, it seems that is evidence of the reliability of the news, and of the reliability of my memory, and additional evidence that the event actually occurred that way.
Yeah, true. But having been there the previous night, and making good observations the previous night, certainly makes the news report go from pretty strong evidence to almost nothing.
EDIT: Really the important thing I think, is that if your observations are good enough than the evidence from the news report is “worthless”, in the sense that you shouldn’t pay to find out whether there was a news report that backs up your observations. It’s not worth the time it takes to hear it..
Maybe I’m missing your point altogether, but it seems this is only true if the only thing I care about is the truth of that one theory of mine. If I also care about, for example, whether news reports are typically reliable, then suddenly the news report is worth a lot more.
Suppose A gives me information about B, and B gives me information about C; they’re dependent. (Remember, probabilistic dependence is always mutual.) A gives me information about C (through B) only if I don’t know B. If I know B, then A is conditionally independent of C, and so learning A tells me nothing about C.
(As for cases when the evidence from race is not so easy to screen off… well, I’ve never heard anybody being accused of racism for pointing out that Africans have longer penises than Asians.)
Minor note, this appears to actually not be the case. Most studies have no correlation between race and penis size. See for example here. The only group that there may be some substantial difference is that Chinese babies may have smaller genitalia after birth but this doesn’t appear to hold over to a significant difference by the time the children have reached puberty. Relevant study.
Huh, according to this map the average Congolese penis is nearly twice as long as the average South Korean penis. (ISTR that stretched flaccid length doesn’t perfectly correlate with erect length.)
The evidence someone’s race constitutes about that person’s qualities is usually very easily screened off, as I mentioned here. And given that we’re running on corrupted hardware, I suspect that someone who does try to “performs Bayesian inference that somehow involves probabilities conditioned on the race of a person” ends up subconsciously double-counting evidence and therefore end up with less accurate results than somebody who doesn’t. (As for cases when the evidence from race is not so easy to screen off… well, I’ve never heard anybody being accused of racism for pointing out that Africans have longer penises than Asians.)
I have seen accusations for racism as responses to people pointing that out.
Also, according to the U.S. Supreme Court even if race is screened off, you’re actions can still be racist or something.
In real life, you don’t have the luxury of gathering forensic evidence on everyone you meet.
I’m not talking about forensic evidence. Even if white people are smarter in average than black people, I think just talking with somebody for ten minutes would give me evidence about their intelligence which would nearly completely screen off that from skin colour. Heck, even just knowing what their job is would screen off much of it.
Also, as Eric Raymond discusses here, especially in the comments, you sometimes need to make judgements without spending ten minutes talking to everyone you see.
There’s this thing called Affirmative Action, as I mentioned elsewhere in this thread.
...
I facepalmed. Really, Eric? Sorry, I don’t think that a moral realist is perceptive enough to the nuances and ethical knots involved to be a judge on this issue. I don’t know, he might be an excellent scientist, but it’s extremely stupid to be so rash when you’re attempting serious contrarianism.
Yep, let’s all try to overcome bias really really hard; there’s only one solution, one desirable state, there’s a straight road ahead of us; Kingdom of Rationality, here we come!
(Yvain, thank you a million times for that sobering post!)
You know, there are countries where the intentional homicide rate is smaller than in John Derbyshire’s country by nearly an order of magnitude.
That thing doesn’t exist in all countries. Plus, I think the reason why you don’t see that many two-digit-IQ people among (say) physics professors is not that they don’t make it, it’s that they don’t even consider doing that, so even if some governmental policy somehow made it easier for black people with an IQ of 90 to succeed than for Jewish people with the same IQ, I would still expect a black physics professor to be smarter than (say) a Jewish truck driver.
That’s not the point. The point is that the black physics professor is less smart than the Jewish physics professor.
But the difference is smaller than for the median black person and the median Jewish person. (I said “even just knowing what their job is would screen off much of it”, not “all of it”.)
The bell curve has both the mean and the deviation, you can have a ‘race’ with lower mean and larger standard deviation, and then you can e.g. filter by reliable accomplishment of some kind, such as solving some problem that smartest people in the world attempted and failed, you may end up with situation that the population with lower mean and larger standard deviation will have fewer people whom attain this, but those whom do, are on average smarter. Set bar even higher, and the population with lower mean and larger standard deviation has more people attaining it. Also, the Gaussian distribution can stop being good approximation very far away from the mean.
edit: and to reply to grand grand parents: I bet i can divide the world into category that includes you, and a category that does not include you, in such a way that the category including you has substantially higher crime rate, or is otherwise bad. Actually if you are from US, I have a pretty natural ‘cultural’ category where your murder rate is about 5..10x of normal for such average income. Other category is the ‘racists’, i.e. the people whom use skin colour as evidence. Those people also have substantially bad behaviour. You of course want to use skin colour as evidence, and don’t want me to use your qualities as evidence. See if I care. If you want to use the skin colour as evidence, lumping together everyone that’s black, I want to use ‘use of skin colour as evidence’, lumping you together with all the nasty racists.
IIRC, no substantial difference was found in the standard deviations among races. (Whereas for genders, they have the same mean but males have larger sigma, so there are both more male idiots than female idiots and more male geniuses than female geniuses.)
Isn’t IQ defined to be a Gaussian (e.g. IQ 160 just means ‘99.99683rd percentile among people your age’), rather than ‘whatever IQ tests measure’? If so, a better statement of that phenomenon would be “IQ tests are inaccurate for extreme values.”
I want to use ‘use of “use of skin colour as evidence” as evidence’ as evidence, but I’m not sure what that’s evidence for. :-)
Even a small difference translates into enormous ratio between numbers of people, several standard deviations from the mean...
Yes, and it is defined to have specific standard deviation as well. That definition makes it unsuitable measure. The Gaussian distribution also arises from sum of multiple independent variables. The statement was about intelligence though, which is different thing from both “what IQ tests measure” and “how IQ is defined”.
Another huge failing of IQ is the non-measure of ability to build and use a huge search-able database of methods and facts. Building such database is a long-term memory task and can not be tested in short time span; the existing knowledge can’t be tested without massive influence by the background. Likewise, the IQ test lacks any problems that are actually difficult enough to have some solution methods that some people would know before the test, and some won’t.
Effectively, the IQ tests do not test for heavily parallel processing capability.
For example, I do believe that it would be possible to build ‘superhuman AI’ that runs on a cellphone and aces IQ tests, and could perhaps deceive a human in brief conversation. The same AI would never be able to invent a stone axe from scratch, let alone anything more complicated; it’d be nothing but a glorified calculator.
Well, the people who use skin colour as evidence, i would guess, are on average less well behaved than rest of society… so you can use it to guess someone’s criminality or other untrustworthiness.
Indeed, when I last took a few IQ tests I felt like I was being tested tested more for familiarity with concepts such as exclusiveOR, cyclical permutations, and similar basic discrete maths stuff than for processing power. (Of course, it does take insight to realize that such concepts are relevant to the questions and processing power to figure out the answer within the time frame of the test, but I think that if I had never heard about XOR or used Sarrus’ rule I would have scored much worse.)
ETA: This is also why I suspect that the correlations between race and IQ aren’t entirely genetic. If Einstein’s twin brother had grown up in a very poor region with no education...
A distribution with mean 100 and st. dev. 14 will exceed one with mean 90 and st. dev. 16 for all x between about 93 and about 170, and there aren’t that many people with IQs over 170 anyway.
But can we detect such a tiny difference as between std dev 14 and std dev 16 ? After we have to control for really many factors that are different between groups in question?
Also, that was my point, at the level of very high (one in million) intelligence, i.e. actual geniuses, the people you’d call genius without having to detect them using some test. I have a pet hypothesis about the last biological change which caused our technological progress. Little mixing with Neanderthals, raising the standard deviation somewhat.
The IQ test I think get useless past some point, when the IQ test savants that solve it at such level (but can’t learn very well for example, or can’t do problems well that require more of parallel processing), start to outnumber geniuses.
What sort of effect size do you expect here? Why?
You have the neonazis among those who use skin colour as evidence of criminality, but not among those who don’t. I don’t know of other differences that were demonstrated, my expectation for other effects is zero. I should expect the overall effect on order of at least the proportion of race motivated violence to overall violence; my expectation is somewhat higher than this though because I would guess that the near-neonazis are likewise more violent, including within-race crime.
Doh, missed the extra nesting. I doubt it’ll be evidence for much… both neonazis and liberal types use that as evidence, the former as evidence of ingroup-ness and the latter as evidence of badness, so I don’t see for what it would be discriminating.
I can’t remember whether I read this from someone else or came up with it on my own, but when people ask “do you oppose homosexual marriage” in questionnaires to find out political orientations, people answering “yes” will include both those who oppose homosexual marriage but are OK with heterosexual marriage, and those who oppose all marriage, and those groups are very different clusters in political space (paleo-conservatives the former, radical anarchists the latter). (Of course, the latter group is so much smaller than the former than if you’re doing statistics with large numbers of people this shouldn’t be such an issue.)
What if verbal ability and quantitative ability are often decoupled?
I wasn’t talking about “verbal ability” (which, to the extent that can be found out in ten minutes, correlates more with where someone grew up than with IQ), but about what they say, e.g. their reaction to finding out that I’m a physics student (though for this particular example there are lots of confounding factors), or what kinds of activities they enjoy.
If you’re able to drive the conversation like that, you can get information about IQ, and that information may have a larger impact than race. But to “screen off” evidence means making that evidence conditionally independent- once you knew their level of interest in physics, race would give you no information about their IQ. That isn’t the case.
Imagine that all races have Gaussian IQ distributions with the same standard deviation, but different means, and consider just the population of people whose IQs are above 132 (‘geniuses’ for this comment). In such a model, the mean IQ of black geniuses will be smaller than the mean IQ of white geniuses which will be smaller than the mean IQ of Jewish geniuses- so even knowing a lower bound for IQ won’t screen off the evidence provided by race!
Huh, sure, if the likelihood is a reversed Heaviside step. If the likelihood is itself a Gaussian, then the posterior is a Gaussian whose mean is the weighed average of that of the prior and that of the likelihood, weighed by the inverse squared standard deviations. So even if the st.dev. of the likelihood was half that of the prior for each race, the difference in posterior means would shrink by five times.
Right- there’s lots of information out there that will narrow your IQ estimate of someone else more than their race will, like that they’re a professional physicist or member of MENSA, but evidence only becomes worthless when it’s independent of the quantity you’re interested in given the other things you know.
Can you give an example of evidence becoming worthless? (I can’t think of any.)
You have a theory that a certain kind of building is highly prone to fire. You see a news report that mentions that a building of that kind has burnt down on Main Street. The news report supports your theory—unless you were a witness to the fire the previous night.
If you were promoting the theory before that point, the police may still have some pointed questions to ask you.
I’m talking about how valuable the evidence is to you, the theory-promoter. If you were there, then the news report tells you nothing you didn’t already know.
I understood your point. I was simply making a joke.
In this case, if the news report is consistent with my recollections, it seems that is evidence of the reliability of the news, and of the reliability of my memory, and additional evidence that the event actually occurred that way.
No?
Yeah, true. But having been there the previous night, and making good observations the previous night, certainly makes the news report go from pretty strong evidence to almost nothing.
EDIT: Really the important thing I think, is that if your observations are good enough than the evidence from the news report is “worthless”, in the sense that you shouldn’t pay to find out whether there was a news report that backs up your observations. It’s not worth the time it takes to hear it..
Hm.
Maybe I’m missing your point altogether, but it seems this is only true if the only thing I care about is the truth of that one theory of mine. If I also care about, for example, whether news reports are typically reliable, then suddenly the news report is worth a lot more.
But, sure, given that premise, I agree.
Suppose A gives me information about B, and B gives me information about C; they’re dependent. (Remember, probabilistic dependence is always mutual.) A gives me information about C (through B) only if I don’t know B. If I know B, then A is conditionally independent of C, and so learning A tells me nothing about C.
So essentially… a new fact is useless only if it’s a subset of knowledge you already have?
That seems like a fine way to put it.
Minor note, this appears to actually not be the case. Most studies have no correlation between race and penis size. See for example here. The only group that there may be some substantial difference is that Chinese babies may have smaller genitalia after birth but this doesn’t appear to hold over to a significant difference by the time the children have reached puberty. Relevant study.
Huh, according to this map the average Congolese penis is nearly twice as long as the average South Korean penis. (ISTR that stretched flaccid length doesn’t perfectly correlate with erect length.)
Oddly salient for such a trivial result. Should a study qualify for an Ig Nobel if you can use it to settle bar bets?