I don’t understand what you mean by “matter.” People don’t care about hair color because hair color is not very predictive of other traits that people care about, but this doesn’t seem to be true of race.
What traits, aside from skin colour and immunity or vulnerability to sunburn, are strongly correlated with race and cared about in more than an aesthetic sense?
That depends on what you mean by “strongly.” I would tentatively posit that even if race isn’t strongly predictive in an absolute sense of other traits that people care about, it is relatively predictive compared to other traits that are easy to unambiguously learn about a person. For example, if I wanted to predict the performance of a high school student on standardized tests, I think race would be a better predictor than height or weight, and I don’t know enough to confidently say whether it would a better predictor than income level.
I’ve recently begun to suspect that a possibly substantial amount of what gets labeled “racism” is just using race as weak Bayesian evidence in the spirit of http://lesswrong.com/lw/aq2/fallacies_as_weak_bayesian_evidence/ (edit: and then subsequently failing to distinguish between the probability of a statement being true having increased and the statement becoming true).
Hmmm. It seems to me that what is happening here is that race is reasonably correlated with culture, and culture is very strongly correlated with upbringing, and upbringing is very strongly correlated with academic performance. (Note that income level->culture is also a fairly strong correlation).
Race is also highly visible, and (often, but not always) easily discerned. Hence, a correlation (via culture) between race and academic performance would be very visible.
If the correlation between race and culture is thus dissolved, or at least dramatically reduced, then race will become far weaker evidence as to (say) academic performance, eventually dipping below random noise levels. Once the correlation between race and non-aesthetic traits that people care about is generally recognised as being below the level of random noise, then I would say that race will no longer matter.
(Culture, of course, will still matter. I don’t really see any good way around that).
Why does it matter to you how strong the correlation between race and culture is? Isn’t the real problem that people are mishandling Bayesian updates based on race? That could be solved by teaching people how to perform Bayesian updates more accurately. It wouldn’t be a world in which “race doesn’t matter,” but it would be a world in which the extent to which race does matter is recognized and not exaggerated or ignored.
I can think of at least two other causal paths from race to academic performance. One is the attitudes a person’s peer group is likely to hold towards academic performance (even if they don’t make a point of affiliating with other people based on race, other people may make a point of affiliating with them based on race), and more generally how the people around a person treat them based on race. The other is genetics. (I imagine this is not a particularly popular thing to say but I recently realized that I do not have a solid statistical foundation for dismissing it.)
I disagree. Many statistical effects of race are screened off by fairly easily obtained information, but people act as though this is not the case. Moreover, if you, say, beat someone for being black, that’s really not tied to any sort of problem with your use of Bayesian updating.
Some such information is degraded, yes, but not all, and not to uselessness. And yes, people are beaten in the first world in this day and age for being black or for being white, and I find it difficult to blame either of those on the use or misuse of Bayesian updating (except to the extent that observing a person’s race might tell you “I can get away with this”).
I do not accept your contention that people just happen to be exactly the correct degree of racist.
I do not accept your contention that people just happen to be exactly the correct degree of racist.
People are usually not “exactly correct” about anything, so statements like this are almost automatically true. But is this your true rejection?
Imagine that tomorrow some magic will turn all people into exactly the correct degree of racists. That means for example that if a person with a given skin color has (according to the external view) probability X to have some trait, they will expect that trait with probability exactly X, not more, not less.
Would such society be more similar to what we have now, or to a perfectly equal society?
I dunno what that society would be more similar to. I expect it’d be a fair distance from either, and that there would remain significant problems apart from inequality of social status, economic status, etc. Eugine_Nier’s assertion was that it would be identical (read: very similar) to what we have now. I disagreed.
Just for the record, my estimate is that it would be cca 70% as much “racist” as what we have today. (I don’t have a high confidence in this number, I just though it would be fair to write my opinion if I am asking about yours.) So cca 30% of the racism can be explained by people updating incorrectly, but that still leaves the remaining 70% to be explained otherwise. Therefore focusing on the incorrect updates misses the greater part of the whole story.
I think that affirmative action hurts both ways. And it also keeps the feeling of resentment alive, which again hurts people.
As a simple example, in my country most people in IT are male. So on one hand you have the “prejudice” that women in general are not good with computers, but on the other hand, if you meet a female programmer, you know that she specifically is good enough. She passed the filter.
I imagine that in an alternative reality where IT companies would be legally required to have 50% female programmers, the “prejudice” would expand, and it would say that women programmers are not good with computers. A female programmer would have to work harder to pass the filter. Even participating in a successful project would not be enough, because others would think that the males in her team did most of the work, and she was there mostly for political reasons. To prove herself, she would have to win some programming competition (and tell everyone about it). But those who can do it, they have no problem finding a programming job in our world, too.
Affirmative action would work best if you could legislate it and make everyone forget that it exists. Perhaps legislating it and making taboo of discussing it openly, is a step in this direction. Still, if the differences in abilities are real, people will notice the result, even if they are not informed about the causes.
In the alternative reality where IT companies are legally required to have 50% women programmers, and the law is successfully kept secret from everyone except the HR departments, programmers would still notice the differences in their colleagues’ skills. Although… this knowledge would exist only among the programmers, because only they see it firsthand. You could still convince the public that what the programmers see is not real, that it is merely their sexism.
So now I think that social enginnerings of this kind are successful only if people are prevented from discussing them openly. Even a lie told with good intentions makes the truth forever your enemy. Of course that makes it difficult to evaluate whether the policy really helps or not.
I’d have expected affirmative action to have substantial ill effects, but no one seems to be saying that the quality of American goods has dropped noticeably since the late sixties.
My tentative explanation is that hiring and promotion are much more random than people want to think.
Who’s “focusing”? I would argue, if we take your numbers, that the incorrect 30% are disproportionately problematic compared to the remaining 70%, and that there are other, non-epistemic problems involved in racism. Eugine_Nier said that “the problem” is the 70%. That’s the disagreement that’s going on here. My claim is not that modern-day racism is on average a greater distortion of the facts than an inability to perceive race would be.
There are at least two confounding factors for the crime statistics. One is that the justice system is pretty sloppy, and more so for black men. Another is that even if your crime statistics are accurate, it’s hard to identify a criminal’s exact motives. Was a beating part of a robbery? Was it a simple attack initiated by one side, or was it a quarrel that escalated?
Another possibility is that race affects how many people are treated in the educational system, and that affects how much effort they put into schoolwork.
My cousin is of mixed ethnicity (black father and white mother), and if half of what he says is true and not just teenaged exaggeration, a good chunk of his disciplinary record at school is probably (I’d assign over 70% assuming he’s completely truthful) based on race, and nothing he does. He isn’t as interested in academics as my sister or I were, but the only actual academic losses I’ve noticed were in his first quarter of mathematics in eighth grade (he wound up in the most advanced math class available, which he wasn’t particularly thrilled about, and it was a new teacher and a new curriculum and the entire class was left in the dust for a few weeks).
Also, black people are usually not in such high academic standing as he is, and when I was his age, in the same school, I heard people talk about perceived racism from teachers toward the black minority that were in the honors/AP/etc classes.
All anecdotal hearsay, but it’s strong enough evidence for me that I tend to agree with the idea that race correlates with intelligence and crime because the culture expects it to more than because of genetic reasons.
[edit] Oh, I’ll also add that my evaluation of the likelyhood that my cousin is being completely honest in his accounts is only slightly above 50% at this point. He’s way more honest than his younger brother (who is a pathological liar caught in a self-enforcing death-spiral (and they have different fathers—the younger one’s father is white)), but is no stranger to trolling, and even when he’s speaking truthfully his accounts might be muddled in bias. But a good number of them seem hard to interpret as anything but consistent unfair treatment in a context where what sticks out about him is race. He did not offer the explanation of racism, though; that was my conclusion after a dozen or so separate incidents.[/edit]
Why does it matter to you how strong the correlation between race and culture is?
I don’t care about the correlation between race and culture in and of itself. I want to remove or reduce (preferably remove) the percieved correlation between race and academic performance; and it seems to me that the best way to do this is to remove the correlation between race and culture (as the correlation from culture to academic performance does not look removable).
Isn’t the real problem that people are mishandling Bayesian updates based on race? That could be solved by teaching people how to perform Bayesian updates more accurately. It wouldn’t be a world in which “race doesn’t matter,” but it would be a world in which the extent to which race does matter is recognized and not exaggerated or ignored.
That is a good strategy, and quite possibly superior to my suggestion. The biggest trouble is that it requires a substantial majority of people to be willing to learn how to properly perform Bayesian updates, which I fear may make it less practical. (Not that my idea was necessarily all that practical to begin with).
I can think of at least two other causal paths from race to academic performance. One is the attitudes a person’s peer group is likely to hold towards academic performance (even if they don’t make a point of affiliating with other people based on race, other people may make a point of affiliating with them based on race), and more generally how the people around a person treat them based on race.
Hmmm. This is a possible path; intuitively, I’d expect it to matter about as much as the neighbourhood one grows up in. That is, I would expect any non-cultural effects to be more or less random noise.
The other is genetics. (I imagine this is not a particularly popular thing to say but I recently realized that I do not have a solid statistical foundation for dismissing it.)
That is also possible. Intuitively (which is very poor evidence, I know) I would expect this to matter less than culture. I do know some very intelligent people of many races; so individual variance seems large enough to defeat any systemic genetic bias that may exist.
Experimental evidence of the effects of culture versus genetics could be discovered by studying people of one race raised in the culture of another race (e.g. by adoption).
I don’t care about the correlation between race and culture in and of itself. I want to remove or reduce (preferably remove) the percieved correlation between race and academic performance
I think a better strategy is to remove the actual correlation between race and academic performance, and possibly the one between race and criminality for that matter.
One place to start is to change the culture that leads to said problems.
I think a better strategy is to remove the actual correlation between race and academic performance, and possibly the one between race and criminality for that matter.
That is a necessary prerequisite, yes. As long as such an actual correlation is in place, it will be observed and will result in a perceived correlation.
What traits, aside from skin colour and immunity or vulnerability to sunburn, are strongly correlated with race and cared about in more than an aesthetic sense?
Intelligence and criminality, to give the two most important examples.
I’d be interested to see a citation for the intelligence claim. I could believe a very weak correlation to genetics, but find a strong one unlikely.
There may be a strong correlation to intelligence via culture; which implies that some cultures are flawed, holding people back from achieving what they might in a better culture; implying in turn that flawed cultures should be improved/debugged.
criminality
Citation?
Again, I suspect—though I’m not certain—that what we have here is a cultural tendency pretending to be a racial tendency. If that is correct, then a member of the wrong race faces severe and unfair disadvantages even if he belongs to a less-criminality-inclined culture.
It sounds like you’re using the word “correlation” to refer to different modes of causation, which is potentially confusing; “correlation” just refers to certain kinds of association.
It’s trivial to dig up citations for correlations between race & IQ. Distinguishing between the two causal models of racial genetic differences → IQ and racial genetic differences ↔ culture → IQ, which I think is what you’re getting at, is a distinct and more vexed issue. Still, the first citation in that Wikipedia article is of a paper that clearly favours the first model over the second:
The hereditarian model of Black–White IQ differences proposed in Section 2 (50% genetic and 50% environmental), far from precluding environmental factors, requires they be found. Although evidence in Sections 3 to 11 provided strong support for the genetic component of the model, evidence from Section 12 was unable to identify the environmental component. On the basis of the present evidence, perhaps the genetic component must be given greater weight and the environmental component correspondingly reduced.
As it happens, I find this particular paper flawed in various ways, but it is a citation of the sort you’re asking for.
I don’t understand what you mean by “matter.” People don’t care about hair color because hair color is not very predictive of other traits that people care about, but this doesn’t seem to be true of race.
What traits, aside from skin colour and immunity or vulnerability to sunburn, are strongly correlated with race and cared about in more than an aesthetic sense?
That depends on what you mean by “strongly.” I would tentatively posit that even if race isn’t strongly predictive in an absolute sense of other traits that people care about, it is relatively predictive compared to other traits that are easy to unambiguously learn about a person. For example, if I wanted to predict the performance of a high school student on standardized tests, I think race would be a better predictor than height or weight, and I don’t know enough to confidently say whether it would a better predictor than income level.
I’ve recently begun to suspect that a possibly substantial amount of what gets labeled “racism” is just using race as weak Bayesian evidence in the spirit of http://lesswrong.com/lw/aq2/fallacies_as_weak_bayesian_evidence/ (edit: and then subsequently failing to distinguish between the probability of a statement being true having increased and the statement becoming true).
Hmmm. It seems to me that what is happening here is that race is reasonably correlated with culture, and culture is very strongly correlated with upbringing, and upbringing is very strongly correlated with academic performance. (Note that income level->culture is also a fairly strong correlation).
Race is also highly visible, and (often, but not always) easily discerned. Hence, a correlation (via culture) between race and academic performance would be very visible.
If the correlation between race and culture is thus dissolved, or at least dramatically reduced, then race will become far weaker evidence as to (say) academic performance, eventually dipping below random noise levels. Once the correlation between race and non-aesthetic traits that people care about is generally recognised as being below the level of random noise, then I would say that race will no longer matter.
(Culture, of course, will still matter. I don’t really see any good way around that).
Why does it matter to you how strong the correlation between race and culture is? Isn’t the real problem that people are mishandling Bayesian updates based on race? That could be solved by teaching people how to perform Bayesian updates more accurately. It wouldn’t be a world in which “race doesn’t matter,” but it would be a world in which the extent to which race does matter is recognized and not exaggerated or ignored.
I can think of at least two other causal paths from race to academic performance. One is the attitudes a person’s peer group is likely to hold towards academic performance (even if they don’t make a point of affiliating with other people based on race, other people may make a point of affiliating with them based on race), and more generally how the people around a person treat them based on race. The other is genetics. (I imagine this is not a particularly popular thing to say but I recently realized that I do not have a solid statistical foundation for dismissing it.)
At this point I think the problem is that they are updating correctly.
I disagree. Many statistical effects of race are screened off by fairly easily obtained information, but people act as though this is not the case. Moreover, if you, say, beat someone for being black, that’s really not tied to any sort of problem with your use of Bayesian updating.
Or would be if people weren’t actively rigging said information such that this is not the case. And that’s before getting into tail-effects.
Which really doesn’t happen these days. (It’s certainly much rarer than someone being beaten up for being white.)
Some such information is degraded, yes, but not all, and not to uselessness. And yes, people are beaten in the first world in this day and age for being black or for being white, and I find it difficult to blame either of those on the use or misuse of Bayesian updating (except to the extent that observing a person’s race might tell you “I can get away with this”).
I do not accept your contention that people just happen to be exactly the correct degree of racist.
People are usually not “exactly correct” about anything, so statements like this are almost automatically true. But is this your true rejection?
Imagine that tomorrow some magic will turn all people into exactly the correct degree of racists. That means for example that if a person with a given skin color has (according to the external view) probability X to have some trait, they will expect that trait with probability exactly X, not more, not less.
Would such society be more similar to what we have now, or to a perfectly equal society?
I’d bet on closer to a perfectly equal society, but it’s rather hard to do the experiment.
It’s certainly my (a) true rejection of “the problem is that [people] are updating correctly”. What did you expect I was rejecting?
I dunno what that society would be more similar to. I expect it’d be a fair distance from either, and that there would remain significant problems apart from inequality of social status, economic status, etc. Eugine_Nier’s assertion was that it would be identical (read: very similar) to what we have now. I disagreed.
I confess, I was sacrificing some precision for snark. I meant “the problem is that [people] are updating correctly, to the extant they are”.
Just for the record, my estimate is that it would be cca 70% as much “racist” as what we have today. (I don’t have a high confidence in this number, I just though it would be fair to write my opinion if I am asking about yours.) So cca 30% of the racism can be explained by people updating incorrectly, but that still leaves the remaining 70% to be explained otherwise. Therefore focusing on the incorrect updates misses the greater part of the whole story.
Really? I’d estimate more like 120%.
Edit: especially consider affirmative action and the desperate impact doctrine.
I think that affirmative action hurts both ways. And it also keeps the feeling of resentment alive, which again hurts people.
As a simple example, in my country most people in IT are male. So on one hand you have the “prejudice” that women in general are not good with computers, but on the other hand, if you meet a female programmer, you know that she specifically is good enough. She passed the filter.
I imagine that in an alternative reality where IT companies would be legally required to have 50% female programmers, the “prejudice” would expand, and it would say that women programmers are not good with computers. A female programmer would have to work harder to pass the filter. Even participating in a successful project would not be enough, because others would think that the males in her team did most of the work, and she was there mostly for political reasons. To prove herself, she would have to win some programming competition (and tell everyone about it). But those who can do it, they have no problem finding a programming job in our world, too.
Affirmative action would work best if you could legislate it and make everyone forget that it exists. Perhaps legislating it and making taboo of discussing it openly, is a step in this direction. Still, if the differences in abilities are real, people will notice the result, even if they are not informed about the causes.
In the alternative reality where IT companies are legally required to have 50% women programmers, and the law is successfully kept secret from everyone except the HR departments, programmers would still notice the differences in their colleagues’ skills. Although… this knowledge would exist only among the programmers, because only they see it firsthand. You could still convince the public that what the programmers see is not real, that it is merely their sexism.
So now I think that social enginnerings of this kind are successful only if people are prevented from discussing them openly. Even a lie told with good intentions makes the truth forever your enemy. Of course that makes it difficult to evaluate whether the policy really helps or not.
I’d have expected affirmative action to have substantial ill effects, but no one seems to be saying that the quality of American goods has dropped noticeably since the late sixties.
My tentative explanation is that hiring and promotion are much more random than people want to think.
Well two points:
1) There is a huge confounding factor, namely technological progress.
2) In general, labor intensive goods aren’t even produced in America anymore.
Who’s “focusing”? I would argue, if we take your numbers, that the incorrect 30% are disproportionately problematic compared to the remaining 70%, and that there are other, non-epistemic problems involved in racism. Eugine_Nier said that “the problem” is the 70%. That’s the disagreement that’s going on here. My claim is not that modern-day racism is on average a greater distortion of the facts than an inability to perceive race would be.
Taboo “perfectly equal society”.
Evidence? Also, are you including assault by the police in your comparison?
Look at crime statistics.
Sure, it doesn’t change its truth.
There are at least two confounding factors for the crime statistics. One is that the justice system is pretty sloppy, and more so for black men. Another is that even if your crime statistics are accurate, it’s hard to identify a criminal’s exact motives. Was a beating part of a robbery? Was it a simple attack initiated by one side, or was it a quarrel that escalated?
I don’t think this is a valuable response to being asked for evidence.
Another possibility is that race affects how many people are treated in the educational system, and that affects how much effort they put into schoolwork.
My cousin is of mixed ethnicity (black father and white mother), and if half of what he says is true and not just teenaged exaggeration, a good chunk of his disciplinary record at school is probably (I’d assign over 70% assuming he’s completely truthful) based on race, and nothing he does. He isn’t as interested in academics as my sister or I were, but the only actual academic losses I’ve noticed were in his first quarter of mathematics in eighth grade (he wound up in the most advanced math class available, which he wasn’t particularly thrilled about, and it was a new teacher and a new curriculum and the entire class was left in the dust for a few weeks).
Also, black people are usually not in such high academic standing as he is, and when I was his age, in the same school, I heard people talk about perceived racism from teachers toward the black minority that were in the honors/AP/etc classes.
All anecdotal hearsay, but it’s strong enough evidence for me that I tend to agree with the idea that race correlates with intelligence and crime because the culture expects it to more than because of genetic reasons.
[edit] Oh, I’ll also add that my evaluation of the likelyhood that my cousin is being completely honest in his accounts is only slightly above 50% at this point. He’s way more honest than his younger brother (who is a pathological liar caught in a self-enforcing death-spiral (and they have different fathers—the younger one’s father is white)), but is no stranger to trolling, and even when he’s speaking truthfully his accounts might be muddled in bias. But a good number of them seem hard to interpret as anything but consistent unfair treatment in a context where what sticks out about him is race. He did not offer the explanation of racism, though; that was my conclusion after a dozen or so separate incidents.[/edit]
I’d say that this is another very strong possibility.
I don’t care about the correlation between race and culture in and of itself. I want to remove or reduce (preferably remove) the percieved correlation between race and academic performance; and it seems to me that the best way to do this is to remove the correlation between race and culture (as the correlation from culture to academic performance does not look removable).
That is a good strategy, and quite possibly superior to my suggestion. The biggest trouble is that it requires a substantial majority of people to be willing to learn how to properly perform Bayesian updates, which I fear may make it less practical. (Not that my idea was necessarily all that practical to begin with).
Hmmm. This is a possible path; intuitively, I’d expect it to matter about as much as the neighbourhood one grows up in. That is, I would expect any non-cultural effects to be more or less random noise.
That is also possible. Intuitively (which is very poor evidence, I know) I would expect this to matter less than culture. I do know some very intelligent people of many races; so individual variance seems large enough to defeat any systemic genetic bias that may exist.
Experimental evidence of the effects of culture versus genetics could be discovered by studying people of one race raised in the culture of another race (e.g. by adoption).
I think a better strategy is to remove the actual correlation between race and academic performance, and possibly the one between race and criminality for that matter.
One place to start is to change the culture that leads to said problems.
That is a necessary prerequisite, yes. As long as such an actual correlation is in place, it will be observed and will result in a perceived correlation.
Intelligence and criminality, to give the two most important examples.
I’d be interested to see a citation for the intelligence claim. I could believe a very weak correlation to genetics, but find a strong one unlikely.
There may be a strong correlation to intelligence via culture; which implies that some cultures are flawed, holding people back from achieving what they might in a better culture; implying in turn that flawed cultures should be improved/debugged.
Citation?
Again, I suspect—though I’m not certain—that what we have here is a cultural tendency pretending to be a racial tendency. If that is correct, then a member of the wrong race faces severe and unfair disadvantages even if he belongs to a less-criminality-inclined culture.
I never said anything about causation or genetics. I was just talking about correlation.
It sounds like you’re using the word “correlation” to refer to different modes of causation, which is potentially confusing; “correlation” just refers to certain kinds of association.
It’s trivial to dig up citations for correlations between race & IQ. Distinguishing between the two causal models of racial genetic differences → IQ and racial genetic differences ↔ culture → IQ, which I think is what you’re getting at, is a distinct and more vexed issue. Still, the first citation in that Wikipedia article is of a paper that clearly favours the first model over the second:
As it happens, I find this particular paper flawed in various ways, but it is a citation of the sort you’re asking for.
Thank you, that was exactly the sort of citation I was asking for.