“It is, however, proper application of Bayesian evidence.”
Nonsense.
If the only relevant pieces of information you had were the race of each man, and the average intelligence of each race, then of course it would be rational to estimate that the man from the ‘smarter’ race were the smarter of the two. But this is very far from the truth. In the Obama-Bush example, there is more than enough evidence on the public record to swamp any racially determined prior.
I think the principle of ‘treating people as individuals’ exists to combat a couple of things. One is the tendency to form stereotypes on flimsy or non-existent evidence, to over-estimate the generality and force of those stereotypes that are factually based, and to treat prejudice (i.e. group membership-based priors) as a substitute for even very easily-gathered and reliable evidence about the individual. The other is the direct emotional harm done to people by treating them as members of a group first, and individuals second (if at all). It is possible for this harm to outweigh the benefits of otherwise-rational discrimination.
Note: I honestly have no idea about the relationship between race and intelligence, so I deliberately set aside the question of who, if anyone, would have the higher prior in the Obama-Bush comparison. These aren’t politically correct weasel words; I would have a hard time properly defining intelligence, let alone measuring it in a culturally neutral way, but if I do see good evidence for a racial intelligence gap then I will readily accept it. All of this is beside the point of the dispute between TheAncientGreek and Eugine_Nier, though, for the reasons I gave above.
If the only relevant pieces of information you had were the race of each man, and the average intelligence of each race, then of course it would be rational to estimate that the man from the ‘smarter’ race were the smarter of the two.
Even then, assuming the difference between the averages is one standard deviation of either race’s distribution and each race’s distribution is Gaussian, there is only 76% probability that the smarter guy is the one from the smarter race, which hardly counts as “must” in my book.
If both are per-selected to be in the upper Nth percentile for their race, likely given the way affirmative action works, the probability is much higher.
But this is very far from the truth. In the Obama-Bush example, there is more than enough evidence on the public record to swamp any racially determined prior.
In principal, yes. In practice there is also so much noise put out by spin doctors that it might very well make sense to fall back on the prior.
Well, then let’s do a back of the envelope calculation based on statistics of today’s student bodies. Let’s stipulate that due to affirmative action, students of color are at the bottom of their class.
Harvard Law School’s class body is 40% of students of color. The 25 percentile LSAT score at Harvard Law is 170, and score 170 is 97.5th percentile among people who take the LSAT. So at least (40%-25%)/40% = 37.5% of the students of color at Harvard Law have a 97.5th percentile LSAT. Let’s assume that a black editor of the Harvard Law Review will be among those top students of color, so at least 25% percentile among HLS students. Based on his race, university and editorship, we estimate that Obama has a 97.5th percentile LSAT score or greater.
Compared to Harvard Law School, Harvard Business School relies less on standardized tests and enrolls fewer students of color (25% vs 40% for Harvard Law). George Bush, being white, is in the upper 75% section of the class. I know of no indication to suggest that he was an exceptional student, so I’ll put him in the middle of the white range (62.5 percentile). That corresponds to an GMAT score percentile of about 96 or 97 percent.
No, this is not a rigorous analysis, but it’s an improvement over a simple race prior (which I think is pretty crazy)
Nonsense.
If the only relevant pieces of information you had were the race of each man, and the average intelligence of each race, then of course it would be rational to estimate that the man from the ‘smarter’ race were the smarter of the two. But this is very far from the truth. In the Obama-Bush example, there is more than enough evidence on the public record to swamp any racially determined prior.
I think the principle of ‘treating people as individuals’ exists to combat a couple of things. One is the tendency to form stereotypes on flimsy or non-existent evidence, to over-estimate the generality and force of those stereotypes that are factually based, and to treat prejudice (i.e. group membership-based priors) as a substitute for even very easily-gathered and reliable evidence about the individual. The other is the direct emotional harm done to people by treating them as members of a group first, and individuals second (if at all). It is possible for this harm to outweigh the benefits of otherwise-rational discrimination.
Note: I honestly have no idea about the relationship between race and intelligence, so I deliberately set aside the question of who, if anyone, would have the higher prior in the Obama-Bush comparison. These aren’t politically correct weasel words; I would have a hard time properly defining intelligence, let alone measuring it in a culturally neutral way, but if I do see good evidence for a racial intelligence gap then I will readily accept it. All of this is beside the point of the dispute between TheAncientGreek and Eugine_Nier, though, for the reasons I gave above.
It’s statements like this which make me extremely skeptical of the idea that human rationality has increased over the last 100 years.
Even then, assuming the difference between the averages is one standard deviation of either race’s distribution and each race’s distribution is Gaussian, there is only 76% probability that the smarter guy is the one from the smarter race, which hardly counts as “must” in my book.
If both are per-selected to be in the upper Nth percentile for their race, likely given the way affirmative action works, the probability is much higher.
In principal, yes. In practice there is also so much noise put out by spin doctors that it might very well make sense to fall back on the prior.
Well, then let’s do a back of the envelope calculation based on statistics of today’s student bodies. Let’s stipulate that due to affirmative action, students of color are at the bottom of their class.
Harvard Law School’s class body is 40% of students of color. The 25 percentile LSAT score at Harvard Law is 170, and score 170 is 97.5th percentile among people who take the LSAT. So at least (40%-25%)/40% = 37.5% of the students of color at Harvard Law have a 97.5th percentile LSAT. Let’s assume that a black editor of the Harvard Law Review will be among those top students of color, so at least 25% percentile among HLS students. Based on his race, university and editorship, we estimate that Obama has a 97.5th percentile LSAT score or greater.
Compared to Harvard Law School, Harvard Business School relies less on standardized tests and enrolls fewer students of color (25% vs 40% for Harvard Law). George Bush, being white, is in the upper 75% section of the class. I know of no indication to suggest that he was an exceptional student, so I’ll put him in the middle of the white range (62.5 percentile). That corresponds to an GMAT score percentile of about 96 or 97 percent.
No, this is not a rigorous analysis, but it’s an improvement over a simple race prior (which I think is pretty crazy)