Only then could propensity to crime be an inherent, genetic thing. And the inference that that was the case may still be wrong, for example if black people are bombarded by messages that they are supposed to be become criminals, or are otherwise influenced by the people around them. It would be very difficult to seperate the inherent genetic traits from those that are caused by percieved race.

I think childhood role models and so on is a part of one's upbringing and "society", don't you?

Let's start with something simple that I hope we can agree on. Group people by genetic heritage, and by social class, and by intelligence, and by antisocial behaviour. "Genetic heritage" is clear enough, I hope. "Class" is rather slippery in this analysis unless we are careful with it. Let's pin it down right now to be talking about environment, not where a person ends up. We'll see why later. Anyways, if you make these groupings, you will find that there is a lot of mutual information between them. That is, they are not independent. If you don't believe this, assume it for now.

Perhaps we are talking past each other. I'm not claiming you can't get any information from someone's race, I'm saying that this is due to historical/memetic causes. It's the differenc between loaded dice and an opponent who regularly lies about the results, if you see what I mean.

At this point, I hope I've said nothing controversial. The redefinition of "class" rubs me the wrong way, but I couldn't think of anything else to call that node. Ok, let's move on to the implications.

"Upbringing"? "Background"? I'm OK with class, TBH, as long as we both know what we mean.

Let's boot up the racist and see what he says about all this. The racist says "I don't care which of the two it is, and here's why:". Uh oh, here we go.

Here we go indeed. The racist is supposed to hold a belief abut how he world is, i.e. that intelligence and so on are as much racial characteristics as skin tone.

Let's do a little thought experiment: group someone's genes by those that define who they are as a person, and those that define what they look like. Let's say there's no overlap between these, that is, that a given gene cannot both impact personhood and appearance. (there's reasons to suspect overlap, but this is a thought experiment). Let's further say that, for obvious reasons (subpopulations), having gene A, which affects appearance, is quite strongly corellated with having gene alpha, which affects personality. Extend this to most of the genes so that you can largely predict someones appearance genes from which cluster of personality genes they come from, and the other way around. Does it seem unreasonable to talk about which cluster you belong to without specifying which of appearance or personality you are refering to, given that you can say things like "people with visual trait X have personality trait Y", which I hope seems reasonable itself, in this case. Note that this is the imaginary world where "strong racism" is correct.

Fair enough. This is what our pet racist here believes is true, yes?

Ok, given that, if you're still reading, let's draw a parallel to the weak racist world where genes affect appearance and such, and memes affect intelligence and personality and such, and these factors are both highly heritable and highly corellated. (This is our world). When asked to comment, the weak racist says "Why should it matter whether a highly heritable component of who someone is is genetic or memetic or on the 13th chromosome or the 14th? Can't we just point to the empirical clusters and say 'that there is a meaningful cluster', given that it does seem to cluster in a meaningful way?". At this point the lines are open and the objections are coming in fast:

No no no no no.

Consider various brands of Wiggin.

• For whatever reason, Wiggins are usually born to disadvantaged families, and must live off. ketchup and steal simply to survive. They often have trouble breaking these habits when they are fortunate enough to escape their poverty
• The genes responsible for black hair and green eyes are heavily correlated with the ones causing kleptomania and a craving for ketchup.
• Wiggins are constantly expected to steal and eat ketchup, and people generally behave as society expects.

Now, in all these situations Wiggins do, in fact, steal and eat ketchup. However, in the first case, once we know that this particular wiggin grew up in a well-to-do environment, we should no longer expext them to steal our ketchup. In addition, we should expect anyone who grew up in a disadvantaged home to act "wiggin-like", not just those with black hair and green eyes.

In the second case, there is no need to consider the upbringing of a particular Wiggin, since they all have similar odds of stealing our ketchup. In this case, however, we can perform genetic tests to identify whatever genes may be causing this problem; and of course there is a case to be made for sterilizing Wiggins to prevent their criminal genes from spreading - especially to non-wiggins, who would interfere with our ability to judge the likelihood of ketchup-theft by a particular individual.

In the third case, we have something of a dilemma. On the one hand, we need to protect our ketchup from thieving Wiggins. On the other hand, assuming any Wiggin will try and steal it will only encourage them. In this case, while we realize that a wiggin-like appearance is a risk factor for ketchup-theft, we must strive to treat them equally; only hiding our ketchup if we know they already steal (and we should do likewise to non-wiggins.)

"But there are places (like Vancouver) where, for selection and social reasons, race is independent of other things, therefore race is not interesting." Yes, then no. In our imaginary strongly racist world, there are places where green eyes and black hair does not corellate with a ketchup fetish and kleptomania, but in most of the world it does, so "wiggin" is still a meaningful term. More generally, just because you can find a subset of your survey population that does not have the corellations you find in the whole, doesn't mean you can reject the corellations in the whole. Especially given that if you look at enough subpopulations, you'll find ones that go just about any way you like, so you might as well cherry-pick your data if you are going to do that. All such a non-corellation proves is that the variables of interest don't have a common atomic cause (or that you have selection effects in your data).

What this proves is that you should take location into account when estimating the odds of a particular individual acting like a Wiggin.

"What about a guy who wears a cardigan and goes to harvard? If you learn he is black, should you then conclude he is stupid and violent like the stereotype?" No, because no matter how things go, who you became screens off any possible cause.

Not necessarily. If race determines violent tendencies and intelligence, then, while he may well be unusually intelligent, we should still be wary of him attacking us. Furthermore, we should increase our probability that he is unusually stupid for a cardigan-wearing harvard-goer, and achieved that status by some other means than intelligence.

There. Now we have comprehensive lines of retreat. Now and only now are we prepared to go take an unbiased look at the data, because none of the possibilities are scary anymore. I haven't looked very hard, but I think it's the third case. Not that I really care; I've got plans however it happens to be.

I'm pretty sure that admitting that, if racists were right, genocide would be justifiable, is a line of retreat. But then I already admitted that, so whatever. The problem is that you can't get from "I update my probability of a harvard degree downward when I learn that he's black" to "racists are right". You have to check if the environment is causing all or most of the differences.