I show you a picture of an asian person (if you’re good at distinguishing them, you notice they’re of Japanese ethnicity, specifically) that you do not know, and it is obvious that I’ve photoshopped clothing, background, and other environmental visual cues that could reveal that person’s culture. You only have their body frame and their face to work with.
What is your probability assignment that this person is of generic asian (japanese) culture, as opposed to any other culture (e.g. that of amazon hunter-gatherer tribes)? Is this probability equal to that for any other culture, as per an even-distribution hypothesis?
Look at the context. Racism only predicts violence and civilsiation inasmuch as it predicts culture, and culture predicts those things better—hell, you couldn’t get a razor blade between culture and civilsiation. So why does Nyan_Sandwich
call himself a proto-racist?
The primary observation is one of race. You can visually see that someone is of asian race. You cannot immediately ascertain a specific culture without first learning and recognizing in practice behaviors strongly associated with that culture.
e.g. If you don’t know anything about japanese culture at all, you will not know that a person of japanese race who does not get upset when a stranger who is also japanese calls them by first name without honorifics is most likely not of typical japanese culture, nor will you understand why another does get upset in the same situation. Thus you cannot use their culture as a predictor, since you don’t have any signals that tell you which culture they’re part of. Race is much easier to use as a data point.
Racism only predicts violence and civilsiation inasmuch as it predicts culture, and culture predicts those things better (...)
This is not obvious, nor does it follow trivially from any logical assertions I’ve seen yet. I’ve never seen claims either way backed by sufficient evidence to move my prior significantly in either direction.
You cannot immediately ascertain a specific culture without first learning and recognizing in practice behaviors strongly associated with that culture.
That does not have the slightest bearing on what is most stronglty correlated wtih what, what the causal mechanisms are, and why on Earth Nyan-Sandwich would want to call himself a proto-racist.
I call myself a racist in that I would predict differing values for intelligence, propensity to violence, etc based on observing someone’s race. I find it interesting that there are people who would not. The ones that are especially attached to racial equality have to go to all sorts of lengths to justify why race isn’t evidence of these things.
I call myself a proto-racist because despite being racist on matters of fact, I try to not make the (default) step from there to hatred or smugness. I think it fucking sucks that some people are disadvantaged in intelligence, or in ability to function as a member of civilized society. I think we should do something nice (help) instead of something mean (genocide).
If culture comes form acculturation , it doens’t come from genes, and therefore has nothing significant to do with race. The statistical correlations you make so much of aren’t worth making anything of unless they indicate mechanisms.
statistical correlations aren’t worth making anything of unless they indicate mechanisms.
tell it to the statistics establishment. Methinks I can make better predictions using not-causally-explained statistics than I can without. For example, If I learn of a person who is black and american, I can predict that he is 5x (or whatever it is) more likely to be in prison. I can predict that he is more likely to be a part of that awful antisocial gansta culture.
Of course, if I then learn that at this very moment, he is wearing a cardigan, a lot of that goes away.
If you restrict yourself to causal models, you do very poorly. I might even be tempted to say “I guess you’re fucked then”
If you throw out information you have reason to believe is true but can’t explain the mechanism for your model is more coherent but less powerful. Does that make sense?
No. How exactly are you defining a causal vs a statistical model? What I find confusing is in the Newtonian physics limit of what you can know, I don’t think you can do better than a causal model, in some sense. I understand that it can happen that non-causal models can predict better if knowledge is not complete, I am just trying to find a way to state that formally.
Let’s talk about fluid dynamics. In FD, we have many equations that were determined by measuring things and approximating their relationship. For example, the darcy weisbach equation for drag in a pipe: dP = fd*L/D*rho*v^2/2. This equation (and other like it) is called a corellation, or an empirical equation, as opposed to a theoretical model. To demonstrate the power of corellations, consider that we still can’t predict fd from theory (except for laminar flow). At this point, it’s just a lack of computing power, the use of which would be esentially the same as measurement anyways. There were times in the past, though, where we didn’t know even in principle how to get that from theory.
Bascially, you need to be able to look at the world and describe what you see, even if you can’t explain it. If we’d taken the policy of ignoring corellations that couldn’t be understood causally, we still wouldn’t have airplanes, plumbing, engines, etc.
I don’t think these sorts of equations are good examples of what you are trying to say, since laws of physics and related equations are counterfactual and thus causal. That is, if I were to counterfactually change the length of the pipe in your equation, it would still predict the loss correctly. Invariance to change is precisely what makes these kinds of equations useful and powerful, and this invariance is causal. The fact that the equation is ‘ad hoc’ rather than deduced from a theory is irrelevant to whether the equation is causal or not. Causality has to do with counterfactual invariance (see also Hume’s counterfactual definition).
I think a better example would be something like the crazy “expert voting” algorithm that won the Netflix prize. I think in that case, though, given sufficient knowledge, a causal model would do better. Not because it was causal, mind you, but just because observing enough about the domain gives you as a side effect causal knowledge of the domain. In the Netflix prize case, which was about movie recommendations, ‘sufficient knowledge’ would entail having detailed knowledge of decision and preference algorithms of all potential users of the system. At that point, the model becomes so detailed it inevitably encodes causal information.
The people who supply statistics to people who are looking for causal mechanisms.
For example, If I learn of a person who is black and american,
“American” isn’t a race. An american of any race has a n enhanced likelihood of being in jail, becaue the US imprisons a lot of poeple. Have I converted you to Americainism?
I thought genes had something to do with race, but feel free to clarify.
Across cultures?
Currently unknown, since they are strongly correlated anyway; race is also a good predictor of cultures.
Not really. There are people of just about every race in just about every culture.
I show you a picture of an asian person (if you’re good at distinguishing them, you notice they’re of Japanese ethnicity, specifically) that you do not know, and it is obvious that I’ve photoshopped clothing, background, and other environmental visual cues that could reveal that person’s culture. You only have their body frame and their face to work with.
What is your probability assignment that this person is of generic asian (japanese) culture, as opposed to any other culture (e.g. that of amazon hunter-gatherer tribes)? Is this probability equal to that for any other culture, as per an even-distribution hypothesis?
Look at the context. Racism only predicts violence and civilsiation inasmuch as it predicts culture, and culture predicts those things better—hell, you couldn’t get a razor blade between culture and civilsiation. So why does Nyan_Sandwich call himself a proto-racist?
The primary observation is one of race. You can visually see that someone is of asian race. You cannot immediately ascertain a specific culture without first learning and recognizing in practice behaviors strongly associated with that culture.
e.g. If you don’t know anything about japanese culture at all, you will not know that a person of japanese race who does not get upset when a stranger who is also japanese calls them by first name without honorifics is most likely not of typical japanese culture, nor will you understand why another does get upset in the same situation. Thus you cannot use their culture as a predictor, since you don’t have any signals that tell you which culture they’re part of. Race is much easier to use as a data point.
This is not obvious, nor does it follow trivially from any logical assertions I’ve seen yet. I’ve never seen claims either way backed by sufficient evidence to move my prior significantly in either direction.
That does not have the slightest bearing on what is most stronglty correlated wtih what, what the causal mechanisms are, and why on Earth Nyan-Sandwich would want to call himself a proto-racist.
“Want to”? Perhaps he merely thinks it’s an accurate description.
I call myself a racist in that I would predict differing values for intelligence, propensity to violence, etc based on observing someone’s race. I find it interesting that there are people who would not. The ones that are especially attached to racial equality have to go to all sorts of lengths to justify why race isn’t evidence of these things.
I call myself a proto-racist because despite being racist on matters of fact, I try to not make the (default) step from there to hatred or smugness. I think it fucking sucks that some people are disadvantaged in intelligence, or in ability to function as a member of civilized society. I think we should do something nice (help) instead of something mean (genocide).
Come on, man. Do you even probability?
If culture comes form acculturation , it doens’t come from genes, and therefore has nothing significant to do with race. The statistical correlations you make so much of aren’t worth making anything of unless they indicate mechanisms.
tell it to the statistics establishment. Methinks I can make better predictions using not-causally-explained statistics than I can without. For example, If I learn of a person who is black and american, I can predict that he is 5x (or whatever it is) more likely to be in prison. I can predict that he is more likely to be a part of that awful antisocial gansta culture.
Of course, if I then learn that at this very moment, he is wearing a cardigan, a lot of that goes away.
If you restrict yourself to causal models, you do very poorly. I might even be tempted to say “I guess you’re fucked then”
I don’t like this. Not sure why.
Could you clarify what you mean, here?
If you throw out information you have reason to believe is true but can’t explain the mechanism for your model is more coherent but less powerful. Does that make sense?
No. How exactly are you defining a causal vs a statistical model? What I find confusing is in the Newtonian physics limit of what you can know, I don’t think you can do better than a causal model, in some sense. I understand that it can happen that non-causal models can predict better if knowledge is not complete, I am just trying to find a way to state that formally.
Let’s talk about fluid dynamics. In FD, we have many equations that were determined by measuring things and approximating their relationship. For example, the darcy weisbach equation for drag in a pipe:
dP = fd*L/D*rho*v^2/2. This equation (and other like it) is called a corellation, or an empirical equation, as opposed to a theoretical model. To demonstrate the power of corellations, consider that we still can’t predict fd from theory (except for laminar flow). At this point, it’s just a lack of computing power, the use of which would be esentially the same as measurement anyways. There were times in the past, though, where we didn’t know even in principle how to get that from theory.Bascially, you need to be able to look at the world and describe what you see, even if you can’t explain it. If we’d taken the policy of ignoring corellations that couldn’t be understood causally, we still wouldn’t have airplanes, plumbing, engines, etc.
I don’t think these sorts of equations are good examples of what you are trying to say, since laws of physics and related equations are counterfactual and thus causal. That is, if I were to counterfactually change the length of the pipe in your equation, it would still predict the loss correctly. Invariance to change is precisely what makes these kinds of equations useful and powerful, and this invariance is causal. The fact that the equation is ‘ad hoc’ rather than deduced from a theory is irrelevant to whether the equation is causal or not. Causality has to do with counterfactual invariance (see also Hume’s counterfactual definition).
I think a better example would be something like the crazy “expert voting” algorithm that won the Netflix prize. I think in that case, though, given sufficient knowledge, a causal model would do better. Not because it was causal, mind you, but just because observing enough about the domain gives you as a side effect causal knowledge of the domain. In the Netflix prize case, which was about movie recommendations, ‘sufficient knowledge’ would entail having detailed knowledge of decision and preference algorithms of all potential users of the system. At that point, the model becomes so detailed it inevitably encodes causal information.
The people who supply statistics to people who are looking for causal mechanisms.
“American” isn’t a race. An american of any race has a n enhanced likelihood of being in jail, becaue the US imprisons a lot of poeple. Have I converted you to Americainism?
Culture is culture, not race.