OtherWorld: Yes. I got that. But thanks for spelling it out. My perspective is unchanged.
US Politics Mega-thread - Page 8463
Forum Index > Closed |
Read the rules in the OP before posting, please. In order to ensure that this thread continues to meet TL standards and follows the proper guidelines, we will be enforcing the rules in the OP more strictly. Be sure to give them a re-read to refresh your memory! The vast majority of you are contributing in a healthy way, keep it up! NOTE: When providing a source, explain why you feel it is relevant and what purpose it adds to the discussion if it's not obvious. Also take note that unsubstantiated tweets/posts meant only to rekindle old arguments can result in a mod action. | ||
Plansix
United States60190 Posts
August 17 2017 17:12 GMT
#169241
OtherWorld: Yes. I got that. But thanks for spelling it out. My perspective is unchanged. | ||
mozoku
United States708 Posts
August 17 2017 17:17 GMT
#169242
On August 18 2017 02:07 OtherWorld wrote: Show nested quote + On August 18 2017 01:45 Plansix wrote: If the discussion is going to relying the google definition of racism, my comment need not be applied. The discussions of intent and systematic racism require a less simplistic definition that does not revolve an expressed belief of racial superiority. Belief of racial separation (ie, identifying finite groups of humans like "Blacks", "Asians", "Whites", etc, while science shows that such groups do not exist) =/= belief of racial superiority Show nested quote + On August 18 2017 01:47 mozoku wrote: On August 18 2017 01:35 OtherWorld wrote: ^Well I mean, technically, racism is nothing else than an ideology (-ism) that's based on race (rac-). Thus, yes, a "racist policy" should be a policy that's made according to an ideology based on race. Which includes race-based affirmative action in racist policies, since you first have to believe that there are different races to implement them. Don't you need stereotypes (i.e. empirical distributions or priors) to decide that "racist" affirmative actions result in the greatest utility for society though? Supporters of affirmative action are implicitly acknowledging that there exists a tradeoff between employing stereotypes (which they contradictorily deem universally bad) and maximizing utility. That's very true, but then that falls in the "the end justify the means" type of action, since you're using a mean (racism/racial discrimination) that's opposite to the values you defend (no racism/racial equality) in order to attain your goal. And I think a quick glance at history is enough to know that "the end justify the means" is very rarely a good way to do things. Oh I agree there are legitimate utilitarian arguments in favor of affirmative action, even if I'm unsure if I agree with them. This is genuine btw, I'm unsure if affirmative action policies have produced real value, but I'm not sure that eliminating them is necessarily an improvement either. It's a complicated issue. | ||
xDaunt
United States17988 Posts
August 17 2017 17:17 GMT
#169243
On August 18 2017 01:56 IgnE wrote: Show nested quote + On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. | ||
OtherWorld
France17333 Posts
August 17 2017 17:19 GMT
#169244
On August 18 2017 02:12 Plansix wrote: I find the debate about defining racism about as productive as debates over proving the assistance of god. Basic as fuck. Most of my learning about race and racism has not come from discussions, but from reading books, articles and listening to podcasts like Code Switch. The topic doesn’t really lend itself to back and forth discussion for me. Most of the time find I need work out my own thoughts whatever I read and listened to. OtherWorld: Yes. I got that. But thanks for spelling it out. My perspective is unchanged. Well I mean, defining terms is the first thing people have to do in order to efficiently discuss an issue, to make sure everyone is on the same page. If you're not interested in defining terms, that means you're not really interested in discussing issues to advance forward. Which is fine, I guess. | ||
Plansix
United States60190 Posts
August 17 2017 17:21 GMT
#169245
| ||
farvacola
United States18768 Posts
August 17 2017 17:21 GMT
#169246
| ||
Plansix
United States60190 Posts
August 17 2017 17:23 GMT
#169247
On August 18 2017 02:17 xDaunt wrote: Show nested quote + On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. No we won't. A color blind society can't see racism. We won't let the right propagate something that will never exist so they can increase the burden on the oppressed to prove it is because of of their race. | ||
Trainrunnef
United States599 Posts
August 17 2017 17:25 GMT
#169248
On August 18 2017 02:17 xDaunt wrote: Show nested quote + On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. Because they know people on the right exist throughout the "conservative" spectrum, that are still looking to go back to the pre civil rights life, and for better or worse they tar everyone with the same feathers. The right needs to demonstrably show (not say) that they aren't racist before they stop being called racist, and I don't know that they ever did that. | ||
Plansix
United States60190 Posts
August 17 2017 17:28 GMT
#169249
On August 18 2017 02:19 OtherWorld wrote: Show nested quote + On August 18 2017 02:12 Plansix wrote: I find the debate about defining racism about as productive as debates over proving the assistance of god. Basic as fuck. Most of my learning about race and racism has not come from discussions, but from reading books, articles and listening to podcasts like Code Switch. The topic doesn’t really lend itself to back and forth discussion for me. Most of the time find I need work out my own thoughts whatever I read and listened to. OtherWorld: Yes. I got that. But thanks for spelling it out. My perspective is unchanged. Well I mean, defining terms is the first thing people have to do in order to efficiently discuss an issue, to make sure everyone is on the same page. If you're not interested in defining terms, that means you're not really interested in discussing issues to advance forward. Which is fine, I guess. You are talking about consensus, where we agree on terms to better serve deeper discussion of issues. We don’t define them, as the limitations of the words are only for the purpose of the specific discussion. As I said, discussions of systematic racism and intent are not well served if the consensus is that racism requires an expressed belief of racial superiority. So I choose not to participate in the discussion. | ||
Mohdoo
United States15082 Posts
August 17 2017 17:28 GMT
#169250
On August 18 2017 02:17 xDaunt wrote: Show nested quote + On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. So long as certain races are suffering from the consequences of segregation and the like, we have no reason to move on. Only once there are no longer any systematic, race-specific struggles, does it make sense to see the world as colorless. Do you think it should be a goal of society to undo the damage done by slavery and segregation? | ||
Doodsmack
United States7224 Posts
August 17 2017 17:30 GMT
#169251
On August 18 2017 02:17 xDaunt wrote: Show nested quote + On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. Not sure if you were arguing for ethnostates, but they sure aren't color blind. | ||
xDaunt
United States17988 Posts
August 17 2017 17:31 GMT
#169252
On August 18 2017 02:25 Trainrunnef wrote: Show nested quote + On August 18 2017 02:17 xDaunt wrote: On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. Because they know people on the right exist throughout the "conservative" spectrum, that are still looking to go back to the pre civil rights life, and for better or worse they tar everyone with the same feathers. The right needs to demonstrably show (not say) that they aren't racist before they stop being called racist, and I don't know that they ever did that. We don't need to show or prove anything. Nor should we try at this point. Regardless of what the Right does to appease the SJW Left, the narrative will never change. That much has been made abundantly clear over the years. | ||
Plansix
United States60190 Posts
August 17 2017 17:32 GMT
#169253
| ||
KwarK
United States40776 Posts
August 17 2017 17:33 GMT
#169254
On August 18 2017 02:17 xDaunt wrote: Show nested quote + On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. When society is treating people of different races differently then being colour blind is being willfully blind to injustice. There is nothing virtuous about refusing to see a racial subtext to injustice because you don't want to think about race. That's the problem with "all lives matter". They're insisting upon a universalist approach for the express purpose of dismissing specific problems. | ||
IgnE
United States7681 Posts
August 17 2017 17:34 GMT
#169255
On August 18 2017 02:17 xDaunt wrote: Show nested quote + On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. well yeah if you want to use something like "corn rows" as a proxy for race you are going to run into problems. but color-blind society is impossible when the numbers of black people born into single parent poverty are greater per capita than whites, etc. primarily because as mozoku has just illustrated to us, people want to use race as a bayesian prior and real, historically contingent inequality reinforce structuring racist stereotypes that feed back into reproducing that inequality. maybe we should do what plato suggests in the timaeus: Soc. And what about the procreation of children? Or rather not the proposal too singular to be forgotten? for all wives and children were to be in common, to the intent that no one should ever know his own child, but they were to imagine that they were all one family; those who were within a suitable limit of age were to be brothers and sisters, those who were of an elder generation parents and grandparents, and those of a younger children and grandchildren. Tim. Yes, and the proposal is easy to remember, as you say. Soc. And do you also remember how, with a view of securing as far as we could the best breed, we said that the chief magistrates, male and female, should contrive secretly, by the use of certain lots, so to arrange the nuptial meeting, that the bad of either sex and the good of either sex might pair with their like; and there was to be no quarrelling on this account, for they would imagine that the union was a mere accident, and was to be attributed to the lot? Tim. I remember. Soc. And you remember how we said that the children of the good parents were to be educated, and the children of the bad secretly dispersed among the inferior citizens; and while they were all growing up the rulers were to be on the look-out, and to bring up from below in their turn those who were worthy, and those among themselves who were unworthy were to take the places of those who came up? | ||
farvacola
United States18768 Posts
August 17 2017 17:34 GMT
#169256
| ||
xDaunt
United States17988 Posts
August 17 2017 17:34 GMT
#169257
On August 18 2017 02:28 Mohdoo wrote: Show nested quote + On August 18 2017 02:17 xDaunt wrote: On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. So long as certain races are suffering from the consequences of segregation and the like, we have no reason to move on. Only once there are no longer any systematic, race-specific struggles, does it make sense to see the world as colorless. Do you think it should be a goal of society to undo the damage done by slavery and segregation? Well I asked you last week what the Mohdoo reparations scheme would look like, so what would you do? All I know is that Republicans and the Right are called racists whenever we dare point out that maybe black communities have some problems of their own making, even if we start offering some solutions to those problems. Anything short of an unconditional handout as a solution is always construed as racist. Who's really guilty of reductive thinking here? | ||
KwarK
United States40776 Posts
August 17 2017 17:40 GMT
#169258
On August 18 2017 02:34 xDaunt wrote: Show nested quote + On August 18 2017 02:28 Mohdoo wrote: On August 18 2017 02:17 xDaunt wrote: On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. So long as certain races are suffering from the consequences of segregation and the like, we have no reason to move on. Only once there are no longer any systematic, race-specific struggles, does it make sense to see the world as colorless. Do you think it should be a goal of society to undo the damage done by slavery and segregation? Well I asked you last week what the Mohdoo reparations scheme would look like, so what would you do? All I know is that Republicans and the Right are called racists whenever we dare point out that maybe black communities have some problems of their own making, even if we start offering some solutions to those problems. Anything short of an unconditional handout as a solution is always construed as racist. Who's really guilty of reductive thinking here? A member of the party which is on record as having deliberately started the war on drugs as a way to disrupt the black community saying "have you considered whether the problem is that black culture is intrinsically inferior?" gets about as much of a response as it deserves. Conservatives are happy to high five over the stereotype of black fathers being in prison without ever asking themselves who put them there, or why. A colourblind law is the kind of law embodied by Jim Crow era electoral rules such as "if your grandfather could vote, you can vote". It's intellectually dishonest to blame the other side for bringing race into it. Incidentally when I looked a little more into the 1901 Alabama constitution I found an interesting factoid. Poor white districts actually voted to reject it (due to it disenfranchising a lot of them too). It was passed by black votes. Large numbers of black votes. More black votes than even lived in those districts. There's an irony for you. | ||
Mohdoo
United States15082 Posts
August 17 2017 17:40 GMT
#169259
On August 18 2017 02:34 xDaunt wrote: Show nested quote + On August 18 2017 02:28 Mohdoo wrote: On August 18 2017 02:17 xDaunt wrote: On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. So long as certain races are suffering from the consequences of segregation and the like, we have no reason to move on. Only once there are no longer any systematic, race-specific struggles, does it make sense to see the world as colorless. Do you think it should be a goal of society to undo the damage done by slavery and segregation? Well I asked you last week what the Mohdoo reparations scheme would look like, so what would you do? All I know is that Republicans and the Right are called racists whenever we dare point out that maybe black communities have some problems of their own making, even if we start offering some solutions to those problems. Anything short of an unconditional handout as a solution is always construed as racist. Who's really guilty of reductive thinking here? "of their own making" is a little misleading though. If a group is segregated, systematically oppressed, but begrudgingly let into society, those people have wildly different starting conditions. There are reasons we have programs to pull people out of homelessness. Many problems build on themselves once you start in a certain position. You keep pretending the starting conditions don't play a role in the current condition. Being willing to admit black communities hurt themselves in some ways, but not being willing to ask "why is that?" is silly. You are ignoring information. | ||
Plansix
United States60190 Posts
August 17 2017 17:41 GMT
#169260
On August 18 2017 02:34 xDaunt wrote: Show nested quote + On August 18 2017 02:28 Mohdoo wrote: On August 18 2017 02:17 xDaunt wrote: On August 18 2017 01:56 IgnE wrote: On August 18 2017 01:52 xDaunt wrote: On August 18 2017 01:47 IgnE wrote: On August 17 2017 23:39 mozoku wrote: On August 17 2017 14:29 IgnE wrote: On August 17 2017 13:48 mozoku wrote: On August 17 2017 10:50 IgnE wrote: bayesian priors are a belief system that is always only a justification a posteriori. you are sitting here making claims that its "rational" to be scared of blacks on a train because of "statistics" about "the violence of black people" (in comparison to the US) with no reference to any other details and no acknowledgement that the criterion black is always an arbitrary criterion. in other words, if you and i were betting on indidual crimes in trains i am quite positive that i would beat you over the long term by using "average rate of crimes in trains" if you were using a "racial propensity for crime" model. i am aware that this is almost tautological but i am also sure that there are nearly an infinite number of models that would have a better performance than "are the people on this train black?" now i am not arguing that a feeling of fear is never warranted (imagine a gang of bloods dressed all in red with doo rags and face tatts and all those "im black and dangerous" indicators and flashing a gun) but the millions of pieces of data your brain is analyzing has only the slightest resemblance to this "racial stereotyping" abstraction you are talking about, not all of it inherently racist, while your racial stereotyping abstraction is exactly that. for another thing there is no way that anyone working with stereotypes in practical situations has access to rigorous data of the right type. its always an operation of unjustified belief. you seem to have missed the point here ("oh you just dont understand bayes theorem! you see there are these things called 'priors' that are really cool"). yeah i know what bayes theorem is and i know what a prior is. a prior is the conscious assignation of a value to what amounts to more or less of a gut feeling. even framing the question structures priors. why are we asking "whether black people are more dangerous" rather than "whether train riders on a wednesday at noon" are more dangerous? On August 17 2017 11:08 IgnE wrote: @mozoku what i am trying to emphasize is that the context under which stereotypes form is always limited and never fully applicable to the instant situation. you have come up with an example (the train example), which you will shift in the course of this discussion, and perhaps disown entirely as "just an example" (i.e. it's not about the specifics it's about the generality of stereotypes in making efficient decisions about known statistical distributions). and yet the framework about which distribution to use in any given situation is (usually) a pre-conscious given that has no possible rational justification other than belief. and in almost all cases is using bad data. now if you think about how you want to run society, and how those pre-conscious judgments structure relations between people, you might say, "well a consideration about how likely any random black person is to be violent" is a racist consideration because it deliberately chooses his or her race as the arbitrary criterion for making a judgment to the exclusion of literally everything else we know about him or her (and often what we know about ourselves). I brought up Bayes because a stereotype is an informal prior for a person. When there is pre-existing precise language developed for having this discussion, it improves the discussion to make use of it. I didn't bring it up because it's "cool." I never made the claim that race was a strong predictor of violence on trains. In fact, I even acknowledged it was very weak in my post ("the probability of being the victim of a crime is still low"). My point was merely to demonstrate that stereotypes have predictive power in character as well as Mahjong skill prediction. Even if skin color isn't the cause (and it certainly isn't), it has predictive power because it correlates with factors such as socioeconomic status, culture, etc., and that information often isn't known in real world situations. This is where the issue of racism gets tricky. Using the predictive utility of skin color isn't necessarily racist imo; attaching an irrationally strong prior (based on skin color because it's usually the first thing you see about someone) and not conditioning effectively on a person's actions is evidence of an actual racist. Of course, this is from a pure predictive utility perspective. In reality, most people have some sense of moral obligation and probably actively work to widen their prior (i.e. "reduce their bias" in common lingo). However, it's necessarily a trade-off in the sense that actively working to widen your prior for the exclusive purpose of fairness (what is promoted by "social progressives"), while noble, necessarily reduces predictive utility. Note that this doesn't mean that people don't often widen their priors from simply learning more (e.g. maybe spending more time around a certain race and realizing their prior was too narrow)--obviously, this is a best case outcome when it happens. The apparent current progressive "correct" prior is a totally flat (uninformative) prior, which I believe to be silly. To be clear, a flat prior would be to claim that a random Chinese and a random white person have an equal chance in a game of Mahjong. If you acknowledge that skin color on a train has any predictive power for crimes in the case where you lack any other information about the person (which is a fairly realistic assumption for strangers on a train... you might be able to see their facial expression, mannerisms, and clothing but that's really about it [and all of those are also correlated with race anyway]), then you're acknowledging that stereotypes about people's skin color have some predictive utility in terms of character (if you accept propensity to commit crime as an indicator of character, which is admittedly an argument of its own). If you recall, the original point was that the word "racist" has become diluted. My argument is that "racist" has been broadened to include "people who harbor stereotypes", which is a rather ridiculous definition as I argued above--as stereotypes are not necessarily "bad", and can increase utility. [I should also note that I argued that the term "racist" has been diluted because it is used by social progressives to defend socially progressive policies from people who agree that diversity is good, but disagree with the progressives on the merits of current socially progressive policy (e.g. probably Damore imo). But we're not discussing that argument here.] Also keep in mind that I'm only making arguments that demonstrate the existence of the "stereotype utility" phenomenon here. Stereotypes are employed hundreds (if not thousands) of times each day by everyone. It's literally impossible to argue what stereotypes are appropriate for each and every situation, so an argument of existence is going to have to make do if we're keeping this discussion general. Obviously, the magnitude of the "stereotype utility" is going to vary drastically from case to case, so making arguments about it in a general discussion makes little sense. In the train scenario, the "stereotype utility" is obviously small, as I've acknowledged in all my posts. In the Mahjong example, the "stereotype utility" is larger. Let's go back to your original statement: If I'm sitting on a train car with 5 African Americans from the South Side of Chicago, I can observe that I'm x times more likely to be the victim of a crime than if I were sitting among five random members of the general US population. Therefore, I feel more threatened on this train car. It's literally the same example as Mahjong, but now it's politically sensitive. No, it's not fair to the African Americans on the train. And I would be irrational to assume I'll likely be the victim of a crime on that train, since base crime rates are very low. But I'm still logically and mathematically justified in feeling more threatened on that train car than I would with 5 other random US citizens. This is an absurdity meant to prove I don't know what. You've reduced the 5 African Americans from the South Side of Chicago to a relatively limited set of variables based entirely on race and an arbitrarily selected geographic region. Are these individuals 90 years old? Maybe they are returning from south side bingo night. Oh, you assumed you were talking about Young Black Males. Are they wearing ties and carrying The New World Translation of the Holy Bible? Are they in their mid 40s and tired, carrying bags of groceries home? Do you think a 40 something restaurant manager carrying groceries and a gaggle of older women with grey hairs still presents a greater risk than "5 other random US citizens?" What's the crime rate for 50 year old black person of either gender compared to 18-25 white unemployed white male? How do you identify an unemployed person? What color is your skin? If you are white are you more or less likely to be killed by a black man per capita or a white man per capita? Are you an old woman or are you a man in a police uniform? How many members of the public do you think know any accurate statistics on any of the questions I asked? You think being racist sometimes yields utility, and you'd really appreciate it if everyone would stop calling people racist who are racist only sometimes, especially when they were right about it. One of the issues I'm having with what you keep doing is that you never actually attempt to outline what "racism" is. I'm assuming you're of the opinion that the Mahjong scenario is not racist, but the train scenario is. But what is the difference between the train scenario and the Mahjong scenario? If you don't want the term "racist" to be diluted, you need a commonly understood definition for "racism" and only use the term when it meets that definition. Instead, we have a status quo where racism is functionally defined as a "you know it when you see it" thing, which, of course, is a total mess in practice. Earlier in this discussion, I was accused of having a racist analysis for having allegedly incorrect facts (I disagreed, but that's irrelevant)--when I think it was clear to all sides that there was no intent to be racist. In that case, "you know it when you see it" led to "alleged factual inaccuracy = racist." Now in this discussion, you're asserting (see bold) that using a stereotype as a prior is racist (i.e. having stereotypes = racist). Even though, using a stereotype in the Mahjong scenario was not racist. If this happens in a discussion literally about the dilution of the word "racist" (where you'd expect people to be particularly careful), how can you assert that people are more careful in conversations where they have political motives to slander their opponents as a racist and no incentive to be careful about how they use the term? I'd ask you to offer some real-life examples but I know that you are so far deep in abstraction land that you've lost touch with how stereotypes operate in reality, where one datapoint (skin color) swamps all the other uncountable sensory datapoints that we receive during basic, short interactions with people. People aren't actuaries with sets of data in their interactions. You're muddling definitions again. If a stereotype is a prior, then it only swamps all other sensory datapoints when the prior is extremely strong. I've already said that having an irrationally strong prior based on skin color and not conditioning on new data effectively is arguably what defines a racist. You're going back to the classic Frequentist point that "priors are sometimes misapplied, therefore they shouldn't be used." Which I disagree with. If priors are sometimes misapplied, the solution is to be aware, disciplined, and critical of your priors (i.e. be informed and challenge your beliefs). It isn't to stick your head in the sand and apply a flat prior to everything (i.e. assigning equal probability of winning to both players in the Mahjong scenario). or the solution is to use non-racist priors. if you come up to me in tattered clothes without shoes and without having showered for several days there are several priors there, none of them race-based, that condition my response. as ive said repeatedly, how you decide which information to condition your prior is always unjustifiable and faith-based. choosing to use race to condition your prior is racist. i expect you'll say something like "ideally you include all the data," at which point i say, get real, thats not how stereotypes work, the whole premise here was that it's an efficiency shortcut, and then you waffle around a bit with more abstractions while accusing me of conflating definitions etc. luckily i am not a proponent of thought crime so in your mahjong scenario its not a big deal to keep your thoughts on the likelihood of who is a better mahjong player to yourself. but if you went up and said, "hey i bet you could beat this white person here at mahjong," you'd be doing something racist. likewise if you got on the train in chicago and treated a bunch of black people going about their business like potential criminals you'd be doing something racist (and irrational). Yeah? And how many of those are left in the wake of the ever-expanding definition of racism? quite a lot. reality is pretty complex. sorry you cant use "black man" as the determinative prior anymore. Horseshit. The whole problem here is that our SJW friends cry racism/sexism/whatever every time there is anything resembling a disparate impact. They won't allow a color blind society. So long as certain races are suffering from the consequences of segregation and the like, we have no reason to move on. Only once there are no longer any systematic, race-specific struggles, does it make sense to see the world as colorless. Do you think it should be a goal of society to undo the damage done by slavery and segregation? All I know is that Republicans and the Right are called racists whenever we dare point out that maybe black communities have some problems of their own making, even if we start offering some solutions to those problems. The reductive part is that you think you are some truth tellers for saying that. Do you know who knows that black communities have problems of their own making? Black communities. They didn't need your help. But if black people turn around and say “woah there, get your racist white people under control,” there is a resounding uproar from the Right. And then something about gang violence or some shit. | ||
| ||