Thirdly, and of most generalisable importance, sorry if this sounds insulting, but either you do not understand Baysian probability, or more likely you are ignoring it due to motivated cognition or speaking hyperbolically. If HBD is true, then the group intelligence is your prior, the conversation provides more information and allows you to update to a posterior. I agree that a conversation could provide more information on IQ than HBD (assuming for sake of argument that HBD is true), but just because you have updated does not make your prior useless.
I understand Bayesian probability. I also understand that human beings aren’t perfect Bayesian updaters, and find the arguments founded the basis that a theoretical mind operating on optimal rationality could get more information to be not-even-wrong.
Ok, ignoring HBD for a moment, on a scientific level this is just wrong. To take a less inflammatory example, suppose vitamin B12 increases lifespan, but its a very small effect only apparent over a very large number of people, and you get far more information from diet. Therefore, by your logic, B12 is useless and has the same scientific status as aether.
No.
The aether theory was abandoned because people thought it was wrong—when it was replaced by relativity, relativity had no use (at least for a few decades). You are comparing the aether, which makes no predictions, with HBD, which does make predictions, although you do not think these predictions are practically useful.
Correct.
Again, I’m sorry if what I’ve said seems insulting, but in your haste to take down HBD you are also getting rid of probability theory and the scientific method. You are confusing practicality with truth and ‘comparitivly low information’ with ‘zero information’.
No, I’m not pretending humans are perfectly rational agents that can successfully utilize relatively low levels of information.
No, I’m not pretending humans are perfectly rational agents that can successfully utilize relatively low levels of information.
So your argument is that the information conveyed by HBD is so low that humans can’t perceive it?
IIRC the threshold for updating for the average person is 1⁄3 - increasing a Bayes factor by less than this is below the threshold of perception.
Some HBD people claim that people of African descent are one standard deviation below normal IQ, and Ashkenazi Jews are one standard deviation above. Mensa is open to people of about IQ three standard deviations above normal. [edit: apparently its only two, but I’m leaving the following calculation unchanged] If the HBD people are correct, and the standard deviations are the same, then Ashkenazi Jews are 718 times* as likely than African people to have Mensa-level IQ. IIRC the difference in crime rates between Blacks and Asians is also quite large (>10x).
Now, I’m not saying the HBD people are correct here—that’s not my field—but the claims they make do concern very large differences which even a far from optimal Baysian updater could make use of, if the claims are correct.
Now, I’m not saying the HBD people are correct here—that’s not my field—but the claims they make do concern very large differences which even a far from optimal Baysian updater could make use of, if the claims are correct.
What’s the difference in expected crime rate between a black man in a suit and an Asian man in a suit? What if it’s merely business casual? What if it’s jogging paints and a paint-covered t-shirt?
What if they’re both wearing clothes signaling clear gang affiliation?
Once you consider more relevant information, race adds little to no additional information, meaning using it is more likely to move you -away- from the correct answer than towards it.
The thing about suits is its a signal that can be faked, if you have enough money. Perhaps that is why the mafia wear suits (source: TV), so it can help give the impression of being legitimate businessmen.
Sure, in extreme cases such as wearing gang colours or tats that gives a lot of information. But some gangsters wear suits, and not all crime is organised crime, and I’m sure that perfectly ordinary businessmen who have no gang affiliations whatsoever commit crimes sometimes.
In many circumstances race does give comparitivly little knowledge. Asian women commit violent crimes far less then black men, but if an Asian woman’s fingerprints are on the murder weapon, then its her that the police should investigate.
On that note, there are also large differences between men and women. I am a man, but I concede that while it may be sexist for a woman to be far more worried about male assailants than female assailants, this behaviour is also perfectly rational. WRT domestic violence the police do automatically assume that the man is the perpetrator, which is more likely than not the case, but also leaves the system open to abuse if being male is considered sufficent grounds for arrest in absence of any supporting evidence.
The interesting thing about comparing the black/asian difference and male/female difference is that the side which supports discrimination based upon the difference is the left in one case and the right in the other, so its interesting to see whether people have a consistent opinion across both cases. One debate tactic is to counter “all men are potential rapists” and “men should be taught not to rape” is to compare with statements replacing sex with race, like “teach black people not to steal” in an attempt to expose underlying hypocrisy. Of course, these two cases are not exactly symmetrical.
For anyone approximating a Baysian reasoner, for more information to move you away from the truth is absurd. However, I do acknowledge that any discrimination on group based differences is open to abuse by those ‘reasoning’ in bad faith.
The children’s cartoon Dora the Explorer which is heavily oriented towards Hispanics has a resident character named Swiper who tends to steal stuff. In a lot of episodes Dora and friends stand around loudly yelling “Swiper, no swiping!” X-D
WRT domestic violence the police do automatically assume that the man is the perpetrator, which is more likely than not the case, but also leaves the system open to abuse if being male is considered sufficent grounds for arrest in absence of any supporting evidence.
This police assumption likely increases the number of falsely accused and convicted men, and of wrongly non-accused or acquited women. The justice system sometimes has very high conviction rates (i.e. persons convicted out of those brought to trial): above 90% for federal cases in 2001-2012 (random Google link to a PDF link from 2012). Therefore, one must ask what independent evidence we have about how much more likely men are to be the perpetrator in domestic violence cases.
This is a good point. Given that men commit more violence in general, it seems likely that they commit more domestic violence. However, I’m not sure that there is much evidence as to how large this difference is. One could look at the rates of domestic violence in male and female gay relationships, as this removes the ‘the man is always arrested’ bias, but there is evidence of differences in violence behaviour between gay and straight people, so this wouldn’t help all that much.
Perhaps look at the rates of mothers vs fathers beating children?
Different kinds of domestic violence (against children, spouses, parents, etc.) have significant psychological or behavioral differences. I don’t want to generalize from “parents of gender X more likely to beat their children” to “people of gender X more likely to beat their spouse” without evidence.
In any case, how do you propose to look at the rates of any kind of violence? If we don’t trust data from the justice system, or from the police, and we obviously can’t trust self-reporting and surveys, then what do we do?
All data is relative to a definition of what constitutes domestic violence. Fifty or a hundred years ago, men raping their wives wasn’t violence. Today, some surveyors (or police or judges) sometimes consider a wife hitting or raping her husband not to be violence, but the husband hitting his wife to be violence. We need to agree on a definition of violence, and then to find a reliable data source that uses the same definition.
Well, this is one of the many problems with sociology. There are some obvious approaches to use, such as finding crimes which are solved far beyond doubt, such as where there are many witnesses or DNA evidence, and hope this generalises to crimes which are harder to solve, such as domestic violence.
Of course, as you point out, these different kinds of violence might not generalise, as different people commit different crimes for different reasons. So I really don’t know what to do about crimes that happen in private where there are no witnesses, short of putting cameras in every room of every house.
I understand Bayesian probability. I also understand that human beings aren’t perfect Bayesian updaters, and find the arguments founded the basis that a theoretical mind operating on optimal rationality could get more information to be not-even-wrong.
No.
Correct.
No, I’m not pretending humans are perfectly rational agents that can successfully utilize relatively low levels of information.
So your argument is that the information conveyed by HBD is so low that humans can’t perceive it?
IIRC the threshold for updating for the average person is 1⁄3 - increasing a Bayes factor by less than this is below the threshold of perception.
Some HBD people claim that people of African descent are one standard deviation below normal IQ, and Ashkenazi Jews are one standard deviation above. Mensa is open to people of about IQ three standard deviations above normal. [edit: apparently its only two, but I’m leaving the following calculation unchanged] If the HBD people are correct, and the standard deviations are the same, then Ashkenazi Jews are 718 times* as likely than African people to have Mensa-level IQ. IIRC the difference in crime rates between Blacks and Asians is also quite large (>10x).
Now, I’m not saying the HBD people are correct here—that’s not my field—but the claims they make do concern very large differences which even a far from optimal Baysian updater could make use of, if the claims are correct.
*From running in python:
import scipy
scipy.stats.norm.cdf(-2)/scipy.stats.norm.cdf(-4)
It’s two standard deviations. (source)
Fixed, thanks.
What’s the difference in expected crime rate between a black man in a suit and an Asian man in a suit? What if it’s merely business casual? What if it’s jogging paints and a paint-covered t-shirt?
What if they’re both wearing clothes signaling clear gang affiliation?
Once you consider more relevant information, race adds little to no additional information, meaning using it is more likely to move you -away- from the correct answer than towards it.
The thing about suits is its a signal that can be faked, if you have enough money. Perhaps that is why the mafia wear suits (source: TV), so it can help give the impression of being legitimate businessmen.
Sure, in extreme cases such as wearing gang colours or tats that gives a lot of information. But some gangsters wear suits, and not all crime is organised crime, and I’m sure that perfectly ordinary businessmen who have no gang affiliations whatsoever commit crimes sometimes.
In many circumstances race does give comparitivly little knowledge. Asian women commit violent crimes far less then black men, but if an Asian woman’s fingerprints are on the murder weapon, then its her that the police should investigate.
On that note, there are also large differences between men and women. I am a man, but I concede that while it may be sexist for a woman to be far more worried about male assailants than female assailants, this behaviour is also perfectly rational. WRT domestic violence the police do automatically assume that the man is the perpetrator, which is more likely than not the case, but also leaves the system open to abuse if being male is considered sufficent grounds for arrest in absence of any supporting evidence.
The interesting thing about comparing the black/asian difference and male/female difference is that the side which supports discrimination based upon the difference is the left in one case and the right in the other, so its interesting to see whether people have a consistent opinion across both cases. One debate tactic is to counter “all men are potential rapists” and “men should be taught not to rape” is to compare with statements replacing sex with race, like “teach black people not to steal” in an attempt to expose underlying hypocrisy. Of course, these two cases are not exactly symmetrical.
For anyone approximating a Baysian reasoner, for more information to move you away from the truth is absurd. However, I do acknowledge that any discrimination on group based differences is open to abuse by those ‘reasoning’ in bad faith.
The children’s cartoon Dora the Explorer which is heavily oriented towards Hispanics has a resident character named Swiper who tends to steal stuff. In a lot of episodes Dora and friends stand around loudly yelling “Swiper, no swiping!” X-D
This police assumption likely increases the number of falsely accused and convicted men, and of wrongly non-accused or acquited women. The justice system sometimes has very high conviction rates (i.e. persons convicted out of those brought to trial): above 90% for federal cases in 2001-2012 (random Google link to a PDF link from 2012). Therefore, one must ask what independent evidence we have about how much more likely men are to be the perpetrator in domestic violence cases.
This is a good point. Given that men commit more violence in general, it seems likely that they commit more domestic violence. However, I’m not sure that there is much evidence as to how large this difference is. One could look at the rates of domestic violence in male and female gay relationships, as this removes the ‘the man is always arrested’ bias, but there is evidence of differences in violence behaviour between gay and straight people, so this wouldn’t help all that much.
Perhaps look at the rates of mothers vs fathers beating children?
Different kinds of domestic violence (against children, spouses, parents, etc.) have significant psychological or behavioral differences. I don’t want to generalize from “parents of gender X more likely to beat their children” to “people of gender X more likely to beat their spouse” without evidence.
In any case, how do you propose to look at the rates of any kind of violence? If we don’t trust data from the justice system, or from the police, and we obviously can’t trust self-reporting and surveys, then what do we do?
All data is relative to a definition of what constitutes domestic violence. Fifty or a hundred years ago, men raping their wives wasn’t violence. Today, some surveyors (or police or judges) sometimes consider a wife hitting or raping her husband not to be violence, but the husband hitting his wife to be violence. We need to agree on a definition of violence, and then to find a reliable data source that uses the same definition.
Well, this is one of the many problems with sociology. There are some obvious approaches to use, such as finding crimes which are solved far beyond doubt, such as where there are many witnesses or DNA evidence, and hope this generalises to crimes which are harder to solve, such as domestic violence.
Of course, as you point out, these different kinds of violence might not generalise, as different people commit different crimes for different reasons. So I really don’t know what to do about crimes that happen in private where there are no witnesses, short of putting cameras in every room of every house.
And yet you belive yourself sufficiently rational that you can foresee all the implications encouraging people to believe things likely to be false.