= 27d567bc2d48d9f40e9ccc20ea370b15)
Vaniver comments on Why Are Individual IQ Differences OK? - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (526)
So what you're saying is that it's ethically wrong to use Bayesian reasoning.
If you intend to use an African prejudgment heuristic like 1 (below) rather than reacting as if you've done an equation that takes into account other relevant data like 2 (below), then I think your probability equation needs an upgrade.
1) African prejudgment heuristic: "The IQ test(s) said African's IQs are lower than those of whites, therefore this specific African individual is likely to be relatively stupid compared to my white friends."
2) Reasoning that Takes Relevant Data Into Account: "The IQ test(s) said African's IQs are lower than those of whites. However, there are known flaws with IQ tests such as cultural bias, so that figure might be wrong. Most published research findings are false (PLOS Medicine), so I should apply healthy skepticism to all the research I read. This is not likely to be an accurate piece of data to use as a Bayesian prior. Once I've decided on a prior to use, I should then adjust for other relevant data (things I know about the specific individual).
Then, if one wants to behave rationally after one has decided what to believe, I think one must continue by thinking something like this:
"Considering things like...
A. ...the fact my prior is likely to be inaccurate (there isn't an accurate one for this subject as far as I'm aware)...
B. ...the fact that even if the IQ study is correct and the IQ test it used was accurate, there's a decent chance (25%) that this specific individual has an IQ above the African average - meaning I need to avoid the logical fallacy called hasty generalization...
C. ...the risk of lost utility via damaging this individual's reputation, emotional health or opportunities for success by pre-judging them...
D. ...the risk that this makes for bad social signaling and witnesses may retaliate against me with one or more forms of social rejection if I pre-emptively treat an African like an idiot...
...do I really want to treat this person as if they are less intelligent?"
I think you may have reacted to my "I hope it is false." statement or my "it's ethically wrong to prejudge individual Africans" statement - but that shouldn't matter to your probability calculation. What should matter is to get an accurate idea of reality. Along with saying other things, I also provided other factors which are relevant, as credible sources can confirm. If one wants to be a good Bayesian probabilist, after one specifies some prior probability, one must then update it in the light of new, relevant data. [1] This situation where you focused on one part of my comment and ignored the rest reminds me of those math problems where there's an irrelevant statement thrown in just to distract you. I of course did not intend to distract you, but since you seem to think that Bayesian reasoning in this case means ignoring all the other data I presented, it appears that you have skipped the parts of the process where you ensure an accurate prior and update your prior with new, relevant data.
Life is really, really complicated. I doubt it's ever wise to just grab a prior and run with it and I certainly hope that you do not reason this way.
This reads like a classic case of motivated cognition. You don't want to believe the conclusion; therefore, you selectively look for potential flaws then declare that there's still a chance. The reason I believe the connection between race and intelligence is not just because of the tests but because more or less every relevant aspect of reality (e.g., the statistic on race and crime, the nearly complete lack of blacks in intelligence intensive fields, e.g., math, programing, the state of majority black countries) looks the way one would expect it to look if the connection existed.
Who is claiming otherwise?
Yes there is. You just don't want to believe it exists.
Actually the chance of this particular black being above the African average is 50% (more if I condition on the fact that he is in the USA). The probability that he is above the white average is significantly less. The probability that he is above some high cutoff is can be even lower.
This is a general argument against using evidence of any kind.
This is potentially a problem, although here the problem is arguably with the witnesses behaving irrationally and/or participating in out of control signaling arms races than with the strategy itself.
Did you stop to make distinction between me being influenced by motivated cognition and alternate explanations like:
Me seeing significant flaws in data that would otherwise support your conclusion. Part of this may be that I've spent a significant amount of time reading about IQ and giftedness and I have learned that there are a lot of pitfalls to doing IQ related research.
Me simply being unaware of relevant data. (This might be the case in the event that the people who supplied my data were influenced by motivated cognition or confirmation bias.)
You seeing motivated cognition in my words because of being influenced by motivated cognition yourself?
There is an alternate explanation for those which does not have the same issues that IQ tests and studies have: The effects of slavery and prejudice. We are certain that slavery and prejudice has influenced them, and that it has existed for a long time. To know this, one must only look at the KKK or investigate the history of black enslavement. Imagine a third world country. Imagine that an equal proportion of those inhabitants are removed and used as slaves. Imagine an equal proportion of them dying. Imagine that they're freed, but all of them - not some but all - are freed into a situation of extreme poverty where they don't even own a home or have the ability to read. Many still aren't being taught to read. Consider also that even though there have been advances in medicine, poverty means you can't afford health insurance or medical treatments. Don't think that disability and chronic illness are uncommon - they're not. Not even in America. They're probably especially common for the poor. Don't think that severe worker abuse ended with slavery, either - do some research on sweatshops in America sometime. Now take into account the effects of stress, and the human element - how those effects can compound into things like mental illnesses and drug addictions. Would you predict that the majority of these people who started out with literally nothing and without even the education to read would manage to avoid pitfalls like disability, mental illness, drug addiction and sweatshops and carve an opportunity to excel out of poverty and ignorance over the course of 150 years? I would not expect that. I would expect most of them to have fared poorly.
I don't see a good way to tell the difference between a low IQ score due to actually being less intelligent versus a low IQ score due to nurture-related reasons such as the following:
If you want to attribute the IQ scores to race, not poverty or circumstances, then there needs to be a good way to distinguish between nurture and nature as a cause for low IQ scores. Do you have one?
If it's true I want to believe it. However, it's hard to believe it exists without a citation. Do you have one?
I respect you more for being able to say something that supports my view better than it does yours. +1 karma for that. I still think the number is 25%, however I do not view this as a key point in our disagreement, so I will leave it at that.
You appear to have taken that statement as an argument regarding what to believe. It was not. I deliberately put that part after the section where I was discussing deciding what to believe, and put it under "if one wants to behave rationally".
As the saying goes, "Life is an IQ test." The predictive ability of IQ on income (and most other statistics of interest) is very similar for each race, which suggests that differences in measured IQ scores map onto differences in life outcomes. To the extent that a medical condition makes someone test poorly, it generally also makes them live poorly. (There are a handful of prominent exceptions to this- like dyslexia- which don't significantly impact the main point.)
Now, where those IQ differences come from in the first place is an interesting question. Many people have looked at it, and the consensus answer for individual IQ differences is "somewhere between 50% and 80% of it is genetic," and the extrapolation from that to group IQ differences is somewhat controversial but seems straightforward to me. A perhaps more interesting question is "what knobs do we have to adjust those IQ differences?"
Here's Jason Richwine explaining that all serious scientists have agreed on the basics of IQ for decades, and the media is completely mistaken on the state of reality and scientific consensus.
It seems pretty relevant to me, because it looks like basic statistical innumeracy on your part, unless you think the IQ distribution of African Americans is tremendously skewed such that the mean intelligence is the 75th percentile of intelligence, rather than the 50th percentile like it would be in a symmetric distribution. (Or you think that the African average is higher than the African American average, which is very much not the case.)
As a stand-alone statement, I would probably leave this alone. But as a response to "What about nurture", the first thing that comes to mind is:
Has Vaniver adequately corrected for the just world fallacy?
Ok, that's interesting, but it does nothing to rule out nurture factors that would impact both IQ and income.
I agree that IQ is mostly genetic and that IQ does seem to correlate with a lot of factors. I'm not saying IQ does not exist. What I am saying is that, specifically when it comes to black people, there are other factors that are definitely influencing performance and IQ scores. Therefore I reject claims about IQ and race that haven't controlled for known factors.
Actually, when I read that section (it starts with "What scholars"), I parsed it like this:
Jason explicitly says that there's a scientific consensus on many issues that seem controversial to journalists.
Jason states that virtually all psychologists (not scientists) believe there is a general mental ability factor (That he's not saying is specifically connected to race).
Then, without qualifying these statements with anything along the lines of "most x believe", he states: "In terms of group differences, people of northeast Asian descent have higher average IQ scores than people of European lineage, who in turn have higher average scores than people of sub-Saharan African descent."
I will not assume that this sequence of claims means that the group differences statement is also something scientists have a consensus about. If I did, that would be a non-sequitur.
Also, below that, he writes:
"It is possible that genetic factors could influence IQ differences among ethnic groups, but many scientists are withholding judgment until DNA studies are able to link specific gene combinations with IQ."
This is where I stop reading the article because it is clear to me that it does not say "there's a scientific consensus that there's a link between race and IQ". If you have a credible source for that claim, I'll be curious about it. No more Politico articles please.
It might or might not have been an error. In any case, I'm not going to go digging for that right now because I still think knowing the percentage is irrelevant to the current point. Whether the figure is 50% or 25%, it is still true that a significant proportion of people will have an IQ above average and therefore it would be hasty generalization to assume that a person of a certain group was an idiot. That is one way in which the exact number is irrelevant. However, that point about hasty generalization is much more irrelevant at this particular moment because we haven't even decided on a prior, let alone have we got a decent posterior - so the step where we have a concern about making a hasty generalization based on our probabilities should be in this disagreement's future. If it becomes relevant, I will dig around, but not right now.
It's not clear to me why you would be interested in nurture factors. There are two things going on here: the ability of IQ to measure intelligence, and the historical causes of intelligence.
With the exception of disorders that prevent people from testing well without significantly impacting life outcomes, the historical causes of intelligence don't appear to have much to do with the ability of IQ to measure intelligence. A nurture factor (like, for example, being breastfed as a child or being struck on the head) actually alters someone's intelligence, and their intelligence influences both their test scores and their income.
Again, this looks like it's mixing up the historical causes and the predictive ability. If the predictive ability is the same independent of race (it is), then it doesn't matter why the racial IQ averages are the way they are. What we would need to show to discount IQ measurements is that the IQ measurements are not as predictive for members of one race than another.
As an example of a real bias like this, girls tend to get better grades than standardized test scores alone would predict. In order to get an accurate estimate of what a girl's grades would be from her standardized test scores, you need to adjust upwards because she's a girl. Symmetrically, boys score better than one would expect from their grades, and so when predicting scores one needs to adjust upwards. When moving in the opposite direction, one would need to adjust downwards; a girl's grades overestimate her standardized test scores. But note that this doesn't mean we throw out the data- it's still predictive! We just adjust it the correct quantitative amount.
Now, do we know the historical causes of that effect? I'm not familiar with that field, but it seems like there are lots of plausible theories that probably have support. Even without knowing the causes, though, we can use our estimates of the size of the effect in order to predict more accurately.
What would you call a scientist who studies intelligence? (I suppose I should also make clear that by "serious" I mean a scientist speaking confidently in their field of expertise.)
One does not say "most scientists believe that hydrogen has one proton," one says "hydrogen has one proton."
Here's the APA report he references. The group means section starts on page 16.
In general, though, asking for citations like this is really frustrating, because it doesn't seem like the true rejection. The linked Richwine article referenced more serious sources that you could find if interested, and even if you didn't notice that Googling "racial IQ averages" leads to this as the fourth hit, and if sufficiently motivated you could find the paper that journalist was writing about, and so on.
But if you're not curious enough to seek out this information, and you don't seem to have updated on the other information I've provided, what reason do I have to expect that the difference between my position and your position is that I have citations, and as soon as I share them you'll adopt my position?
Sure. When you think in distributions, an estimate generally comes with both a mode and a precision (or, relatedly, the standard deviation). Knowing someone is African American gives you an estimate with a mode of 85 and standard deviation of 15, which has a non-trivial but small chance of being over 120. Knowing someone got a 120 on a recent IQ test gives you an estimate with a mode slightly south of 120 and a standard deviation of probably 2-5, depending on the precision of the test.
That's a saying? Are there also sayings "Life is a test of height.", "Life is a test of immune system efficiency" and "Life is a test of facial symmetry"? We may as well round out the set. Anything of comparable test significance that I missed? Oh! "Life is a test of breast perkiness" and "Life is a test of capacity for situation-appropriate violence."
(I actually agree with the point of the rest of the paragraph.)
Those may be qualitatively similar but I would suggest they are quantitatively different. I would be surprised if facial symmetry did not correlate with income, health, social status, and so on, but I would expect the correlation to be much lower than the correlation with IQ. The saying means that most metrics of life success are moderately highly g-loaded, and so it makes sense that IQ correlates positively with basically everything good.
I think you either drastically overestimate the impact of IQ or underestimate the predictive value of those other factors of life success---particularly in environments different to those experienced by the white male nerd class of first world countries. Regardless, the paragraph with that alleged saying redacted would be far more persuasive than the one with it included. It strongly undermines the credibility of your point. I currently agree with you despite the opening sentence, not because of it.
Perhaps we should switch to numbers to make it easier to communicate. I haven't looked at any numbers recently, so I don't expect these guesses to be particularly accurate, but I'd guess IQ correlates .2 to .6 with most interesting measures of life success, and height correlates around 0 to 0.1 (after controlling for IQ). I'd guess facial symmetry has correlation >0.3 with other health-related things, a correlation around 0.4 with social things, and a correlation below 0.3 with the rest.
If you have a source handy that estimates these sorts of correlations, I'd like to see it, but I don't think it's important enough to spend time hunting something down.
I would be very surprised if IQ correlates at .6 with, say, wealth or income. Parental wealth and income possibly correlate no more than 0.5 to childrens' incomes, and it would be frankly remarkable for IQ to be (1) transmitted intergenerationally to a large degree, and (2) more closely correlated to financial outcomes than one's parent's financial outcomes, since your parents often give you not only your genes, but your inheritance/early support, financial assumptions, and first set of career contacts.
Do you similarly control the IQ finding for height? For obvious reasons that is necessary for consistency.
I'd certainly expect (based on loose memories of studies encountered) height to predict more than '0 to 0.1' and IQ to predict a heck of a lot less than 0.6, especially when considering populations that are not limited to first world nations. I wish it were otherwise, naturally. I'd love the world to be more biased in favour of my own highest stat.
In any case, I would consider it legitimate to readers who encounter "Life is an IQ test" delivered in response to the quoted context to be sufficient evidence that the comment is not worth reading and downvoting and ignoring it. It was only because I was more patient that normal that I bothered to read further and found you had a good point hidden in the detail. Do with that information what you will.
EDIT: sorry, misunderstood your comment.