This is true, but links to a number of key studies would certainly be better than nothing. I’d like to follow up on the claims of the book, but I’m not about to spring for the cost of the hardcover.
I’ll admit to being skeptical that “the actual evidence is that beliefs about group differences tend to be highly accurate and proportional.” I suspect that this is true in certain domains, and not in others. My priors for any book of the sort being referenced, which draws on legitimate research to paint a grand narrative in opposition to the general impression of a field suggest that
*It’s very likely that the common wisdom of the field is wrong, or at least not on as steady ground as one might otherwise believe.
*It’s also very likely that the book leaves out countervailing evidence which doesn’t fit the grand narrative, and the reality is messier, and less easily resolved with a single clear vision, than the author would have readers believe.
Your general impression is about a hyper-politicized topic. ‘Most stereotypes are accurate’ is exactly the sort of technical claim which goes against political sacred cows I would expect researchers in a field to not play up and the few counter-examples get a great deal of press as proof of certain sacred cows. I see this all the time in intelligence-related stuff: a study claiming IQ gains or that IQ is not correlated with something gets publicized, while the studies showing the opposite get ignored or misinterpreted; hence you run into people who think that the general impression of the field is that IQ has been debunked, while it’s never been in better shape and tied to more things and closer to being nailed down into specific aspects of the brain and genes.
I see the same thing all the time as well, but I’m also used to seeing people drawing grand narratives based on an opposition to biased research which are, themselves, biased research. The fact that there are political reasons for scientists to downplay any research which indicates the accuracy of stereotypes is not sufficient to disabuse my skepticism in this case.
ETA: There’s plenty of intellectual status to be sought in meta-contrarianism. That being the case, I think one should be wary of adjusting too much on the knowledge that there are strong political biases favoring the position a person is arguing against.
The actual evidence is that beliefs about group differences tend to be highly accurate and proportional
How can something tend to be highly accurate?
ETA: Well, someone didn’t like that, but didn’t say why. Let me pose a more pointed question then. What distinction is being drawn between “tends to be highly accurate” and “is generally somewhat accurate”? If it takes me ten throws to score a treble 20 at the dartboard, am I “tending to be highly accurate”? If I score 70% in an exam, am I “tending to ace the exam”?
Perhaps the cited book answers this question. I have just checked it out from my library.
If I score 70% in an exam, am I “tending to ace the exam”?
You’re looking at the wrong problem and numbers.
If you score 70% in an exam, you are not very accurate.
If that was the only exam on which you scored 70%, and in all your other exams (of which there were more than ten) you had scores better than 95%, then you tend to be highly accurate, even though on that exam you were not accurate.
In other words, the claim by kaetl is that on average, some particular belief about group difference will probably be very accurate, because most of them are, but there are some that are not accurate at all. Which is why they tend to be highly accurate, but they’re not always highly accurate (or even accurate at all).
If it takes me ten throws to score a treble 20 at the dartboard, am I “tending to be highly accurate”? If I score 70% in an exam, am I “tending to ace the exam”?
Let’s say you have written 5 exams and I know the scores of 3 of them. 70% 75% 73%.
If I want to describe your performance I makes sense to say: “You tend to score between 70%-75% on exams.
Whenever you draw conclusions from cognitive science experiments to reality it’s useful to use language that doesn’t signal that you are 100% certain even if the experiments found highly accurate results, meaning they had very low p values,
Whenever you draw conclusions from cognitive science experiments to reality it’s useful to use language that doesn’t signal that you are 100% certain even if the experiments found highly accurate results, meaning they had very low p values,
So should one say, not “tend to be highly accurate”, but “probably tend to be highly accurate”? Or “may probably tend to be highly accurate”?
At some point you have to stop nesting dubifiers, and I think the right point is at the outset: one is enough.
Hmm, how does this take on the research square with group-dependent actions that can’t really be “accurate”? E.g. cars being more expensive when you’re black?
That seems like it could be accurate. If the salesmen selling the cars to blacks are losing sales (net commissions) due to quoting higher prices at black customers, then they’re not being accurate.
About the xkcd comic: I am not actually convinced that IS how it works.
The actual evidence is that beliefs about group differences tend to be highly accurate and proportional, see http://www.rci.rutgers.edu/~jussim/socialperception.html
Do you know where one could find the original research that book draws on, hopefully ungated?
You mean an entire book’s worth of citations, probably at least into the hundreds based on the summaries of chapters?
I suspect there’s no one convenient location for them at all, gated or ungated...
This is true, but links to a number of key studies would certainly be better than nothing. I’d like to follow up on the claims of the book, but I’m not about to spring for the cost of the hardcover.
I’ll admit to being skeptical that “the actual evidence is that beliefs about group differences tend to be highly accurate and proportional.” I suspect that this is true in certain domains, and not in others. My priors for any book of the sort being referenced, which draws on legitimate research to paint a grand narrative in opposition to the general impression of a field suggest that
*It’s very likely that the common wisdom of the field is wrong, or at least not on as steady ground as one might otherwise believe.
*It’s also very likely that the book leaves out countervailing evidence which doesn’t fit the grand narrative, and the reality is messier, and less easily resolved with a single clear vision, than the author would have readers believe.
Your general impression is about a hyper-politicized topic. ‘Most stereotypes are accurate’ is exactly the sort of technical claim which goes against political sacred cows I would expect researchers in a field to not play up and the few counter-examples get a great deal of press as proof of certain sacred cows. I see this all the time in intelligence-related stuff: a study claiming IQ gains or that IQ is not correlated with something gets publicized, while the studies showing the opposite get ignored or misinterpreted; hence you run into people who think that the general impression of the field is that IQ has been debunked, while it’s never been in better shape and tied to more things and closer to being nailed down into specific aspects of the brain and genes.
I see the same thing all the time as well, but I’m also used to seeing people drawing grand narratives based on an opposition to biased research which are, themselves, biased research. The fact that there are political reasons for scientists to downplay any research which indicates the accuracy of stereotypes is not sufficient to disabuse my skepticism in this case.
ETA: There’s plenty of intellectual status to be sought in meta-contrarianism. That being the case, I think one should be wary of adjusting too much on the knowledge that there are strong political biases favoring the position a person is arguing against.
How can something tend to be highly accurate?
ETA: Well, someone didn’t like that, but didn’t say why. Let me pose a more pointed question then. What distinction is being drawn between “tends to be highly accurate” and “is generally somewhat accurate”? If it takes me ten throws to score a treble 20 at the dartboard, am I “tending to be highly accurate”? If I score 70% in an exam, am I “tending to ace the exam”?
Perhaps the cited book answers this question. I have just checked it out from my library.
You’re looking at the wrong problem and numbers.
If you score 70% in an exam, you are not very accurate.
If that was the only exam on which you scored 70%, and in all your other exams (of which there were more than ten) you had scores better than 95%, then you tend to be highly accurate, even though on that exam you were not accurate.
In other words, the claim by kaetl is that on average, some particular belief about group difference will probably be very accurate, because most of them are, but there are some that are not accurate at all. Which is why they tend to be highly accurate, but they’re not always highly accurate (or even accurate at all).
Pedantry:
You mean “exam” here, I think.
You’re right though.
Oh, yeah. Thanks for the heads-up! (edited grandparent)
I’d be curious to see your thoughts on the book if you feel like posting them.
Let’s say you have written 5 exams and I know the scores of 3 of them. 70% 75% 73%. If I want to describe your performance I makes sense to say: “You tend to score between 70%-75% on exams.
Whenever you draw conclusions from cognitive science experiments to reality it’s useful to use language that doesn’t signal that you are 100% certain even if the experiments found highly accurate results, meaning they had very low p values,
So should one say, not “tend to be highly accurate”, but “probably tend to be highly accurate”? Or “may probably tend to be highly accurate”?
At some point you have to stop nesting dubifiers, and I think the right point is at the outset: one is enough.
Given that people are in generally massively overconfident in the conclusions that they draw, I advocate to use more dubifiers rather than less.
Is any distinction normally drawn between “X tends to do Y” and “X generally does Y”?
Hmm, how does this take on the research square with group-dependent actions that can’t really be “accurate”? E.g. cars being more expensive when you’re black?
That seems like it could be accurate. If the salesmen selling the cars to blacks are losing sales (net commissions) due to quoting higher prices at black customers, then they’re not being accurate.