I expect people to have a better idea of group sizes than normal distributions, but the worry is with ‘chosen uniformly at random,’ since people tend to be in IQ bubbles. If you can think of 100 people you know dumber than you, and 10 people you know smarter than you, then you might say “well, that implies a group size of about 10,” without realizing that half of those people you know from graduate school where the average IQ is in the top percentile.
But it seems like we can compare that to SAT scores and measured IQ (if provided) and that’ll give us useful information.
I’ve also edited it so it’s a little less ambiguous (were you randomly selected before or not?) but I want to add the super formal mathematical version to make it totally unambiguous, which may or may not be a good idea.
Yeah, I’d expect some pretty serious regression to the mean, or more accurately to slightly above the mean, going on with the group-of-adults method. People of low intelligence hang out with other people of low intelligence and probably have a slightly inflated opinion of their relative capabilities; likewise for people of high intelligence. But people outside your social circle are totally invisible.
At least, assuming people actually use the group-of-adults method, which I don’t really expect them to.
At least, assuming people actually use the group-of-adults method, which I don’t really expect them to.
That is, you would expect people to say “well, I want my IQ to be 145, so let’s calculate the percentile for that, and then use that percentile to calculate the group size”? Or that they’d just leave the question blank?
I’d expect a lot of people to leave the question blank if there are other IQ questions. I’d also expect a lot of people to work backwards from scores, more or less formally—I don’t think I’d expect many people to actually do the math or track down a normal distribution calculator, but thinking like “well, I got 99th percentile on that standardized test in high school, so I’ll say 100” would probably be common.
To be fair, that’s probably more accurate than what you’d get by counting up the number of people you know who’re smarter than you.
I expect people to have a better idea of group sizes than normal distributions, but the worry is with ‘chosen uniformly at random,’ since people tend to be in IQ bubbles. If you can think of 100 people you know dumber than you, and 10 people you know smarter than you, then you might say “well, that implies a group size of about 10,” without realizing that half of those people you know from graduate school where the average IQ is in the top percentile.
Meh. If I was asked that question, I’d try to guesstimate how many people smarter than me there are in my town and divide by its population.
I like the group-of-adults formulation, it’s clever and I think it’s more amenable to getting at people’s real estimates.
I expect people to have a better idea of group sizes than normal distributions, but the worry is with ‘chosen uniformly at random,’ since people tend to be in IQ bubbles. If you can think of 100 people you know dumber than you, and 10 people you know smarter than you, then you might say “well, that implies a group size of about 10,” without realizing that half of those people you know from graduate school where the average IQ is in the top percentile.
But it seems like we can compare that to SAT scores and measured IQ (if provided) and that’ll give us useful information.
I’ve also edited it so it’s a little less ambiguous (were you randomly selected before or not?) but I want to add the super formal mathematical version to make it totally unambiguous, which may or may not be a good idea.
Yeah, I’d expect some pretty serious regression to the mean, or more accurately to slightly above the mean, going on with the group-of-adults method. People of low intelligence hang out with other people of low intelligence and probably have a slightly inflated opinion of their relative capabilities; likewise for people of high intelligence. But people outside your social circle are totally invisible.
At least, assuming people actually use the group-of-adults method, which I don’t really expect them to.
That is, you would expect people to say “well, I want my IQ to be 145, so let’s calculate the percentile for that, and then use that percentile to calculate the group size”? Or that they’d just leave the question blank?
I’d expect a lot of people to leave the question blank if there are other IQ questions. I’d also expect a lot of people to work backwards from scores, more or less formally—I don’t think I’d expect many people to actually do the math or track down a normal distribution calculator, but thinking like “well, I got 99th percentile on that standardized test in high school, so I’ll say 100” would probably be common.
To be fair, that’s probably more accurate than what you’d get by counting up the number of people you know who’re smarter than you.
Meh. If I was asked that question, I’d try to guesstimate how many people smarter than me there are in my town and divide by its population.