the problem is that such assessments are meaningfully noncausal in ways that have serious issues, in particular that attributes that do not cause low intelligence are correlated with perceived low intelligence due to incorrect training data—this doesn’t invalidate that it’s correlated by any means. quanticle points out that one of the concerning correlations is race. as I said in my comment—naming it racism is a bit odd to someone who tags racism as an agentic action on the part of a person (eg, you object that the post didn’t mention race), but quanticle raises an issue I agree with, and which I would continue to assert is an issue with both studies you link: these perceptual correlations also correlate with variables we’d like to causally normalize away, so that we can consistently detect intelligence across contexts. I do think that the use of the word “racism” is misleading in this context, and that the entire world should probably taboo the word and instead criticize things as being non-causal and correlated with attributes that are downstream of prejudice.
before reading the papers in depth: I’d bet that a significant portion of the findings turn out to be due to health causing intelligence to be higher, and so variables that let you predict health will predict intelligence as well. I also expect some detectable genetic correlate.
(but the long-term solution is that we should give everyone friendly-superintelligence healthcare that allows them to reach maximum human health and customize their own intelligence level ;)
the problem is that such assessments are meaningfully noncausal in ways that have serious issues, in particular that attributes that do not cause low intelligence are correlated with perceived low intelligence due to incorrect training data—this doesn’t invalidate that it’s correlated by any means. quanticle points out that one of the concerning correlations is race. as I said in my comment—naming it racism is a bit odd to someone who tags racism as an agentic action on the part of a person (eg, you object that the post didn’t mention race), but quanticle raises an issue I agree with, and which I would continue to assert is an issue with both studies you link: these perceptual correlations also correlate with variables we’d like to causally normalize away, so that we can consistently detect intelligence across contexts. I do think that the use of the word “racism” is misleading in this context, and that the entire world should probably taboo the word and instead criticize things as being non-causal and correlated with attributes that are downstream of prejudice.
before reading the papers in depth: I’d bet that a significant portion of the findings turn out to be due to health causing intelligence to be higher, and so variables that let you predict health will predict intelligence as well. I also expect some detectable genetic correlate.
(but the long-term solution is that we should give everyone friendly-superintelligence healthcare that allows them to reach maximum human health and customize their own intelligence level ;)