Great post and I agree that the ultimate goal is to identify the causal variants and, ideally, their mechanisms. The approach you outline in fact sounds very similar to one that Turkheimer has supported (or at least validated):
"If, outside the context of race, hereditarians could identify a gene with a generalizable and quantifiable (and especially a positive) effect on intelligence, the hereditarians would be halfway there. Much as we might doubt the likelihood of success, they would hear no objections to their effort from us. Gene hunting might be controversial, but it isn’t abhorrent. If such a gene were ever documented, it would be a short step to testing whether it occurred more frequently in some groups than others." ~ Matthew, Tabery, Turkheimer (2025) (https://pmc.ncbi.nlm.nih.gov/articles/PMC12131876/)
Isn't that lovely, everyone is in agreement!
Interestingly, in the case of diabetes, there is already a well established causal variant that is correlated with IAM ancestry because it rose in frequency via Neanderthal introgression (https://pubmed.ncbi.nlm.nih.gov/24390345/). The sibling admixture paper highlights a similar example for height, so in a sense it is merely confirming with a crude method what we already know more precisely at the variant level. As with a lot of heritability estimation, I think the utility of this approach is essentially in trait prioritization: you run the sibling admixture analyses across a thousand phenotypes in a biobank; the ones -- like education -- that are squarely at the null, you prioritize for more environmental data collection; the ones -- like diabetes -- that are not, you prioritize for additional genetic data collection. This is actually quite boring and straightforward from the perspective of genetic discovery.
So I think the key ethical question is: you have a little tool that can improve the efficiency of your study in this small way but it produces a parameter that is highly prone to interpretation. It might pick up a "red hair" variant which you won't be able to link to a mechanism for many years, during which time the stigma that your population is genetically inferior will build. Do you run the tool?
In the long run, I remain confident in the ability of scientists to explain their work. So if there are potential red hair variants (and you might think of PGS like this, we don't know which parts are causal or how the causal ones are mediated), then there'll be a robust debate about how to interpret that, and ultimately enlightenment will increase.
I think I'd add to the possibility of identifying genes, the possibility of identifying other mechanisms. For example, suppose conscientiousness or agreeableness are indeed very polygenic, and say we can calculate a causal PGS score from family studies. No individual gene identified, just lots of tiny effects, but a reasonable psychological construct at the other end, which we might causally link to social outcomes. That would be useful knowledge. You don't have to start from the biological end of things; it may be possible to work backwards.
this is a really good article and i commend you on it. will type up more of a response but your approach sounds like synthetic controls in econometrics
After reading all that, I have one question: where can I buy these boffinite supplements? I couldn't find them on amazon.
Great post and I agree that the ultimate goal is to identify the causal variants and, ideally, their mechanisms. The approach you outline in fact sounds very similar to one that Turkheimer has supported (or at least validated):
"If, outside the context of race, hereditarians could identify a gene with a generalizable and quantifiable (and especially a positive) effect on intelligence, the hereditarians would be halfway there. Much as we might doubt the likelihood of success, they would hear no objections to their effort from us. Gene hunting might be controversial, but it isn’t abhorrent. If such a gene were ever documented, it would be a short step to testing whether it occurred more frequently in some groups than others." ~ Matthew, Tabery, Turkheimer (2025) (https://pmc.ncbi.nlm.nih.gov/articles/PMC12131876/)
Isn't that lovely, everyone is in agreement!
Interestingly, in the case of diabetes, there is already a well established causal variant that is correlated with IAM ancestry because it rose in frequency via Neanderthal introgression (https://pubmed.ncbi.nlm.nih.gov/24390345/). The sibling admixture paper highlights a similar example for height, so in a sense it is merely confirming with a crude method what we already know more precisely at the variant level. As with a lot of heritability estimation, I think the utility of this approach is essentially in trait prioritization: you run the sibling admixture analyses across a thousand phenotypes in a biobank; the ones -- like education -- that are squarely at the null, you prioritize for more environmental data collection; the ones -- like diabetes -- that are not, you prioritize for additional genetic data collection. This is actually quite boring and straightforward from the perspective of genetic discovery.
So I think the key ethical question is: you have a little tool that can improve the efficiency of your study in this small way but it produces a parameter that is highly prone to interpretation. It might pick up a "red hair" variant which you won't be able to link to a mechanism for many years, during which time the stigma that your population is genetically inferior will build. Do you run the tool?
In the long run, I remain confident in the ability of scientists to explain their work. So if there are potential red hair variants (and you might think of PGS like this, we don't know which parts are causal or how the causal ones are mediated), then there'll be a robust debate about how to interpret that, and ultimately enlightenment will increase.
I think I'd add to the possibility of identifying genes, the possibility of identifying other mechanisms. For example, suppose conscientiousness or agreeableness are indeed very polygenic, and say we can calculate a causal PGS score from family studies. No individual gene identified, just lots of tiny effects, but a reasonable psychological construct at the other end, which we might causally link to social outcomes. That would be useful knowledge. You don't have to start from the biological end of things; it may be possible to work backwards.
yes, but turkheimer isn't going to bother with your criticism xd
https://open.substack.com/pub/ericturkheimer/p/ancestry-and-education?utm_campaign=comment-list-share-cta&utm_medium=web&comments=true&commentId=164986614
this is a really good article and i commend you on it. will type up more of a response but your approach sounds like synthetic controls in econometrics