David Reich’s new book about the ancestral origins of human populations and separately an MIT Technology Review article which names a website that essentially turns 23andme data into what looks like an IQ score have rekindled perennial controversies, like the existence of a genetic basis of cognitive differences observed across various populations, the actual verisimilitude of which is beside the point and beyond my expertise; rather, this is a comment on what looks like the intractability of further progress on debate and relevant to the people of divergent persuasions eagerly awaiting the precipitous flow of fresh genetic data. This controversy more or less hinges on the propriety of psychometric definitions of intelligence.  Critics contend that these measure some joint function of both intelligence and environment.  This would then certainly preclude agreement about the phenotypic definition of intelligence, thereby voiding common inference from future genetic data before we even start (e.g. just as genetic bases for anaerobic performance cannot be identified from the speed of a paraplegic’s sprint).

One might hypothesize that variations associated with things like intelligence also hold within populations, which would be enough to make the point if population is a hidden factor governing the aforementioned “environment”.  Obviously I’m no expert but it seems as if this would curtail the explanatory power of some “attenuated” polygenic score of intelligence since the genetic drivers of intelligence might vary between groups.  For example, a single SNP that accounts for approximately 30 percent of the phenotypic variation in skin tone between Europeans and West African explains a similar portion of the same across South India, but not East Asians.  In other words, Asians and Europeans are white for different reasons.  Therefore, anything short of parity in the measured amount of intelligence for people of two different populations that exhibit the same subset of variations flagged in the association study may well fail to reject other factors to the satisfaction of skeptics who suggest them.  I don’t see why the controversy won’t retain the same character as it has had for the past few decades, which is disagreement about the extent of bias that inheres within psychometrically-derived definitions of intelligence.  (A compact and accessible review of the genomic details was recently published).

This is not to favor either position.  Economists have strong and divergent opinions about the efficiency of markets, even though the precise formulation of the problem boils down to a “joint hypothesis problem” which according to Wikipedia “refers to the fact that testing for market efficiency [because] anomalous market returns may reflect market inefficiency, an inaccurate asset pricing model, or both.”   Likewise, this and other tangential impasses seem related to the fact that it can be effectively impossible to reject hypotheses that derive from a very rich and complex model of the world using tractable empirical methods, since this would involve controlling for the enormity of a subject’s upbringing and socialization with the world.  Yet it is precisely such immeasurable factors like contemporary defense of Confederate monuments, interactions current members of Congress have had with police officers, and “oppression, discrimination, and social resentment” that feature prominently in divergence between Ezra Klein’s and Sam Harris’ interpretation of the data.  It is all but impossible to find both a falsifiable hypothesis and an experiment you could conceivably execute given such a powerful hypothesis class.

This is not to disparage either position.  An alleged victory of 20th century psychometrics was the belief that a human behavior as complex and versatile as intelligence – used to describe a range of pioneers from Shakespeare to Ramanujan to Einstein – could be parsimoniously characterized by aggregating the population-normed performance on tests that were surprisingly simplistic to design and administer.  By contrast, there are others who believe intelligence is exactly as nuanced as the layman would imagine.  It seems plausible in fact that one could conclusively demonstrate that further inference on this particular question is intractable as that would essentially assume the conclusion along the line.  This would in turn make it possible for those like Professor Reich to write about various genomic associations without causing controversy so long as the work is restricted to genomic inquiry rather than claims about the sociological character of things.