Razib Khan - Genomics, Intelligence, and The Church of Science

Dwarkesh Patel introduces Razib Khan as one of the top science bloggers in the world, writing about genetics, history and evolution on his blog Unsupervised Learning and on a podcast of the same name. The opening of the transcript is scrambled by the recording: it begins with a fragment of a tweet about a research institution, and a brief exchange in which Khan, asked to guess Patel's background, says he knows Patel is Gujarati and guesses half Patel and half Bania. Patel confirms the guess, and the transcript marks the audio as laughter and music.

Khan describes his background as a long-standing interest in history and demographics combined with scientific training. Over twenty years, he says, genetics has become a major tool, in biomedicine and for historical inference, and as a geneticist he can use it directly. He gives an example of his working style: while the interview was being recorded he had decided to compute pairwise genetic distances between populations for a post, simply because he could. He says he often replicates what others have already done himself.

Patel opens with the long-run question, assuming no gene editing in the near future. He lays out two pictures. One, following Charles Murray's Coming Apart, expects fat tails in the intelligence distribution because of assortative mating. The other expects a slight dysgenic effect, because people of higher intelligence have lower fertility. He asks which describes the equilibrium better.

Khan answers that there is no equilibrium yet and will not be one until reproductive differentials equalize, which could take centuries. His mechanism is timing. People with genes for educational attainment delay childbearing, many of them end up with no children because they invest in education first, and their generation times are longer. He says the math is not difficult. Correcting himself from "strong" selection, he says there is negative selection on these genes, and that everyone who has looked at the question says so, at least in the developed world.

negative selection on genes for educational attainment

Patel pushes back from first principles: he would naively expect that, with any long-term selection pressure, smart and educated people would have the cognitive tools to reproduce and to survive and thrive. Khan separates the two outcomes. Surviving and thriving is one thing, he says, and having offspring is another, and the incentives in modern society lead many people to equate thriving with being child-free. As he describes it, professional, managerial, college-educated people try to establish themselves in their twenties and think little about children, start considering them in their thirties, and sometimes wait too long, either because of fertility problems or because they cannot find a partner.

Khan grants that in the very long term there must be a limiting principle, but he argues that one does not need to be especially bright to survive and have many children. He adds that there is clear evidence that not being bright is good for reproductive output, and the two briefly refer to the film Idiocracy. He offers no figures in this passage, so the claim rests on his assertion and on the timing argument above.

Patel asks what explains the level of endogamy between jatis in India, citing a post by Khan about genetic evidence that groups living in the same village for thousands of years did not intermarry. He contrasts this with the United States, where even under slavery there was mixing, and mentions Sally Hemings and Thomas Jefferson. He asks what kind of social structure could produce such a result over a thousand years.

Khan's short answer is that nobody really knows. He cites evidence from Andhra Pradesh in south India, examined by David Reich, which implies an endogamy rate of about 99.5 per generation, and compares it with African Americans, whose endogamy rate he says was about 95 in his youth and is about 85 today. A 5 percent out-marriage rate, he notes, is ten times larger than what he is describing for India. Patel adds that average African Americans have about 20 percent European ancestry, which shows how much looser the American barriers were.

Patel wonders whether infanticide or social taboos with low reproductive fitness could be involved, but admits he does not know, and notes that for humans the pattern does not make intuitive sense. Khan agrees that the data is what it is. Other populations tend to show the universal human pattern of mixing, he says, and somehow Indians escaped it. Several Gujarati acquaintances (the transcript reads "gymnastics") have asked him what the explanation is, he says, as if there were a secret sauce, because the pattern makes little sense for a mammal.