Some Thoughts About Polygenic Scores

Since everything seems to be heritable, in principle you could construct polygenic scores to measure any human trait. But there are a couple of problems, one is the very large sample size, the other is the problem of measuring certain traits. How do you measure “honesty?” Loyalty? Conscientiousness? Personality tests are vulnerable to social desirability bias and self deception. Even if you have an ironclad test, the large sample size required means it can be very costly to take such a sample. This is one reason polygenic scores for educational attainment work better than polygenic scores for IQ itself, it is easier to just ask people about educational attainment than it is to give >100,000 people an IQ test.

Because of this, it will be easier to construct polygenic scores for demographic variables. Has anyone tried to construct a polygenic score for divorce? I’m sure the data to do so is out there. Polygenic scores for self-reported alcohol consumption have been constructed. What about for one’s risk of ever getting arrested? On this, people could lie, but one could work with a background check company to determine the true records. Or what about a polygenic score for one’s credit score? The IRBs aren’t going to like some of these studies, you can be sure of that.

To measure things like honesty, one can imagine all kinds of clever experiments. Bring test subjects into a room where they will “find” a “missing” wallet. But the problem with this is that at least some of the test subjects will see through such trickery, thus reducing the test to one of IQ and familiarity with research practices. It’s easier to manipulate children, however, and, through the school system, easier to get a representative sample.(The IRBs are an ever present headache, though.) One can imagine things like the marshmallow test, used to construct polygenic scores, which are then measured against the performance of adults. If your hypothesis is that the marshmallow test will correlate with income because of genetic factors, there will be no more need to wait decades for the original test subjects to grow up to put this to the test.

Would it work? Maybe or maybe not.(The original marshmallow test didn’t work.) It’s not obvious that someone’s behavior as a child in a test environment will correlate much with their behavior in the real world as an adult. But if these studies prove unimpressive, or the IRBs stymie efforts to do them, academic fraudsters and trolls could just make up fake polygenic scores. There will be known variants which have greater weight in the polygenic scores for income or educational attainment than they do for intelligence, thus it will be known that they cause higher income or educational attainment through some trait other than intelligence as measured by IQ tests. Maybe it’s honesty, maybe it’s popularity, maybe it’s how hardworking one is. Academic fraudsters could simply take these “mystery variants” and pretend that they were “discovered” by their clever experiment, with some variants which affect neither IQ nor income thrown in to make it look believable. Trolls could “anonymously” release such “research” with the claim that it is being suppressed. They would be especially likely to get away with it if they claimed the scores were constructed using unethical (and thus hard to reproduce) experiments. For instance, they could say they or some other anonymous person did a test where children in a classroom where asked to rate the popularity of other children they know, with the resulting scores used to construct polygenic scores for popularity which predict income for adults. No IRB would approve of a replication of such an experiment.

A limiting factor here, however, will be the sample size required for such experiments. It is believable that someone secretly did an experiment on 100 children, but not on 10,000 if that is how many scientists believe will be needed to create polygenic scores for such a trait. Thus, it will depend on to what extent the algorithms will improve. If a high sample size is needed, the trolls will take to claiming the fake polygenic scores are the products of experiments in foreign countries, particularly those known to be hostile to or inaccessible by foreign media. One will wonder how much the conspiracy theorists will become interested in this subject. I could imagine claims that North Korea has factored some weird polygenic score into its Songbun system. But on the whole they won’t be very interested in it, as it will be too technical for their target audience.

Once they are created and widely known to society, trolls will use polygenic scores to harass people by leaking their scores. These scores won’t really be that meaningful on an individual level, with only weak predictive power. An IQ test will probably always work better than a polygenic IQ score in predicting real-world intellectual achievement, as it will account for the unshared environmental variance in intelligence. But that won’t really be the point. Suppose there’s a politician or celebrity you think is an idiot or dishonest or whatever. You can’t force them to take an IQ test, but you could acquire a swab of their saliva, test it, and then release the polygenic score. Or you could not do it, but claim that you did. When a couple gets married and you strongly suspect they are going to eventually get divorced,(such as, to pick a random example, Prince Harry and Meghan Markle) you could gleefully release the couple’s combined polygenic score for divorce risk. Of course, like most methods of trolling, this only works to the extent that the targets decide to be offended by the trolls’ actions. But you can count on the fact that many will be offended.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s