20 Jan The Ethics of GPS
First, a note from Twitter. Several people (eg Michel Nivard and Ruben Arslan) thought I was being a little “misleading” in my characterization of what GPS can do as opposed to twin and family studies. That may be; I can get carried away when I am pushing back against over-interpretation of genetic findings. I was thinking, for example of children-of-twins studies I have done about the effects of parental divorce on children. Twin parents who are discordant for marital status offer some controls for the obvious genetic confounds inherent in simply comparing children from divorced and non-divorced families. But those studies are limited because only one parent is a twin; the spouse contributes different, uncontrolled genetic variance. In a large enough sample you could use GPS to control for genetic background in both parents, and for that matter in the kids. It’s not a new kind of knowledge, but it is a new (and with the right data improved) method for obtaining similar knowledge. I am not anti-GPS any more than I am anti-twin-study.
What I do oppose is the interpretation that a GPS for intelligence is an estimate of an individual’s “genetic propensity for intelligence” or anything like that. Genetically-based predictors are not measures of genetic propensity. The ginger child makes the point: a GPS is a measure of SNPs correlated with anything and everything that produces differences in IQ, whether those processes are internal to the individual (myelination or whatever), or external (exposure to lead paint).
Here is another way to say it. When comparing GPS to social indicators like SES, PvS point out,
However, [the SES-IQ] association is confounded by genetics because children inherit the DNA differences that predict their intelligence from their parents.
True enough, but what they don’t acknowledge is that in the exact same way, GPS are confounded by the environment. This is crucially important: GPS, although estimated in the genome, are not essentially genetic. They are a summary of genetic correlations of all associations, genetic or environmental, immutable or controllable, with the phenotype of interest.
Anyway, on to ethics. PvS’s section on ethics, headed “Understanding Ourselves” begins by getting this point wrong:
IQ GPSs will be used to predict individuals’ genetic propensity to learn, reason and solve problems, not only in research but also in society.
Like I said about theory of causation, this is not just a exercise in theoretical bioethics. PvS go on to recommend some truly dreadful social policies, for which the only possible justification is that they believe their GPS is a measure of children’s genetic propensity for IQ. Quoting them at length:
GPSs for intelligence are more than idle fortune telling. Because intelligence is one of the best predictors of educational and occupational outcomes, IQ GPSs will be used for prediction from early in life before intelligence or educational achievement can be assessed. In the school years, IQ GPSs could be used to assess discrepancies between GPSs and educational achievement (that is, GPS-based overachievement and underachievement). The reliability, stability and lack of bias of GPSs make them ideal for prediction, which is essential for the prevention of problems before they occur. A ‘precision education’ based on GPSs could be used to customize education, analogous to ‘precision medicine’.
It sounds reasonable, but think about it for a minute. We are going to assign children to classrooms and curricula (precision education!) based on their GPS? That would be the worst possible kind of biologically determinist discrimination. If you want to assign children based on their IQ, maybe, but why not just use the phenotype? What does a GPS add to the equation, unless you believe that it represents a genetically determined “propensity” for learning? I can hear the poor mother: “No, no, Johnny is really a very intelligent little boy, it’s just that he has red hair! Everyone picks on him!” “I’m sorry, Ma’am, the GPS shows your son has a low genetic propensity to learn.”
Once again, all of this turns on the ginger child and the absence of knowledge of mechanism. Suppose someone recommended a program to use genetic testing for trisomy to identify children with Down’s syndrome, so the parents and school system could start making plans for the special needs the child is certain to have. Who could object, and what is the difference with GPS? The difference is that our knowledge of the biology of trisomy allows us to rule out the possibility that something external and modifiable is the cause of both the low genetic score and the reduced IQ. For highly polygenic complex traits like IQ, there is not only no way to know this, but in fact we know perfectly well that a million such external factors do exist. A GPS does nothing to rule any of them out.
Or try it this way. If what you want is some kind of genetic estimate of a child’s IQ to use in making educational decisions, why not just use the parents’ IQs? Parental IQ is much more correlated with child IQ than a GPS will ever be, and the association is by and large “genetic.” Wouldn’t it be pretty unfair, pretty bio-determinist, eugenic, and dystopian, to assign children to slow classrooms because their parents have low IQs? It is no different with a GPS. It’s a modern way of stating the principle in the second paragraph of the American Declaration of Independence: individuals have a right to be judged on their phenotype, not their genotype. That’s what it means to be created equal in a world that contains individual differences.