The NY Times has published an interesting piece about Cambridge Analytica that raises several issues related to heterogeneous treatment effects (hang in here for a few sentences!). The key idea here is “tailoring”. If everyone was 6 feet tall, then a tailor would have an easy job. In a world where everyone's body shape is different, then the tailor will confront many challenges and will have a lot of work to do.
Take this analogy to medicine. Precision medicine seeks to tailor the right drug therapy to each person where a person has a specific life history, genetic disposition and already is taking a given subset of medicines. In this case, a Big Data scientist seeks to predict the interactive effects of adding a new medication to the patient's drug intake before the patient takes such a drug. The same drug may have different effects on different people depending on a whole set of observable attributes. This tailoring raises the likelihood that doctors can improve our quality of life using the Big Data they can now access. Why?
Suppose there are 10 drugs and they interact in strange ways that we do not know about. So if you take drug 3 and drug 7, suppose that this combination makes you vomit once a week. The doctor would want to know this before he prescribes this. With 10 drugs, there are 2^10 combinations of outcomes. To study non-parametrically the effects of all of these combinations will require a large set of people. If you then say that men and women have different reactions to the drugs, that whites and non-whites have different reactions; then you need huge data sets to engage in precision medicine.
Now, the Cambridge Analytica article raises all of these same issues but now focused on politics. Rather than trying to make a person healthier, the piece is bout how political operatives can use treatments (such as advertisements about robbers) to rev up different types of people (left wing, right wing). This is a form of tailored advertising. In the past, a George Bush would put a single “Willie Horton Ad” on the news and everyone would see it. Facebook and social media platforms have much more tailored ads and the NY Times is hinting that our democracy will be affected by such tailoring.
So, how can precision medicine be good for us but precision politics will be bad? Would democracy work better if there weren't such tailored ads? The NY Times must believe that the treatment effects of being exposed to such ads are huge. While we want medicines to be effective, is it the case that we do not want political influencers to be effective? Are we back to concerns about “manipulation”? Galbraith wrote that Madison Avenue Ad Men create our preferences. Is that now the case for voting? Would election results be more “authentic” without such tailored influence?