Suppose data scientists can profile who are the likely super-spreaders based on Observables "X". Define P(X) as one's probability of being a super-spreader. Let Vaccine(X) = probability Mr X wants the vaccine. The key parameter = corr(P(X),Vaccine(X)). https://www.nytimes.com/2020/08/03/opinion/coronavirus-vaccine-efficacy-trials.html?smid=tw-share

Suppose that super-spreaders refuse to take the vaccine -- so corr(P(X),Vaccine(X) <0, then we have a big problem. If there is a sharp positive correlation between these two variables, then the infection rate will decline sharply as the vaccination begins.

In allocating the vaccine, what minimization problem are we solving? In this Thread, I am focused on minimizing the negative externality of contagion. How do we stop super-spreaders? Vaccinate them first. Is it incentive compatible for them to reveal their "type" to society?