Prof. Scott Sheffield

Title: Pandemics and paradox

Abstract: In one of the simplest epidemic models, one lets p_n denote the number of new infections during week n and assumes that (during the early stages of the epidemic) p_{n+1} = R_0 p_n c_n where c_n measures the “fraction of usual contact” that takes place between people during the nth week. Within this simplistic model, intermittent strategies (taking c_n small some weeks and large other weeks) lead to lower infection rates than consistent strategies with the same total amount of contact.

But what happens in more life-like models (such as SEIR and its network-based variants)? What factors cause intermittent strategies to outperform consistent strategies? What changes when there are multiple subpopulations with different controls? I will discuss some work with public health researchers that explores these questions for simple examples. Some of the models are counterintuitive: the strategy one expects to be “best possible” turns out to be “worst possible” and vice versa.