“Causal Inference in Complex Non-linear Systems: A Cautionary Tale” Eric Winsberg
Non-linear phenomena like climate systems and epidemics display rich patterns of behavior across various temporal and spatial scales. These rich patterns often produce in students of these phenomena the temptation to attribute to them underlying causes. In this talk, Professor Eric Winsberg reviews examples of this propensity from the fields of climate change and epidemiology by experts and lay people alike, proving that causal inference in these contexts is especially epistemically perilous. He discusses, in particular, a method of causal attribution that has been especially popular during the COVID-19 epidemic: the method of synthetic controls, and he will explain why, despite appearances to the contrary, it is not a magic bullet for overcoming these problems.
