Title: Noisy Stereotypes
Presented By: Boris Babic, New York University

Abstract:
In a world characterized by socioeconomic and other inequalities, some stereotypes will be statistically sound. In those cases, many philosophers have argued, epistemic rationality can come apart from our moral obligations to each other. For example, Tamar Gendler puts the point as follows: "As long as there’s a differential crime rate between racial groups, a perfectly rational decision maker will manifest different behaviors, explicit and implicit, toward members of different races. This is a profound cost: living in a society structured by race appears to make it impossible to be both rational and equitable." In a recent paper, Babic and Johnson-King argue that such normative conflicts can usually be avoided: what ordinary morality demands, epistemic rationality typically permits. In this project, however, we explain that as data gets large, Babic and Johnson- King's approach to avoiding normative conflicts becomes less persuasive. More constructively, we build on their project and develop an alternative model of reasoning about stereotypes under which one can indeed avoid normative conflicts, even in a big data world, but only if such data contains some noise. Fortunately for those seeking to avoid normative conflicts, big data is typically noisy.