Anti-discriminatory Privacy

The paper examines the information dynamics of privacy and discrimination to design anti-discriminatory privacy rules, especially for statistical and algorithmic discrimination. To do so, it uses empirical studies of informational anti-discriminatory rules and explores how privacy rules can overcome the limitations that these rules faced. It proposes that taste-based discrimination and statistical discrimination, a traditional distinction in economics, have the same information dynamic and should therefore be addressed similarly by privacy law. The common element between different kinds of discrimination is that, to effectively prevent them, informational rules must focus on blocking information flows that can be used to shift discrimination to other groups (e.g. former inmates versus black men). Anti-discriminatory privacy rules, in other words, should block not only undesirable information but also their proxies. The paper develops a theory on how to identify such proxies based on the cross-elasticity of information. It then applies this idea to algorithmic discrimination and proposes that the literature has so far brought legal solutions to an information problem. The paper proposes an information solution to the informational problem instead.