HPS/LCSSP Seminar

Abstract: One of the predominate conceptions of discrimination in law and social science defines it causally. Something is discrimination on the basis of, e.g., race or sex only if the outcome would not have occurred had the person been, or been perceived to be, a different racial or sex category. This counterfactual definition has been embraced by the United States Supreme Court in interpreting federal anti-discrimination statutes and the Equal Protection Clause, and it structures much of the methodological work of social scientists, causal inference researchers, and computer scientists who design studies and make policy-relevant empirical claims. But this definition is indeterminate and it suppresses assumptions that must ground any definition of discrimination. Using examples from law and social science, I show how the causal definition does not yield determinate designations of discrimination so much as it expresses the analyst's prior sociological and normative commitments. I close by arguing that any account of discrimination must confront two foundational questions. The first is sociological: What is the category on the basis of which the conduct is said to be discriminatory? The second is normative: What constitutes fair or just treatment in light of the nature of that category? Disagreement over these questions is currently obscured by conclusory doctrinal labels or methodological terms. Because any definition of discrimination inevitably embeds answers to these questions, those answers must be made explicit.