Fair Inference On Outcomes
This work addresses fairness issues in machine learning and statistics for applications like healthcare and policy-making, offering a generalized causal framework that builds on existing theories.
The paper tackles the problem of ensuring fairness in statistical inference on outcomes, such as classification and treatment effect estimation, by formalizing discrimination as the effect of sensitive covariates along specific causal pathways. It proposes learning fair outcome models through constrained optimization and addresses related inference complications using causal and semi-parametric methods.
In this paper, we consider the problem of fair statistical inference involving outcome variables. Examples include classification and regression problems, and estimating treatment effects in randomized trials or observational data. The issue of fairness arises in such problems where some covariates or treatments are "sensitive," in the sense of having potential of creating discrimination. In this paper, we argue that the presence of discrimination can be formalized in a sensible way as the presence of an effect of a sensitive covariate on the outcome along certain causal pathways, a view which generalizes (Pearl, 2009). A fair outcome model can then be learned by solving a constrained optimization problem. We discuss a number of complications that arise in classical statistical inference due to this view and provide workarounds based on recent work in causal and semi-parametric inference.