The Fairness Hierarchy: A viewpoint from causal inference

Fairness in machine learning prediction has attracted growing attention in recent years. In this article, we propose a causal–inference–based framework for fair prediction, defined through path-specific counterfactual interventions. Instead of imposing fairness via constraints on predictive objectives or model parameters, our approach specifies fairness directly at the level of counterfactual prediction semantics. Given a learned causal graph, we construct a predictive distribution for the outcome $Y$ using a structural causal model and generate counterfactual predictions by selectively intervening on causal paths emanating from sensitive attributes. By allowing or blocking the propagation of sensitive information along designated paths, possibly involving multiple sensitive sources, our framework induces a hierarchy of interpretable fairness notions, generalizing standard path-specific causal semantics. Our empirical experiments demonstrate how different fairness levels can be instantiated and compared in practice.

Machine learning systems are widely used in important decision-making tasks, but they may produce unfair outcomes across different groups of people. Understanding and improving fairness in these systems is therefore an important challenge. In this work, we study fairness from a causal perspective by examining how sensitive attributes affect outcomes through different causal pathways. This pathway-based view provides a clearer interpretation of fairness and connects several existing fairness notions through a causal decomposition of influence pathways. Overall, our goal is to advance the understanding and analysis of fairness in data-driven decision systems.