Explaining Quantitative Measures of Fairness
Quantifying the fairness of a machine learning model has recently received considerable attention in the research community, and many quantitative fairness metrics have been proposed. In parallel to this work on fairness, explaining the outputs of a machine learning model has also received considerable research attention. Here we connect explainability methods with fairness measures and show how recent explainability methods can enhance the usefulness of quantitative fairness metrics by decomposing them among the model's input features. Explaining quantitative fairness metrics can reduce our tendency to rely on them as opaque standards of fairness, and instead promote their informed use as tools for understanding model behavior between groups.
Lundberg, S. M. Explaining Quantitative Measures of Fairness.
Lundberg, Scott M. “Explaining Quantitative Measures of Fairness” (n.d.).
Lundberg, Scott M. Explaining Quantitative Measures of Fairness.