Characterizing Fairness Over the Set of Good Models Under Selective Labels
This work is significant for practitioners and auditors in high-stakes decision-making, providing a method to either replace existing models with fairer ones or audit for predictive bias, especially in scenarios with selective labels.
This paper addresses the Rashomon Effect in algorithmic risk assessments, where multiple models show similar overall performance but vary in individual predictions and fairness properties. The authors developed a framework to characterize predictive fairness across the set of 'good models,' specifically tackling the challenge of selectively labeled data where selection and outcome are unconfounded given observed features.
Algorithmic risk assessments are used to inform decisions in a wide variety of high-stakes settings. Often multiple predictive models deliver similar overall performance but differ markedly in their predictions for individual cases, an empirical phenomenon known as the "Rashomon Effect." These models may have different properties over various groups, and therefore have different predictive fairness properties. We develop a framework for characterizing predictive fairness properties over the set of models that deliver similar overall performance, or "the set of good models." Our framework addresses the empirically relevant challenge of selectively labelled data in the setting where the selection decision and outcome are unconfounded given the observed data features. Our framework can be used to 1) replace an existing model with one that has better fairness properties; or 2) audit for predictive bias. We illustrate these uses cases on a real-world credit-scoring task and a recidivism prediction task.