Multiaccuracy and Multicalibration via Proxy Groups
This addresses fairness challenges in high-stakes decision-making for applications with incomplete sensitive data, representing an incremental extension of proxy methods to newer fairness frameworks.
The paper tackles the problem of ensuring fairness in predictive models when sensitive group data is missing, by showing that proxy-sensitive attributes can provide upper bounds on multiaccuracy and multicalibration violations and that adjusting models based on proxies can mitigate these violations.
As the use of predictive machine learning algorithms increases in high-stakes decision-making, it is imperative that these algorithms are fair across sensitive groups. However, measuring and enforcing fairness in real-world applications can be challenging due to the missing or incomplete sensitive group information. Proxy-sensitive attributes have been proposed as a practical and effective solution in these settings, but only for parity-based fairness notions. Knowing how to evaluate and control for fairness with missing sensitive group data for newer, different, and more flexible frameworks, such as multiaccuracy and multicalibration, remain unexplored. In this work, we address this gap by demonstrating that in the absence of sensitive group data, proxy-sensitive attributes can provably used to derive actionable upper bounds on the true multiaccuracy and multicalibration violations, providing insights into a predictive model's potential worst-case fairness violations. Additionally, we show that adjusting models to satisfy multiaccuracy and multicalibration across proxy-sensitive attributes can significantly mitigate these violations for the true, but unknown, sensitive groups. Through several experiments on real-world datasets, we illustrate that approximate multiaccuracy and multicalibration can be achieved even when sensitive group data is incomplete or unavailable.