Productionized Fairness Measurement Under Privacy Constraints

2026-06-25

AI Summary

This work addresses the challenge of evaluating algorithmic fairness under strict privacy constraints, where access to sensitive demographic attributes such as race or ethnicity is limited. To overcome this barrier, the authors propose the PPRE framework, which integrates a Bayesian-enhanced surname geocoding estimator with sparse self-reported survey data. The framework leverages privacy-enhancing technologies—including secure two-party computation, differential privacy, and additively homomorphic encryption—to enable deployable, privacy-preserving measurement of racial and ethnic fairness. The approach is successfully applied to assess fairness from both candidate and viewer perspectives, demonstrating its effectiveness and strong privacy guarantees. Furthermore, it establishes a transferable infrastructure paradigm for cross-context fairness analysis, offering a scalable solution for real-world deployment while rigorously safeguarding user privacy.

Abstract

Fairness measurements in the form of disaggregated evaluations often rely on demographic signals that are legally constrained or culturally sensitive. Race and ethnicity signals are among the more difficult signals to curate and use for this task. This paper presents Privacy-Preserving Probabilistic Race/Ethnicity Estimation (PPRE) as a method for enabling fairness measurements with respect to race/ethnicity for U.S.\ LinkedIn members in a privacy-preserving manner. PPRE applies privacy technologies (specifically: secure two-party computation, differential privacy, and additive homomorphic encryption) on top of two race/ethnicity demographic signal sources (the Bayesian Improved Surname Geocoding estimator and a sparse golden survey set of self-reported demographics) to power a fairness measurement solution with respect to US-based race/ethnicity demographics. We detail its privacy guarantees and demonstrate its application on candidate- and viewer-side fairness measurements. We close with a transferable framework for institutions seeking to implement similar privacy-preserving measurement infrastructure.

Problem

• fairness measurement
• privacy constraints
• race/ethnicity
• demographic signals
• sensitive data

Innovation

• Privacy-Preserving Computation
• Fairness Measurement
• Race/Ethnicity Estimation
• Differential Privacy
• Secure Two-Party Computation