Fair Learning with Private Demographic Data
arXiv:2002.11651
Abstract
Sensitive attributes such as race are rarely available to learners in real world settings as their collection is often restricted by laws and regulations. We give a scheme that allows individuals to release their sensitive information privately while still allowing any downstream entity to learn non-discriminatory predictors. We show how to adapt non-discriminatory learners to work with privatized protected attributes giving theoretical guarantees on performance. Finally, we highlight how the methodology could apply to learning fair predictors in settings where protected attributes are only available for a subset of the data.
arXiv commentICML 2020
References in corpus (12)
- Equality of Opportunity in Supervised Learning
- Extremal Mechanisms for Local Differential Privacy
- Fairness Without Demographics in Repeated Loss Minimization
- Discrete Distribution Estimation under Local Privacy
- Learning Non-Discriminatory Predictors
- Robust Optimization for Fairness with Noisy Protected Groups
- Noise-tolerant fair classification
- Assessing Algorithmic Fairness with Unobserved Protected Class Using Data Combination
- Blind Justice: Fairness with Encrypted Sensitive Attributes
- Differentially Private Fair Learning
- Equalized odds postprocessing under imperfect group information
- The Cost of a Reductions Approach to Private Fair Optimization