Fair Classification with Noisy Protected Attributes: A Framework with Provable Guarantees
This addresses fairness in machine learning for applications where protected attributes are imperfectly measured, offering a practical solution with theoretical backing, though it builds incrementally on prior work by extending to more general constraints and noisy settings.
The paper tackles the problem of learning fair classifiers when protected attributes are noisy, presenting an optimization framework that works with various fairness constraints and multiple attributes, and provides provable guarantees on accuracy and fairness. Empirically, it achieves fairness guarantees like statistical rate or false positive rate with minimal accuracy loss on real-world datasets, even under large noise.
We present an optimization framework for learning a fair classifier in the presence of noisy perturbations in the protected attributes. Compared to prior work, our framework can be employed with a very general class of linear and linear-fractional fairness constraints, can handle multiple, non-binary protected attributes, and outputs a classifier that comes with provable guarantees on both accuracy and fairness. Empirically, we show that our framework can be used to attain either statistical rate or false positive rate fairness guarantees with a minimal loss in accuracy, even when the noise is large, in two real-world datasets.