Addressing Fairness in Classification with a Model-Agnostic Multi-Objective Algorithm
This work addresses fairness in classification for applications involving sensitive attributes like race and gender, but it is incremental as it builds on existing relaxations and multi-objective approaches.
The paper tackled the problem of fairness in classification by proposing a model-agnostic multi-objective algorithm that optimizes for multiple fairness notions and sensitive attributes, using a differentiable relaxation based on the hyperbolic tangent function. Experiments on public datasets showed that the method suffers a significantly lower loss of accuracy compared to current debiasing algorithms relative to the unconstrained model.
The goal of fairness in classification is to learn a classifier that does not discriminate against groups of individuals based on sensitive attributes, such as race and gender. One approach to designing fair algorithms is to use relaxations of fairness notions as regularization terms or in a constrained optimization problem. We observe that the hyperbolic tangent function can approximate the indicator function. We leverage this property to define a differentiable relaxation that approximates fairness notions provably better than existing relaxations. In addition, we propose a model-agnostic multi-objective architecture that can simultaneously optimize for multiple fairness notions and multiple sensitive attributes and supports all statistical parity-based notions of fairness. We use our relaxation with the multi-objective architecture to learn fair classifiers. Experiments on public datasets show that our method suffers a significantly lower loss of accuracy than current debiasing algorithms relative to the unconstrained model.