Balancing Fairness and Accuracy in Data-Restricted Binary Classification

TL;DR

This work proposes a framework that explicitly models fairness-accuracy tradeoffs in data-restricted environments and evaluates the behavior of the optimal Bayesian classifier using a discrete approximation of the data distribution, allowing it to systematically isolate the effects of fairness constraints.

Abstract

Applications that deal with sensitive information may have restrictions placed on the data available to a machine learning (ML) classifier. For example, in some applications, a classifier may not have direct access to sensitive attributes, affecting its ability to produce accurate and fair decisions. This paper proposes a framework that models the trade-off between accuracy and fairness under four practical scenarios that dictate the type of data available for analysis. Prior works examine this trade-off by analyzing the outputs of a scoring function that has been trained to implicitly learn the underlying distribution of the feature vector, class label, and sensitive attribute of a dataset. In contrast, our framework directly analyzes the behavior of the optimal Bayesian classifier on this underlying distribution by constructing a discrete approximation it from the dataset itself. This approach enables us to formulate multiple convex optimization problems, which allow us to answer the question: How is the accuracy of a Bayesian classifier affected in different data restricting scenarios when constrained to be fair? Analysis is performed on a set of fairness definitions that include group and individual fairness. Experiments on three datasets demonstrate the utility of the proposed framework as a tool for quantifying the trade-offs among different fairness notions and their distributional dependencies.

Figures (6)

• Figure 1: Outline of proposed framework for trade-off analysis. (a) A discrete approximation of the population distribution is constructed by using a generator to densely sample it and applying vector quantization. (b) The accuracy-fairness trade-off is analyzed under four data-restricting situations in which the sensitive attribute is (is not) available and the features used for classification are (are not) required to be decorrelated from it.
• Figure 2: Pareto frontiers capturing the accuracy-fairness trade-off for three datesets under awareness and unawareness of the sensitive attribute. Each plot provides curves for different pairings of fairness constraints; namely, DP, EA, PE, EOd, and Ind.
• Figure 3: Dimensionality reduction of feature vectors.
• Figure 4: Visualization of the original and decorrelated feature vector distributions when the sensitive attribute is and is not used for decorrelation.
• Figure 5: Pareto frontiers capturing the accuracy-fairness trade-off for three datesets under unawareness of the sensitive attribute. Each plot provides curves for different pairings of fairness constraints; namely, DP, EA, PE, EOd, and Ind.

Theorems & Definitions (15)

• Definition 3.1
• Definition 3.2
• Definition 3.3
• Definition 3.4
• Definition 3.5
• Definition 3.6
• Definition 3.7
• Definition 3.8
• Claim 4.1
• proof