A Unified Approach to Quantifying Algorithmic Unfairness

Abstract

Discrimination via algorithmic decision making has received considerable attention. Prior work largely focuses on defining conditions for fairness, but does not define satisfactory measures of algorithmic unfairness. In this paper, we focus on the following question: Given two unfair algorithms, how should we determine which of the two is more unfair? Our core idea is to use existing inequality indices from economics to measure how unequally the outcomes of an algorithm benefit different individuals or groups in a population. Our work offers a justified and general framework to compare and contrast the (un)fairness of algorithmic predictors. This unifying approach enables us to quantify unfairness both at the individual and the group level. Further, our work reveals overlooked tradeoffs between different fairness notions: using our proposed measures, the overall individual-level unfairness of an algorithm can be decomposed into a between-group and a within-group component. Earlier methods are typically designed to tackle only between-group unfairness, which may be justified for legal or other reasons. However, we demonstrate that minimizing exclusively the between-group component may, in fact, increase the within-group, and hence the overall unfairness. We characterize and illustrate the tradeoffs between our measures of (un)fairness and the prediction accuracy.

Missing the missing values: The ugly duckling of fairness in machine learning

“…Because of the ubiquitous use of machine learning and artificial intelligence (AI) for decision making, there is an increasing urge in ensuring that these algorithmic decisions are fair, that is, they do not discriminate some groups over others, especially with groups that are defined over protected attributes, such as gender, race and nationality 1–7 . Despite all this growing research interest, fairness in decision making did not arise as a consequence of the use of machine‐learning and other predictive models in data science and AI 8,9 .…”

Fairness in Machine Learning: A Survey

“…[173] propose the concept of counterfactual fairness which builds on causal fairness models and is related to both individual and group fairness concepts. [252] proposes a generalized entropy index which can be parameterized for different values of α and measures the individual impact of the classification outcome. This is similar to established distribution indices such as the Gini Index in economics.…”

Quantum Fair Machine Learning

“…constant multiple K) of the metric distance d X of those individuals themselves, that is d Y (f (x 1 ), f (x 2 ) ≤ Kd X (x 1 , x 2 ). The condition is expressed typically as a constraint an optimisation problem; (b) counterfactual fairness, where given causal models (U, V, F ) (U latent background variables, V = S X are observables, S are sensitive variables and F structural equation models, fairness requires that individual outcomes are equivalent if sensitive variables had varied; (c) generalised entropy comparing individualised prediction to average prediction accuracy (see (Speicher et al 2018) where fairness measures are drawn from axiomatic approaches in economics). In the quantum setting, outcomes are also probability distributions but available metrics are more limited where metrics must compare the similarity of quantum input states (the quantum analogue of d X (x 1 , x 2 )) and final output states).…”