Fast Threshold Tests for Detecting Discrimination

By Emma Pierson, Sam Corbett-Davies, Sharad Goel. 2017

53 citations recorded by the source corpus.

Threshold tests have recently been proposed as a useful method for detecting bias in lending, hiring, and policing decisions. For example, in the case of credit extensions, these tests aim to estimate the bar for granting loans to white and minority applicants, with a higher inferred threshold for minorities indicative of discrimination. This technique, however, requires fitting a complex Bayesian latent variable model for which inference is often computationally challenging. Here we develop a method for fitting threshold tests that is two orders of magnitude faster than the existing approach, reducing computation from hours to minutes. To achieve these performance gains, we introduce and analyze a flexible family of probability distributions on the interval [0, 1] -- which we call discriminant distributions -- that is computationally efficient to work with. We demonstrate our technique by analyzing 2.7 million police stops of pedestrians in New York City.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Algorithmic Decision Making and the Cost of Fairness (2017, background): “in the econometric literature on taste-based discrimination [3, 4, 34, 37].”
The Problem of Infra-marginality in Outcome Tests for Discrimination (2016, background): “…jurisdictions to simultaneously infer signal distributions and thresholds for all 6A variant of the threshold test has recently been proposed by Pierson et al. (2017) which can accelerate inference by more than two orders of magnitude, allowing the test to scale to even larger…”
A large-scale analysis of racial disparities in police stops across the United States (2020, methodology): “[20], which we fit separately on state patrol and municipal police stops.”
The Measure and Mismeasure of Fairness: A Critical Review of Fair Machine Learning (2018, methodology): “The ﬁgure was generated by assuming risk scores follow a discriminant distribution (Pierson et al., 2018), a type of logit-normal mixture that is well-suited to modeling risk distributions because it is naturally parameterized in terms of its mean and AUC.”
Repairing without Retraining: Avoiding Disparate Impact with Counterfactual Distributions (2019, background): “…focusing on how disparate impact arises (Chen et al., 2018; Datta et al., 2016; Adler et al., 2018), how it can be measured (Žliobait˙e, 2017; Pierson et al., 2017; Simoiu et al., 2017; Kilbertus et al., 2017; Kusner et al., 2017; Galhotra et al., 2017), and how it can be mitigated (Feldman…”