A Sandbox Tool to Bias(Stress)-Test Fairness Algorithms
Abstract: Motivated by the growing importance of reducing unfairness in ML predictions, Fair-ML researchers have presented an extensive suite of algorithmic “fairness-enhancing” remedies. Most existing algorithms, however, are agnostic to the sources of the observed unfairness. As a result, the literature currently lacks guiding frameworks to specify conditions under which each algorithmic intervention can potentially alleviate the underpinning cause of un- fairness. To close this gap, we scrutinize the underlying biases (e.g., in the training data or design choices) that cause observational unfairness. We present the conceptual idea and a first implementation of a bias-injection sandbox tool to investigate fairness consequences of various biases and assess the effectiveness of algorithmic remedies in the presence of specific types of bias. We call this process the bias(stress)-testing of algorithmic interventions. Unlike existing toolkits, ours provides a controlled environment to counterfactually inject biases in the ML pipeline. This stylized setup offers the distinct capability of testing fairness interventions beyond observational data and against an unbiased benchmark. In particular, we can test whether a given remedy can alleviate the injected bias by comparing the predictions resulting after the intervention in the biased setting with true labels in the unbiased regime – that is, before any bias injection. We illustrate the utility of our toolkit via a proof-of-concept case study on synthetic data. Our empirical analysis showcases the type of insights that can be obtained through our simulations.
Submission Length: Long submission (more than 12 pages of main content)
Changes Since Last Submission: In addition to several changes to the exposition to reflect reviewers' suggestions, we have added a new Appendix C, containing a new case study of confounding bias for a synthetic data set generated through a causal diagram modeling the relationship between an individual’s sex, educational attainment, hours worked per week, and income in dollars. We report the effect of several fairness interventions in the presence of confounding bias in this new setting to showcase the ease the Sandbox can integrate new packages, run on a wide range of synthetic data, and evaluate multiple fairness interventions.
Assigned Action Editor: ~Aditya_Menon1
Submission Number: 721
Loading