Fairlearn: A toolkit for assessing and improving fairness in AI

Abstract

We introduce Fairlearn, an open source toolkit that empowers data scientists and developers to assess and improve the fairness of their AI systems. Fairlearn has two components: an interactive visualization dashboard and unfairness mitigation algorithms. These components are designed to help with navigating trade-offs between fairness and model performance. We emphasize that prioritizing fairness in AI systems is a sociotechnical challenge. Because there are many complex sources of unfairness—some societal and some technical—it is not possible to fully “debias” a system or to guarantee fairness; the goal is to mitigate fairness-related harms as much as possible. As Fairlearn grows to include additional fairness metrics, unfairness mitigation algorithms, and visualization capabilities, we hope that it will be shaped by a diverse community of stakeholders, ranging from data scientists, developers, and business decision makers to the people whose lives may be affected by the predictions of AI systems.