<ul><li>Researchers propose a fair machine learning algorithm to model interpretable differences between observed and desired human decision-making.</li><li>The algorithm aims to reduce disparities in a downstream outcome impacted by human decision, termed representational disparities.</li><li>A neural network is used to learn interpretable representational disparities, which could be corrected by nudges to human decision, mitigating outcome disparities.</li><li>The approach is validated using real-world datasets like German Credit, Adult, and Heritage Health.</li></ul>

Learning Representational Disparities

Discover more