Fair Algorithms for Machine Learning

The widespread use of machine learning to make consequential decisions about individual citizens (such as those involving credit, employment, insurance, and education) has been accompanied by rising alarm over instances of bias or discrimination in the algorithms and models used. While legal, regulatory and watchdog challenges to discriminatory algorithms will play an important role, it is also crucial to examine and quantify the extent to which social norms such as fairness can be "endogenized" into the learning process itself. Can we develop a rigorous and useful science of fair machine learning?