Fair Bayes-Optimal Classifiers Under Predictive Parity

Neural Information Processing Systems 

Increasing concerns about disparate effects of AI have motivated a great deal of work on fair machine learning.