Inherent Trade-Offs in the Fair Determination of Risk Scores

Kleinberg, Mullainathan, and Raghavan (2016) formalize three competing fairness conditions in algorithmic classification and prove that except in highly constrained special cases, no method can satisfy these three conditions simultaneously. This foundational work establishes that practitioners must navigate unavoidable trade-offs when designing risk-scoring algorithms, with competing notions like predictive parity, calibration, and equalized false positive rates unable to be satisfied simultaneously. Accepted to ITCS 2017.