Empirical Price of Fairness in Failure-Aware Kidney Exchange

Fielded kidney exchanges typically use utilitarian or near-utilitarian matching rules, potentially at great cost to certain classes of hardto-match patients. Dickerson, Procaccia, and Sandholm [6] recently adapted the price of fairness, a measure of the tradeoff between fairness and efficiency, to kidney exchange; they showed that the price of fairness is small in theory but often non-negligible in practice. We extend their work by directly comparing fairness in two models of kidney exchange—where edges either can or cannot fail after an algorithmic match but before transplantation. On real and simulated data, even matching under strict fairness constraints in the model with edge failure results in significantly more expected transplants than matching efficiently in the deterministic model.