when the structures that fund, produce, evaluate, and reward knowledge are designed in ways that systematically bias outcomes toward error. Unlike the sunk cost of consensus (Chapter 9), which maintains error through the cost of changing, and the replication problem (Chapter 10), which maintains err