The tensions between explainable AI and good public policy

Explainability alone is not enough to address AI reinforcement of societal biases. Why? Performance-explainability trade-offs and unclear policy aims.