Transparent Artificial Intelligence: Technical Explainability and Its Troubles
Argues that the aims of AI ethics are better served by social transparency (openness about how, by whom, and for what a system is used) than by technical explainability, which carries risks of its own.