A human-in-the-loop system is one where a human makes or strongly influences the final decision. For example, in some safety-ciritical situations, such as medicine or air traffic control, there may be professional or regulatory reasons why a human must oversee a final decision. It is particularly important in autonomous weapons, where many argue that the final decison to fire should come from a human. Sometimes the term is applied more generally when the AI includes human participants, for example online AI-driven chat which may pass more complex queries to a human, altough arguably this is more properly regarded as human computation.
Used in Chap. 17: page 280; Chap. 19: pages 316, 326, 329, 331; Chap. 20: pages 334, 345; Chap. 23: page 385
Also known as human in the loop