Explainable Human-AI Interaction
About This Book
Since its earliest days, artificial intelligence has swung in how it regards people, sometimes aiming to extend human abilities and at other times to take their place. As AI tools move into daily routines at an ever-faster pace, these systems increasingly need to cooperate with the people they serve. A central demand of that cooperation is that an AI's behavior can be understood by the humans working alongside it. Meeting this demand requires agents that reason about people and not only about their own picture of the world. At minimum, an agent needs approximations of a person's tasks and goals, along with a model of how that person sees the agent's own tasks and goals. The first helps the agent anticipate and manage what people need, want, and pay attention to. The second lets the agent act in ways people can interpret, matching what they expect, and offer tailored explanations when they are needed. Drawing on several years of work from the authors' research group, the book examines how an agent can use these mental models either to meet human expectations or to reshape them through explanatory communication. The main emphasis is on cooperative settings, but it also explores how the same models could be turned toward concealment and deception. The book describes several real-world collaborative decision-making systems built on the framework and techniques it develops. Although most of the material comes from the authors' own studies, each chapter connects the work to related research in the wider literature. The technical content is self-contained, so readers with a basic background in AI can follow it.
ISBN: 9781636392905
You might also like
Unlock Exclusive Content
Experience the hidden layers of 'Explainable Human-AI Interaction'. Scan the book cover with the Appilog mobile app to see behind-the-scenes videos, author reviews, and interactive digital stories.
Available on iOS & Android
