People’s responses to autonomous and adaptive systems

摘要

Computer systems are no longer passive tools; instead a growing selection of systems can learn, adapt and make decisions that previously required human intervention. This PhD thesis investigates people’s interaction with such autonomous and adaptive systems. The systems investigated range from spam filters that learn, to social robots that interact with people in ways similar to human-to-human interaction. When systems act in more autonomous ways, they appear harder to understand and become less predictable. They require users to hand over control; users have to put their trust in the system. This thesis presents six user studies and experiments, which explore the complexities of interacting with systems that to a certain extent appear to make their own decisions. Focus is on two themes: the fundamental conflict of system autonomy and the user’s perception of control, and the potential of social behaviours in interaction with highly complex, ‘intelligent’ systems.