Helping users adopt and delegate agency to autonomous agents in everyday life
Jhim Kiel M. Verame · ePrints Soton (University of Southampton) · 2018
Autonomous agents are designed to take actions on behalf of users, acting autonomously upon data from sensors or online sources. However, the performance and actions of such agents are liable to uncertainties. As such, the design of interaction mechanisms that enable users to understand the operation of autonomous agents and flexibly delegate or regain control is an open challenge for HCI. Against this background, in this thesis we report on three studies designed to better understand how to help users interact with autonomous agents. In particular, we begin by understanding how people deal with uncertainties when delegating agency to autonomous services. We then examined the impact of different agent-designs and feedback mechanisms, inspired by the factors that encourages people to delegate agency in an everyday setting. Based on our findings, we discuss key implications for the design of future autonomous technologies and the design of interaction mechanisms to help users make the best use of such systems.