User Perceptions of Responsibility and Autonomy of AI Consent Behaviours

Abstract

Responsibility and autonomy are central to effective human-artificial intelligence (AI) collaboration. The behaviours of (e.g., providing, revoking, or altering) consent empower a human in modulating the levels of autonomy and responsibility they ascribe to an agent. However, how the human users perceive the responsibility and autonomy resulting from different consent behaviours is largely unexplored in the literature. We conduct a vignette-based user study (n=138) to explore the relationship between consent behaviours, and responsibility and autonomy. Our study includes consent behaviours derived from a state-of-the-art consent model, four aspects of responsibility, and three aspects of autonomy. Our findings suggest that the consent behaviours of the agent influence causal perceptions of responsibility and all elements of autonomy perception. Based on these findings, we draw design guidelines for AI agents to exhibit responsible autonomy.