Transparency using audio - the muttering robot

Transparency is an important design consideration for all intelligent autonomous systems. The previous work shows that a real-time visual display of a robot's decisionmaking produces significantly better understanding of that robot's intelligence. We investigate vocalisation of behaviour selection as a possible alternative solution for situations where a visual display of decision-making is either impractical or impossible. In this experiment, we find that vocalisation is associated with a significant improvement in understanding of the robot, comparable with the results obtained using a real-time display. We also find that vocalisation has no significant effect on participants' emotional response, though it may slightly increase positive feelings about the robot. We discuss the relative merits of visual and vocalised transparency mechanisms and suggest possible applications.