Enhancing Robot Explainability in Human-Robot Collaboration

While the explainability of AI is gaining increasing attention from scholars, much is not known about how to enhance robot explainability and its implications for human-robot collaboration. This study sought to understand the impacts of social cues on robot explainability and trust and acceptance of a robotic partner. We proposed a research model in which the non-verbal, verbal, and joint presence of both cues predicted robot explainability, and the moderating role of robot anthropomorphic design was examined. We also proposed that robot explainability promoted trust and acceptance of a robot. We provide evidence for the research model through a mixed-design experiment with 202 individuals. While non-verbal, verbal, and joint presence generally increased robot explainability, we further found that the impact of non-verbal cues was contingent upon the existence of verbal cues and moderated by robot anthropomorphic design. Our findings provide several implications for research and practice.