Exploring the Impact of Confirmation and Interaction During Human-Robot Collaboration with a Proactive Robot Assistant

Abstract

Robots should not make us feel uncomfortable in our own homes, and so we must be able to trust them. Proactive robot assistants, designed to collaborate with humans on Activities of Daily Living (ADLs), operate in a tricky social situation as they emulate complex human-human interactions. This paper considers how humans expect robots to interact in such situations, particularly with regard to action confirmations, and how this impacts trust. A video-based user study was conducted with over 100 participants, comparing three distinct interaction personas. Our findings highlight that: (i) while communication and explanation are a significant factor in improving trust, it may be necessary in many cases to actually confirm actions before executing them; and (ii) there is a divide between individuals who value efficiency and speed versus those who put control above all else.