Appropriate feedback in asymmetric interactions

Based on the observation that human–robot interaction is often laborious because the robot's interactional abilities fail to meet the user's expectations, we argue that feedback can play a central role in regulating expectations and mitigating unnecessary disruptions in the flow of conversation. For feedback to be appropriate in this sense, it needs to take situational information into account. This idea stems from interviews with persons with hearing and mental impairments who display perceptual limitations similar to a robot. The results of these interviews indicated that, depending on the goals of the situation, people with hearing impairments used either mediation (clarification) or concealment strategies to keep the interaction going. With this idea in mind, we analyzed human–robot interactions in two different situations – more task-oriented interactions versus more socially driven interactions – and we observed different feedback behaviors in users and their reactions to the robot's behavior. We use these results to derive a scaffold for modeling appropriate feedback in asymmetric interactions (i.e., in human–robot interactions) and briefly discuss some consequences for both the design of human–robot interaction and for theories of grounding.