SDT: Maintaining the Communication Protocol Through Mixed Feedback Strategies

Abstract

In our previous work, we studied how humans establish a protocol of communication in a context that requires mutual adaptation using our robot Sociable Dining Table (SDT). SDT integrates a dish robot put on the table and behaves according to the knocks that the human emits by guessing the meaning of each knocking pattern. We remarked based on previous experiments, that a personalized communication protocol is established incrementally. In fact, the communication protocol is not personalized only to the pair human-robot but also to the human robot interaction’s (HRI) instance. In the current study, we change the robot’s feedback modality (the way the robot communicates back with the human) in order that the communication protocol can be maintained over different HRI’s instances. We proposed as new feedback modality, 2 mixed-feedback strategies integrating inarticulate utterances (IU) combined with the robot’s visible behavior in order to facilitate the guessing of the robot’s internal state for the human. The first strategy consisted in anticipating the robot’s executed behavior using static IU combined with the robot’s movement (St1), and the other consisted in genuinely suggesting an adaptive IU generation method combined with the robot’s movement too (St2). In the current work, we conducted an HRI experiment to explore whether the communication protocol can be maintained on a long-term basis by integrating the 2 proposed methods. The results provide confirmatory evidence that using IU helps in establishing stable communication protocols. In addition to that it increases the attachment and the robot’s overall subjective ratings. Another important finding is that, among the two methods, the adaptive mixed feedback strategy (St2) affords better subjective results and objective performance.