Shared Knowledge in Human-Robot Interaction (HRI)

Abstract

According to the Theory of Natural Pedagogy, object-directed emotion may provide different information depending on the context: in a communicative context, the information conveys culturally shared knowledge regarding the emotional valence of an object and is generalizable to other individuals, whereas, in a non-communicative context, information is interpreted as a subjective disposition of the person expressing the emotion, i.e., personal preference. We hypothesized that this genericity bias, already present in infants, may be a feature of human communication and, thus, present at all ages. We further questioned the effects of robotic ostensive cues. To explore these possibilities, we presented object-directed emotions in communicative and non-communicative contexts under two conditions: adult participants (N = 193) were split into those who underwent the human-demonstrator condition and those who underwent the robot-demonstrator condition, i.e., a human actor or a robot displayed object-directed emotions, respectively. Questionnaires further assessed the participants’ attachment style and mentalization ability. The results showed that (1) Natural Pedagogy Theory applies to humans across the lifespan; (2) Shared knowledge depends on the contexts (communicative vs. non-communicative) and who is sharing the information (human or robot); and (3) robotic ostensive cues trigger participants’ attention, conversely, in their absence, participants do not turn the robot into a communicative partner by not assigning it a communicative intention due to a difficulty in reading the robot’s mind. Taken together, our findings indicate that robotic ostensive cues may ease the human-robot interaction (HRI), which is also biased by the human attachment style. The study has been preregistered in Open Science Framework, OSF on September 9, 2021 (Registration DOI https://doi.org/10.17605/OSF.IO/9TWY8).