How Movements of a Non-Humanoid Robot Affect Emotional Perceptions and Trust

The way that we move often carries emotional meaning that the people with whom we interact are adept at detecting. Humans treat the movements of robots similarly, attributing the same emotions when the robots move in ways that are analogous to emotionally-charged human movements. However, this HRI work has primarily been done on humanoid or animal-shaped robots. In this paper, we examined whether this effect would hold when people observed the movements of a non-humanoid robot, Cozmo. Moreover, the attribution of emotional stance to another agent is key in the process of predicting the behavior of the other (Eivers AR et al. in Br J Dev Psychol 28:499–504, 2010). This process is laid bare in transactional scenarios where the predicted level of trust guides the humans behavior. The ultimatum game is a transactional framework that we have adapted to allow us to test in stages how humans predict and react to the behavior of the robot. We performed a study in which people played two rounds of the ultimatum game with a non-humanoid robot that moved in either a positive or negative manner. We found that in both rounds people in the Positive Movement condition rated Cozmo’s emotional valence as higher than those in the Negative Movement condition. In the second round, after Cozmo had responded to the first offers that the participants made, Cozmo’s bid response was a significant factor in the Positive Movement condition, in which participants whose first bids were rejected by Cozmo rated its emotional valence as lower than those whose bids were accepted by Cozmo. There was not an effect of movement on trust. We also ran a series of exploratory analyses to explore how various factors affected participants’ reasonings about Cozmo’s behavior, and found that unexpected, non-social behaviors (such as moving in a negative manner or rejecting a participant’s offer) lead to an increase in anthropomorphic behavior explanations.