Play with me: complexity of human-robot interaction affects individuals' variability in intentionality attribution towards robots

Abstract

During daily interactions with other people, humans spontaneously formulate representations of others’ goals, desires, and intentions. Past research argued that even artificial agents that can closely imitate humans’ behavior might induce humans to form these representations. We compared three human-robot interaction (HRI) experiments that explored if the behavior of a humanoid can modulate the user’s attribution of intentions towards the machine. We investigated how different metrics of the same test (InStance Test, IST) can provide useful information about individuals’ tendencies to attribute intentional states to the iCub robot. Our results show that taking into account the variability of responses at the subject level provides useful information, which can be combined with average responses to have a comprehensive understanding of pre- and post-interaction changes. Our results also suggest that taking into account data dispersion of self-report scales could improve our understanding of HRI effects on individuals’ attitudes towards robots