Time to change: deciding when to switch action plans during a social interaction

Building on the extensive cognitive science literature on thesubject, this paper introduces a model of the brain mechanisms underly-ing social interactions in humans and other primates. The fundamentalcomponents of the model are the “Action Observation” and “ActionPlanning” Systems, dedicated respectively to interpreting/recognizingthe partner’s movements and to plan actions suited to achieve certaingoals. We have implemented a version of the model including reach-ing and grasping actions, and tuned on real experimental data comingfrom human psychophysical studies. The system is able to automaticallydetect the switching point in which the Action Planning System takescontrol over the Action Observation System, overriding the automaticimitation behaviour with a complementary social response. With suchcomputational implementation we aim at validating the model and alsoat endowing an artificial agent with the ability of performing meaningfulcomplementary responses to observed actions in social scenarios.