Acquiring and probing self-other equivalencies Using artificial agents to study social cognition

Abstract

As artificial anthropomorphic agents, such as humanoid robots and computer-generated characters, are likely to become widespread in our society, it is important to understand humans' emotional reactions to these agents. "Social resonance" is an emerging framework in social cognition covering cognitive processes used when experiencing an event and when perceiving another individual experiencing the same event. It has been applied to the domains of action, emotion and pain. After presenting this framework and discussing its use to address questions pertaining to artificial agents' social competence I will present experiments designed to test hypotheses concerning the acquisition and probe self-other equivalencies. In a first study, we tested that motor contagion can be bootstrapped by self-observation using neural network computation and robotic simulation. A complementary strategy was to use paradigms derived from experimental psychology to investigate humans' responses to artificial agents, in order to shed new light on parameters influencing resonance in behaving adults. Finally, the idea of closer ties between researchers investigating artificial agents and social cognitive neuroscience is gaining momentum is both communities, and led to the proposal that a new scientific field, android science, should be created. I will thus discuss how theoretical perspectives from cognitive sciences and resources from computer and robotic sciences can be combined to investigate social cognition and to participate to the development of socially competent artificial agents