A Computational Model of Empathy for Interactive Agents

Abstract

Empathy has been defined in the scientific literature as the capacity to relate another’s emotional state and assigned to a broad spectrum of cognitive and behavioral abilities. Advances in neuroscience, psychology and ethology made it possible to refine the defined functions of empathy to reach a working definition and a model of empathy. Recently, cognitive science and artificial intelligence communities made attempts to model empathy in artificial agents, which can provide means to test these models and hypotheses. A computational model of empathy not only would help to advance the technological artifacts to be more socially compatible, but also understand the empathy mechanisms, test theories, and address the ethics and morality problems the Artificial Intelligence (AI) community is facing today. In this paper, we will review the empathy research from various fields, gather the requirements for empathic capacity and construct a model of empathy that is suitable for interactive conversational agents.