Modeling an Empathetic Embodied Conversational Agent

Abstract

Empathy can be seen as a complex socio-emotional behavior, which is a result from the interaction between both cognitive and affective devices and is responsible, for instance, for one person being able to identify and mimicry others emotion. Moreover, the human memory is a powerful tool which allows for each person to be able to store and retrieve information about almost everything that happens. Equipping an embodied conversational agent (ECAs) with the ability of empathy, as well other features like memory, can help to make the interaction with humans smoother and more natural. This work aims to propose an empathetic conversational agent endowed with a human-like memory and the ability to show some extent of empathy. Also, this model endow the virtual agent with a few other abilities, like recognizing the person it is talking to by its face and store/retrieve information based on a human memory model. Some experiments were conducted to gather both quantitative and qualitative information, which show that the proposed model works as intended. Finally, a few avenues for future work are also presented, elucidating what it is planned to do aiming to improve the quality of this work.