Self-explanations of a cognitive agent by citing goals and emotions

Abstract

This paper presents a cognitive (belief-desire-intention based) agent that can self-explain its behaviour based on its goals and emotions. We implement a cognitive agent, embodied by a nao-robot or virtual avatar thereof, to play a quiz with its user. During the interaction the agent intelligently selects questions to optimally educate the user. We show how the simulation of emotions can be used to generate end-user explanations of the agent's behaviour. With this we provide a first proof of concept showing the value of using simulated emotions in addition to goals for generating agent behaviour explanations.