The ideas in this paper stem from original work representing aspects of human emotion in computer models. In previous exercises we have developed computer agents capable of plausibly appraising situations that arise in their world so that that they respond with a wide variety of emotion states such as pity, anger, joy, and admiration. The agents are then capable of expressing these emotions through a variety of theoretical paths, resulting in various multimedia manifestations.
When proposing to add such an affective component to tutoring systems, two primary questions must be addressed. (1) Complexity versus coherence: do we gain enough, through, e.g., interest, sensitivity, and robustness, to make the added complexity worth it? On the one hand we have current systems which do little or nothing with rich personality models, but which are clear in their objectives. On the other hand we have the proposed socially complex systems which risk having the social interaction get in the way of the pedagogical tasks, but show promise as interesting, engaging, tutors. Is there an effective balance between these two? (2) Entertainment versus pedagogical principles: While systems that entertain have the potential to be more effective at engaging students than systems that do not, many issues about the correct balance between entertainment, and pedagogical principles, arise. Is it possible for us to obviate these concerns by integrating the entertainment value into the manipulation of the content material itself using social principles?
People have built-in capabilities for understanding social and emotional complexity. Agents that are able to capitalize on these capabilities, such as those designed for the Affective Reasoner, will be able to initiate complex social interactions in the tutoring paradigm without sacrificing coherence. Furthermore, tutoring goals and principles can comprise the fabric from which an agent's emotion reasoning arises, allowing us a high degree of integration between, on the one hand, entertainment through interaction with the agents, and on the other hand, the underlying material in a content domain.