In this paper we examine ways in which work on the Affective Reasoner, being developed at DePaul University, can be integrated into two extant tutoring systems, the Soar Training Expert for Virtual Environments (STEVE) at the Information Sciences Institute, and the Design-a-Plant (Herman the Bug) project, at North Carolina State University. In the paper we present preliminary design elements of an architecture for including personality and emotional responsiveness in virtual tutoring agents. Part of the paper is high-level discussion, arguing for this as a highly plausible addition to tutoring systems, and presenting motivation for such work; part of it is more technical, serving as initial examples of a design map for applications work we hope to have made progress with before the upcoming IJCAI workshop.
Topics included in the architecture overview include basic emotion architectures for agents; contrasting agent personalities for presenting the same material from different perspectives; models for different relationships with the users; multiple-agent architectures for utilizing simultaneous, but contrasting, motivational techniques; affective user modeling.
Topics in the preliminary design of the two applications systems include the mapping of actual system events as antecedents to plausible emotion responses; discussion of system variables which might be mapped to emotion intensity; and how observed user states my be internally represented by the agents in the agents' emotion reasoning fabric.