In the Affective Reasoner, we refer to temporary changes in an agent's disposition as moods.
Good teachers present the same material in different ways. They are often not always predictable (even if the subject domain is). That AR agents support different moods, which allow for different modes of behavior, but which neither permanently, nor globally, effect an agent's personality, allows us to explore variety in any given agent's approach to a student, without losing coherence. That is: an agent's goals, principles, and modes of expression, are all still present, but tend to be activated at different different thresholds of input, so that while still recognizable, an agent can vary its relationship to the situations that arise in the course of a tutoring session.
Moods can be used to effect in several ways. First, these can be used to increase the lifelike qualities of the agents. For example, if a student repeatedly does poorly on a set of tasks, and it is important to the tutoring agent that they do well, the agent might reasonably be expected to become somewhat depressed about this. Second, such variety may increase the appeal of the agents, and add to their entertainment value. Third, moods can be used for pedagogical effect. For example, if it is important for pedagogical reasons that a student not spend too much time on a particular set of tasks, then an agent can grow tired of the tedium of spending too much time on those tasks, and grow increasingly crabby. Similarly, suppose that if, e.g., Agent-Sarah's goals are repeatedly met by a student she might start telling entertaining jokes, with the idea being that it is more realistic to interact with an agent that tells jokes because she feels good, rather than one that simply always rewards achievement on domain tasks.
With agents that have moods, it is also then possible to ``fake'' these as pedagogical ploys. For example, although it is sometimes important to be particularly demanding of a student, this has to be done with care. A mood-aware agent might use a ploy such as the following, ``Sorry Roy - I was in a bad mood when I made up this next set of exercises. It is pretty hard. I am not in such a bad mood now, but to avoid screwing up the accounting of your progress we have to do this now. I'll help you with it.''
Similarly in a ``patient'' (good) mood, Agent-Sarah might have an increased tolerance for repeated errors.