Next: Moods in the Tutor
Up: Affective Reasoner personality models
Previous: Models with two or
The inspiring of computer agents with personalities, and emotions,
can help to engage the student in several ways:
- There is some evidence to show that autonomous agents with interesting
personalities can help to engage the student (e.g.,
[Lester & Stone1997, Lester et al.
1997]). This is one of the premises upon
which systems such as Design-a-Plant (Herman the Bug)
[Stone & Lester1996, Lester & Stone1997], the Internet Advisor
(Cosmo), and the Soar Training Expert for Virtual Environments (Steve)
[Rickel & Johnson1997]have been designed.
- The AR model allows us to build interesting, rich personalities.
Coming to know these personalities, through using the tutoring system
over time, can be fun.
- The problems that most game characters have is that they are either
(a) predictable, or (b) confusing. Building agent personalities on
rich, theoretically grounded structures consistent with human personality
traits allows us to build complex virtual entities which are nonetheless
cohesive and understandable in their responses, because they follow
well-established, and easily understood, social conventions.
- Agents which have some understanding of a student's emotion states are
better listeners- they may well be perceived as much more able to get
it, than an agent that has no sophisticated knowledge of frustration,
joy, and the like. People often find good listeners to be engaging. That is,
although an AR agent has no understanding of what a student might be
upset, or happy, about, it nonetheless might have a useful knowledge of
how the student feels. For example, an agent might not know what a test
is, but feel pity regarding a student's unhappiness about failing it.
- Different students like different approaches to covering material. Some
like to collaborate, some like to compete; some prefer to stick to
performance issues, others like a more personal relationship. These
contrasting approaches require very different emotion responses for
identical situations, on the part of the cooperating automated
agent. Getting this right may help students to be engaged in the tasks. The AR
allows for different relationships between agents (and students) which
support these different approaches.
Next: Moods in the Tutor
Up: Affective Reasoner personality models
Previous: Models with two or
Clark Elliott
Wed Dec 17 18:41:50 EST 1997