Page 77 -

46 Chapter 2 Understanding and conceptualizing intera k ion

$ww.01asbergen.com 1
8 1 Randy Glasberw.

"If you'd like to press 1, press 3.
If you'd like to press 3, press 8.
If you'd like to press 8, press S..."

A recent development based on the conversing conceptual model is animated
agents. Various kinds of characters, ranging from "real" people appearing at the
interface (e.g., videoed personal assistants and guides) to cartoon characters (e.g.,
virtual and imaginary creatures), have been designed to act as the partners in the
conversation with the system. In so doing, the dialog partner has become highly
visible and tangible, appearing to both act and talk like a human being (or crea-
ture). The user is able to see, hear, and even touch the partner (when it is a physi-
cal toy) they are talking with, whereas with other systems based on a dialog
partner (e.g., help systems) they can only hear or read what the system is saying.
Many agents have also been designed to exhibit desirable human-like qualities
(e.g., humorous, happy, enthusiastic, pleasant, gentle) that are conveyed through
facial expressions and lifelike physical movements (head and lip movements,
body movements). Others have been designed more in line with Disney-like car-
toon characters, exhibiting exaggerated behaviors (funny voices, larger-than-life
facial expressions).
Animated agents that exhibit human-like or creature-like physical behavior as
well as "talk" can be more believable. The underlying conceptual model is con-
veyed much more explicitly through having the system act and talk via a visible
agent. An advantage is that it can make it easier for people to work out that the in-
terface agent (or physical toy) they are conversing with is not a human being, but a
synthetic character that has been given certain human qualities. In contrast, when
the dialog partner is hidden from view, it is more difficult to discern what is behind
it and just how intelligent it is. The lack of visible cues can lead users into thinking
it is more intelligent than it actually is. If the dialog partner then fails to understand
their questions or comments, users are likely to lose patience with it. Moreover,