Page 192 -

5.6 Virtual characters: agents 1 61

Figure 5.1 4 Two versions of
Phil, the agent assistant that
appeared in Apple's promo-
tional video called the
Knowledge Navigator (a) as
a real actor pretending to be
a computer agent and (b) as
a cartoon being an agent.
Phil was created by Doris
Mitsch and the actor Phil
was Scott Freeman.

video called "The Knowledge Navigator." He was designed to respond and behave
just like a well-trained human assistant. In one version, he was played by a real actor
that appeared on a university professor's computer screen. Thus, he was portrayed as
an artificial agent but was played by a real human. The actor was a smartly dressed
assistant wearing a white shirt and bow tie. He was also extremely polite. He per-
formed a number of simple tasks at the computer interface, such as reminding the
professor of his appointments for that day and alerting him to phone calls waiting.
Many people found this version of Phil unrealistic. After viewing the promotional
video, people complained about him, saying that he seemed too stupid. In another
version, Phil was designed as a simple line-drawn cartoon with limited animation (see
Figure 5.14) and was found to be much more likeable (see Laurel, 1993).

Behavior

Another important consideration in making virtual characters believable is how
convincing their behavior is when performing actions. In particular, how good are
they at pointing out relevant objects on the screen to the user, so that the user
knows what they are referring to? One way of achieving this is for the virtual char-
acter to "lead" with its eyes. For example, Silas the dog turns to look at an object or
a person before he actually walks over to it (e.g., to pick the object up or to invite
the person to play). A character that does not lead with its eyes looks very mechan-
ical and as such not very life-like (Maes, 1995).
As mentioned previously, an agent's actions need also to match their underly-
ing emotional state. If the agent is meant to be angry, then its body posture, move-
ments, and facial expression all need to be integrated to show this. How this can be
achieved effectively can be learned from animators, who have a long tradition in
this field. For example, one of their techniques is to greatly exaggerate expressions