Page 191 -

160 Chapter 5 Understanding how interfaces affect users

Mike brings his hands up as if to speak, so Rea does not continue, waiting for
him to speak.
Mike: "Tell me more about it."
Rea: "Sure thing. It has a nice garden . . ."

Which of the various kinds of agents described above do you think are the most convincing?
Is it those that try to be as humanlike as possible or those that are designed to be simple, car-
toon-based animated characters?

Comment We argue that the agents that are the most successful are ironically those that are least 1
like humans. The reasons for this include that they appear less phony and are not trying
to pretend they are more intelligent or human than they really are. However, others 1
would argue that the more humanlike they are, the more believable they are and hence
the more convincing. I

5.6.2 General design concerns
Believability of virtual characters

One of the major concerns when designing agents and virtual characters is how to
make them believable. By believability is meant "the extent to which users inter-
acting with an agent come to believe that it has its own beliefs, desires and person-
ality" (Lester and Stone, 1997, p 17). In other words, a virtual character that a
person can believe in is taken as one that allows users to suspend their disbelief. A
key aspect is to match the personality and mood of the character to its actions. This
requires deciding what are appropriate behaviors (e.g., jumping, smiling, sitting,
raising arms) for different kinds of emotions and moods. How should the emotion
"very happy" be expressed? Through a character jumping up and down with a big
grin on its face? What about moderately happy-through a character jumping up
and down with a small grin on its face? How easy is it for the user to distinguish be-
tween these two and other emotions that are expressed by the agents? How many
emotions are optimal for an agent to express?

Appearance
The appearance of an agent is very important in making it believable. Parsimony and
simplicity are key. Research findings suggest that people tend to prefer simple car-
toon-based screen characters to detailed images that try to resemble the human form
as much as possible (Scaife and Rogers, 2001). Other research has also found that
simple cartoon-like figures are preferable to real people pretending to be artificial
agents. A project carried out by researchers at Apple Computer Inc. in the 80s found
that people reacted quite differently to different representations of the same inter-
face agent. The agent in question, called Phil, was created as part of a promotional