Virtual Babyz - Narrative Intelligence - page 224

so much to making good stories so powerful are not very pliable. To add sig-
nificant interactivity to narrative, we believe it is helpful (at first) to move away
from traditional plot structures and towards a looser character-centric experi-
ence. (By “significant interactivity” we mean interactivity that causes continu-
ous, meaningful and varied changes to the events of a story, not just a glorified
“next” button.)
Aylett (Aylett 1999) points out that as a virtual world becomes closer to
real life, it becomes more likely that narrative could emerge from the virtual
world as it does from human life experience. For example, just as we tend to
tell stories about the events that happen to ourselves on a given day, we could
recognize stories in the events that occur in a virtual world. However to achieve
this life-like effect, the characters and environments in the virtual world must
be endowed with a sufficiently rich set of behaviors. Creating this richness is a
fundamental challenge for the authors of virtual characters and environments.
To this end we put all of our effort into creating as many behaviors and
behavior-alternates as we could for Petz and Babyz. As a rule we found that
creating 6 or 7 alternates for each behavior seemed to reach a critical threshold
for the illusion of life, perhaps surpassing the point where users can keep track
of behavior repetition. In total, the Petz production team of 4 engineers and
4 animators worked for three years to author approximately 100 short-term
narrative goals, constructed from 2000 low-level animation pieces. In one year
of work for this first version of Babyz we authored approximately 50 short-term
narrative goals, constructed from 1000 low-level animation pieces.
Based on customer feedback in the form of emails and bulletin board post-
ings, and on our own observations when playing with the characters, we feel
that the Babyz and Petz characters exhibit a reasonably convincing illusion of
life, and allow for occasional short, simple emergent narratives to occur. But
two shortcomings of the emergent narrative approach stand out, as Aylett ob-
served. First there is a risk that narrative may not emerge at all, and second,
even if it does emerge, it may be boring. In a virtual environment where the
control of agents is decentralized and uncoordinated, just as in real life (osten-
sibly), there is no guarantee that a meaningful and coherent chain of events
(i.e., narrative) will occur at all. And if it does occur it may only be a frag-
ment of what we consider a “good story”. In our own evaluation of Petz we
certainly found many moments when the experience seemed fragmented and
even boring. We attempted to address this shortcoming in Babyz with the addi-
tion of a few explicit long-term narrative behaviors as described earlier, which
are deliberately spawned when it is detected that no emergent narratives have
occurred recently. (However, Petz probably has an overall richer potential for