Artificial Mind for Virtual Characters - Advancing Artificial Intelligence Through Biological Process Applications - page 190

tion and Knowledge are integrated, goal-oriented
cognitive processes. Indeed, we don't simply look
at the environment, but we search for something
we know that can be useful to what we want to
do . Also, we try to do (always checking if it is
really being done) what we know that can satisfy
something we need , which is determined by our
motivations . A motivation thus has an associated
knowledge that has to answer questions like how ,
where and what is needed in order to satisfy it.
This knowledge allows the interpretation of the
current environment so as to guide the search
for information and the choice of an adequate
sequence of actions. So, we can think of a quali-
tative component (Minsky's agent or Franklin's
module) that encapsulates Perception, Knowledge
and Action processing related to its task.
Motivation can then be viewed as a propa-
gating energy that assigns different strengths to
character cognitive component. If it is assumed
that motivation is a consequence of emotional
states, this is coherent with a previous discus-
sion (Vilela and Lima, 2001) of the activating
role of emotion in cognitive processing. So, we
can think of motivational components that define
motivation, and are responsible for the propaga-
tion of the corresponding energy, as quantitative
component.
As already pointed out, an artificial mind has to
show unpredictable, flexible and dynamic choice
of actions, compatible with human being behav-
ior. The following characteristics summarize the
above discussion about which characteristics are
relevant to virtual character mind architecture:

• Qualitative components have a quantitative
element representing their activity level that
codes motivation energy.
• Motivations have competing quantitative
forces that establish priorities and drive
behavior according to activity propagation
among qualitative elements.
• Each motivation triggers only specific
cognitive components by transferring their
energy.
• Only those cognitive components with high
activity level perform their tasks, and this
functions as attentional and intentional
knowledge filtering.
• Activity propagation among qualitative
elements has to allow negotiations and
competition between motivations according
to environmental information.
• Satisfaction and frustration of a single
motivation are specified as zones of the cor-
respondent internal variable values, defined
by threshold parameters.

ARTIFICIAL MIND AS MULTI-AGENT
ARCHITECTURE

Although Minsky's term agent did not origi-
nally mean the same as it does in the context of
Distributed Artificial Intelligence, they are not
incompatible in principle. Usually, characters
are developed as single agent structures, but it
is possible to model them as a composition of
closely interacting sub-agents. Vanputte, Osborn
and Hiles (2002) developed the composite agent
architecture , which combines related reactive
and cognitive agents in a single structure. The
purpose of this combination is to take advantage
of both kinds of agents, improving the effective-
ness of their correspondent functionalities at the
higher-level agent. In the mentioned model, the
composite agent has two agent sets and an inter-
nal environment. One set groups the Symbolic
Constructor Agents ( SCAs ) and process sensory

• Behavior emerges from interaction be-
tween qualitative components (modules or
agents).
• Qualitative components representing cogni-
tive functions encapsulate Perception, Ac-
tion and Knowledge processes for a specific
task.
• Qualitative components representing Moti-
vations propagate activation energy.