Information Technology Reference
In-Depth Information
ing them narratively. This motivates a narrative agent architecture, the Expres-
sivator, which structures agent behavior to support narrative, thereby creating
agents that are intentionally comprehensible. The methodology in this chapter
integrates the narrative traditions of cultural studies with the technical tra-
ditions of Artificial Intelligence, thereby presenting itself, too, as a form of
Narrative Intelligence.
The problem
Building complex, integrated artificial agents is one of the dreams of AI. Clas-
sically, complex agents are constructed by identifying functional components
- natural language processing, vision, planning, etc. - designing and build-
ing each separately, then integrating them into an agent. More recently, some
practitioners have argued that the various components of an agent strongly
constrain one another, and that the complex functionalities classical AI could
come up with could not easily be coordinated into a whole system. They of-
fer other construction methodologies instead. In particular, behavior-based AI
proposes that the agent should be split up, not into disparate cognitive func-
tionalities, but into behaviors, each of which integrates all of the agent's func-
tions for a particular behavior in which the agent engages. Examples of such
behaviors include foraging, sleeping, and hunting.
Even such systems, however, have not been entirely successful in building
agents that integrate a wide range of behaviors. Rodney Brooks, for example,
has stated that one of the challenges of the field is to find a way to build an
agent that can integrate many behaviors, where he defines many to be more
than a dozen (Brooks 1990). Programmers can create robust, subtle, effective,
and expressive behaviors, but the agent's overall behavior tends to gradually
fall apart as more and more behaviors are combined. For small numbers of
behaviors, this disintegration can be managed by the programmer, but as more
and more behaviors are combined their interactions become so complex that
they become at least time-consuming and at worst impossible to manage.
In both cases, divide-and-conquer methodologies lead to integration prob-
lems. With classical agents, who are split up by functionality, there are often
problems with a functional underintegration. This underintegration manifests
itself in various kinds of inconsistency between the different functions, such
as not being able to use knowledge for one function that is available for an-
other. For example, the agent may speak a word it cannot understand or vis-
ibly register aspects of the world that do not affect its subsequent behavior.
