Who are we?
We are a group of AI researchers with interdisciplinary interests who have
found narrative to be an important component of our work!
Michael Mateas is currently
building Terminal Time, a system that constructs ideologically-biased documentary
histories in response to audience feedback. Terminal Time pursues
a set of rhetorical goals in constructing its ideologically biased history.
As these goals are pursued, the system dynamically selects historical events
and constructs relations between these events so as to tell its story.
This project, a collaboration with documentary filmmaker Steffi Domike
and interactive artist Paul Vanouse, makes use of understandings of narrative
in documentary films and the arts.
Phoebe Sengers has studied how
to build behavior-based believable agents that actively structure their
behavior so as to be narratively understandable by human observers. Borrowing
ideas from Jerome Bruner's theory of narrative psychology, Sengers' system,
the Expressivator, actively keeps track of what an agent has communicated
to the human participant. This knowledge is used by the agents to decide
on transitions that maintain coherent narrative flow when switching between
Kerstin Dautenhahn has been working
on socially intelligent agents with a particular focus on autonomous robotic
agents and interaction dynamics of embodied agents. Her current interest
lies in autobiographic agents, namely robotic or virtual agents which develop
a memory in terms of a history of their own sensori-motor experiences in
relationship to interactions with the (social) environment. She recently
started two projects using story-telling approaches for human-agent interaction
involving robotic and virtual agents.
Clark Elliott has built a system that,
while using a fixed script, tells different stories by narrating the stories
with different emotional emphases. This is an alternative to more traditional
approaches to narrative generation which explore a combinatorial space
of primitive story elements. The emotional behavior of the narration agent
is generated by the Affective Reasoner, a cognitive appraisal model of
emotion. Elliott's work demonstrates that human understanding of narrative
makes use of an understanding of emotion.
James Lester has built believable agents
for educational environments that take into account pedagogical and believability
constraints when sequencing their behaviors. Similar in spirit to the work
of Sengers, these agents concern themselves with how they will be interpreted
by human observers when selecting behaviors. Lester has collaborated extensively
with designers, artists, and animators in building his agents, finding
that the representational skills of artists are important in building agents
whose behavior can be interpreted by humans. Recently, Lester has
also begun a project in interactive fiction.
Chrystopher Nehaniv has been
developing algebraic tools for a mathematically rigorous framework expressing
histories and, more generally, subjective views on the temporal experiences
of (possibly non-human) agents. The framework provides formal support to
realizing in robotics and software Dautenhahn's notion of a dynamic autobiographic
agent. By opening up the possibility to use their own and each other's
stories, this work provides narrative grounding to help release such agents
from mere reactivity.
Michael Mateas, co-chair
Department of Computer Science
Carnegie Mellon University
5000 Forbes Ave.
Pittsburgh, PA 15213
Center for Art and Media Technology (ZKM)
Institute for Visual Media
Department of Cybernetics
University of Reading
Whiteknights, PO Box 225
Institute for Applied Artificial Intelligence
School of Computer Science, Telecommunications, and Information Systems
243 South Wabash Ave.
Chicago, IL 60604
Department of Computer Science
North Carolina State University
Raleigh NC 27695-7534
Interactive Systems Engineering
Faculty of Engineering & Information Sciences
University of Hertfordshire
Hatfield, Hertfordshire AL10 9AB