RC5 Of Possible Worlds and Multiple Agents: Using Simulation to Model Narratives

Description

Computational Storytelling is an AI discipline that uses computational means to understand and
model the generation of narratives [1]. Such generative models need to be able to represent the
artifacts to be generated – narratives. Narratives, however, are structurally complex and dynamic
artifacts, which themselves represent many different phenomena [2]. A comprehensive analysis of
the structure, the dynamics and the descriptive range of narratives can be found in Possible Worlds
Narratology (PWN), a post-structuralist narratology operating on the level of plot [3, 4]. Especially,
as opposed to classical narratologies, PWN sees characters not just as semiotic constructs but as
non-actual individuals “[...] whose actions, experience, and destiny form the central concern of
narrative fiction” [4]. Plot, according to PWN, emerges when characters try to adapt the ‘textual
actual world’ to correspond to their wishes, obligations and beliefs, which are represented as
‘textual possible worlds’. Such a proposition-based conceptualization of narratives as the interaction
of character worlds with a textual actual world, and among each other, can account for many
narrative phenomena like genre, conflict and tellability. At the same time it is very reminiscent of
the AI notion of agent-based simulation.
In this course I will show how the concepts of PWN can be modeled with the help of
computational Multi-Agent Systems (MAS) based on the established Belief-Desire-Intention model
[5]. I will introduce Jason [6] as a suitable Java-based MAS framework, demonstrate a work-in-
progress implementation of a computational story model and discuss the system’s limitations.
In the end I will once again draw the connection to Computational Storytelling and outline how a
simulation-based model of PWN can be used to automatically generate and evaluate new stories.
This will highlight the benefits of the proposed system in comparison with current simulation-based
storytelling systems, which are not grounded in narratology theory. Thus, the proposed course can
be seen as a hands-on elaboration of one theme from SC2, and I intend to adopt it onsite for a better
fit with Pablo Gervás’ deliberations.

Objectives

tba

Literature


Gervás, P. 2009. Computational Approaches to Storytelling and Creativity. AI Magazine, 30(3): 49–62.
Currie, G. 2010. Narratives and narrators: A philosophy of stories. Oxford University Press.
Ryan, M.-L. 1991. Possible worlds, artificial intelligence, and narrative theory. Indiana University Press.
Ryan, M.-L. 2012. Possible worlds. In the living handbook of narratology. Hamburg: Hamburg University.
Rao, A. S., and Georgeff, M. P. 1995. BDI agents: From theory to practice. In ICMAS, 1: 312–319.
http://jason.sourceforge.net/

Course location

Forum 2

Course requirements

none

Instructor information.

Instructor's name

Leonid Berov

Email

tba

Vita

Leonid Berov is an early stage PhD researcher who likes to tell stories about stories. In collaboration with the Institute for English and American Studies and the Institute for Cognitive Science in Osnabrück he pursues research on computational models of storytelling, informed by narratology and the phenomenology of composition. His further research interests include creativity in general and computational models of personality.
He is holder of a MSc in Cognitive Science and a BSc in IT-Systems Engineering. When he is not engineering or cognizing, Leonid likes to frolic around with different languages.

Website

tba