Virtual Storyteller

IRIS Wiki - IS Systems - Virtual Storyteller

Availability

The Virtual Storyteller is publicly available on the project homepage.

Technical Description

The Virtual Storyteller builds on the emergent narrative approach as used in Fearnot. Stories are not prespecified, but emerge as a consequence of the interaction of autonomous virtual agents playing the roles of story characters in a storyworld simulation. Novel is that these agents are not only self-interested, goal-directed characters, but also drama-directed, collaborative actors. This means a distributed drama management approach is being adopted. As characters, the agents adopt character goals and create plans in order to attain these goals. As actors, the agents select external events and retrospectively fill in details of the initial state of the simulation, in order to justify the adoption of new goals and facilitate the creation of goal plans (Swartjes, Kruizinga & Theune 2008).

Result Description (end user perspective)

The Virtual Storyteller generates simple stories that result from autonomous character interaction. Currently, there are several experiment domains, most notably, one about pirates and one based on the Little Red Riding Hood folktale. Creating these domains was done with a focus not on achieving specific aesthetic goals, but on better understanding the authoring process involved. Descriptions of this process and resulting stories can be found in (Swartjes & Theune 2009).

Authoring Description

For The Virtual Storyteller, an iterative authoring approach is adopted in which the task of the author is to create a space of possible stories. This space follows from the story content (goals, actions, events, etc.) and processes (goal adoption, action selection, etc.) created by the author. This space cannot be totally preconceived. Often, the authored content and processes give rise to event sequences that the author had not anticipated, but are nevertheless valuable. Hence, authorial intent is envisioned to take shape during authoring, and story generation feedback from the system is considered important (Swartjes & Theune 2009).

Strong Points

• Emergent narrative approach: narrative is not explicitly represented but emerges from autonomous characters.
• Agents are not just self-interested characters but also drama-interested actors. The implementation of this second role is informed by techniques used in improvisational theatre (Swartjes & Vromen 2007). They can select external events and retrospectively adapt the initial state of the story for dramatic reasons.

Limitations

• No emotion model or support for authoring character emotions.
• Model for global narrative control remains limited.
• Currently no interface for user interaction.

Main Publications

• Swartjes, I., & Theune, M. (2009). Iterative Authoring Using Story Generation Feedback: Debugging or Co-creation? In I. A. Iurgel, N. Zagalo, & P. Petta (Eds.), Interactive Storytelling (Vol. 5915, pp. 62–73). Berlin / Heidelberg: Springer.
• Swartjes, I., Kruizinga, E., & Theune, M. (2008). Let's Pretend I Had a Sword: Late Commitment in Emergent Narrative. In U. Spierling & N. Szilas (Eds.), Interactive Storytelling (Vol. 5334, pp. 264–267). Berlin / Heidelberg: Springer.
• Swartjes, I., & Theune, M. (2008). The Virtual Storyteller: Story Generation by Simulation. In A. Nijholt, M. Pantic, M. Poel, & H. Hondorp (Eds.), BNAIC 2008, Proceedings 20th Belgian-Netherlands Conference on Artificial Intelligence (pp. 257–264). Enschede: University of Twente.
• Swartjes, I., & Vromen, J. (2007). Emergent Story Generation: Lessons from Improvisational Theater. In Intelligent Narrative Technologies: Papers from the AAAI Fall Symposium (pp. 147–150). Menlo Park: AAAI Press.
• Swartjes, I., & Theune, M. (2006). A Fabula Model for Emergent Narrative. In S. Göbel, R. Malkewitz, & I. Iurgel (Eds.), Technologies for Interactive Digital Storytelling and Entertainment (Vol. 4326, pp. 49–60). Heidelberg: Springer.

Supporting Narrative Theories

The Virtual Storyteller draws inspiration from improvisational theatre (Swartjes & Vromen 2007). In the poetics of improv, character interaction yields the additive collaborative construction of a story, in contrast to the neo-Aristotelian poetics in which author-given plots determine the necessary character behaviour.

The fabula of the generated narratives is captured and respresented by the system, using a formal model based on Trabasso's causal network theory of story comprehension (Swartjes & Theune 2006).

Computational Model

The Virtual Storyteller is a Multi-Agent System in which agents are implemented using a BDI-based character model. The agents use an adapted version of partial-order planning.