Thespian

IRIS Wiki - IS Systems - Thespian

Availability

Not available now. Check the pages of Mei Si and Stacy Marsela.

Technical Description

Thespian is a character-based approach to Interactive Drama. The choice of the actions in the narrative is based on character's internal state. Character modelling is based on PsychSims developed previously by Stacy Marsella and David Pynadath. Characters have goals, beliefs, and available actions to perform in order to reach their goal. To choose an action, a lookahead mechanism enables to assess the achievement of goals after several actions and other characters' possible reactions. The belief model not only contains beliefs of the storyworld's state but also beliefs about other characters' beliefs, in a recursive manner.

Such a character-based approach is complex and finding the right parameters to produce an interesting story is a tedious task. Therefore, a major feature of Thespian is a fitting algorithm, that allows to automatically calculate the character's parameters according to some desired path, that is story examples that are given by an author.

Result Description (end user perspective)

In the specific interactive drama called "TactLang", to teach soldiers language and culture awareness, Thespian is integrated within a 3D environment. User moves within this environment, and play the role of a character, in a subjective view. To interact with characters, the user speaks in natural language, which is automatically translated into dialog acts by the system.

Authoring Description

Authoring consists in choosing the goals of the characters, as well as actions that can be achieved towards the goals.

Several weighting parameters intervene to obtain suitable agent's behaviours and personnality. To tune these parameters, authors can benefit from a fitting algorithm, as mentioned above. This process simplifies authoring because the authors enter possible story paths rather than abstract parameters for agents. After entering one or more possible story paths, the system calculates the parameters and the author can experiment with the system and possibly add new story paths if necessary.

Another automatic mechanics for authoring is the simulation of potential users. After fitting the parameters, the system generates new stories by simulating user actions. The generated stories can then be shown to the author, who gives feedback that feeds the fitting algorithm (cyclic design).

Strong Points

An advanced character model and the authoring process.

Limitations

As any character-centered approach, it is unsure how global narrative qualities can be achieved by solely tuning agents' parameters.

Main Publications

• Mei Si, Stacy Marsella, and David V. Pynadath (2007). Proactive Authoring for Interactive Drama: An Author's Assistant, in 7th International Conference on Intelligent Virtual Agents, (Paris, France).

• Mei Si, Stacy C. Marsella, and David V. Pynadath (2005). Thespian: Using Multi-Agent Fitting to Craft Interactive Drama, in Proceedings of the International Conference on Autonomous Agents and Multiagent Systems, (Utrecht, Netherlands), pp. 21-28.

• David V. Pynadath and Stacy C. Marsella (2005). PsychSim: Modeling Theory of Mind with Decision-Theoretic Agents, in Proceedings of the International Joint Conference on Artificial Intelligence, pp. 1181-1186.

Supporting Narrative Theories

No explicit reference to a narrative approach is provided.

Computational Model

The agent model is based on Partially Observable Markov Devision Problems, a variant of reinforcement learning.

Type of interaction

3D interaction + speech recognition.