Moe is based on the principle of plot points: a list of important moments (plot points) in the story are identified. They are organized according to a partial order: some plot points must occur only when another plot points has already occured.
Such a partial ordering does produce a large quantity of possible sequences. In order to filter out uniniteresting moves, an evaluation function is proposed, which rates a scenario according to some criteria. Seven criteria are used:
- Thought Flow: logical relation between one event in the User's experience and the next one
- Activity Flow: how bored the User feels
- Options: perceived freedom
- Motivation: User's actions motivated by her goals
- Momentum: Proximity of certain events that should preferentially happen together
- Intensity: Increases the User's excitement
- Manipulation: how manipulated the User feels
Based on this rating, a search algorithm is applied, to select the best move at a given moment in the story.
Result Description (end user perspective)
Initially, Moe was not to be fully implemented: only the evaluation function and the search algorithms were implemented. The target application is a text-based environment, where the user types free text which is interpreted into commands.
Recently, Moe's approach has been implemented in a research 2D adventure game (Sullivan et al. 2008).
In Moe, the aesthetics of a given author are provided by the evaluation function. Since Tea for Three, the story used to test the approach, was written by Peter Weyhrauch himself, the designer of the whole system, it is difficult to assess the accessibility of the approach for less technical authors.
The evaluation function is a major contribution of Moe for two reasons:
- It embeds features that include the interactive nature of experience,
- It is an access point to authors.
Plot points in Moe are fragments that are pre-written by the author. It limits the degree of agency of the system.
- Weyhrauch, P. (1997). Guiding Interactive Drama. Ph.D. Dissertation, Tech report CMU-CS-97-109, Carnegie Mellon University.
- Sullivan, A., Chen, S., Mateas, M. (2008). Integrating Drama Management into an Adventure Game. Proceedings of the Fourth Artificial Intelligence and Interactive Digital Entertainment Conference, October 22-24, 2008, Stanford, California, USA 2008.
Supporting Narrative Theories
Moe is based on planning algorithms. Several adversary search algorithms are tested to optimize the computational costs.