The success of modern computer games depends more and more on interesting stories and less on dazzling graphics. To keep the player’s interest, game developers need to find a balance between the freedom of the user to explore all possible scenarios and the need to maintain a cohesive and consistent narrative. An important element of this narrative structure is time, which plays a crucial role in maintaining the required causality between the key elements of the story being told. The relative duration of scenes is also relevant to player interest, which can drop off when a scene goes on for too long.
Previous approaches frequently relied on pre- and post-condition constructs for actions, thus imposing an implicit temporal structure on the narrative of a game. The approach here relies on temporal aspects, such as duration, concurrency, and overlap, which are explicitly incorporated into the narrative action representation. The authors claim that, this way, the story and discourse can be generated via shared principles; the narrative construction is simplified; and the capability for the integration of action and motion at a technical level is achieved.
As a test case, they used a scene from Shakespeare’s The merchant of Venice. One version was generated by PDDL3.0 (the planning domain definition language), a system with implicit temporal information based on a standard action description for planning approaches. The other version was generated according to the authors’ approach; it used an explicit temporal representation. Their qualitative evaluation compared timing problems for the staging of agent actions, the ability to exploit information about the staging of agent actions, and the relative generative powers of the two systems. Their system can handle narratives that require actions with explicit durations, and it enables the representation of complex relationships between the story and the discourse. Furthermore, the approach offers a uniform and consistent treatment of time in agent-based storytelling.
Overall, the paper gives a good overview of the use of explicit temporal information in interactive storytelling. At times, further background information would be desirable (for example, on the PDDL language used in the example); within the page limits of a conference paper, however, this may not be practical.