February 26, 2008
As the previous chapter described, James Meehan’s Tale-Spin — built on a simulation embodying the “scruffy” artificial intelligence theories of Roger Schank and Robert Abelson — generated coherent accounts of character actions and interactions in a fictional world. This set the foundation for the field of story generation. Considered today, it also raises an inevitable question: What next?
This chapter considers two different responses. From the perspective of today’s media authors, the question of how to generate appropriate character behavior is an important one. In my discussion of Prince of Persia: The Sands of Time I described the basic design of a common, problematic approach to character behavior: the finite-state machine (FSM). As it turns out, an approach somewhat similar to Tale-Spin’s has been used to address some of the limitations of authoring behavior with FSMs — specifically, in the computer game F.E.A.R. — offering one response to “what next” for characters.
But within the history of story generation a rather different answer emerges. As I mentioned in the previous chapter, a primary critique of Tale-Spin within the AI community is that stories are not simply an account of character behavior. To answer “what next” for story generation requires a system attuned to the overall shape of a fiction, rather than just local rules for character action. Scott Turner’s work on Minstrel is an ambitious example of this next stage for story generation.
Through different routes, these two examples also bring this book to central topics for AI, media authoring, and fiction. First, a familiar question: Should AI attempt to build systems that operate the way people do? Second, a new topic: The rise of statistical techniques and their implications for approaches built with hand-authored processes and data.