We study methods for constructing a story from partial evidence where a story is defined as a path along a finite directed graph from the origin to a terminal node. Each node in the graph represents a possible event. A story builder receives evidence, i.e. a subset of events consistent with at least one story, and expands it into a coherent story. The analysis focuses on a stickiness property whereby if the story builder believes in a particular story, given a certain set of facts, then he believes in it given a broader set of facts consistent with the story.
- Non-Bayesian beliefs
- Story builder