A Computational Model of Suspense in Virtual Worlds

Abstract

In the cognitive psychology literature, it has been found that during suspenseful parts of a narration, a person uses a mental model similar to that of problem solving. The person is feeling suspenseful because of the uncertainty that the protagonist of the narration can find a solution to the problem (typically a threat of misfortune). This narrative comprehension employed during the problem solving is very similar to the way in which an AI partial order least commitment hierarchical planner attempts to find a solution plan. So, we have used such a planner, the Longbow planner, to represent the mental model formed by the audience, and came up with a computational model of suspense that can predict the suspense level at any point of a narration (generated by the Longbow planner for that story world).