Plan-Based Intention Revision

Introduction Plan-based story generation has operationalized concepts from the Belief-Desire-Intention (BDI) theory of mind to create goal-driven character agents with explainable behavior. Two classes of agents have emerged from this representation. A ‘conscious rationalizer’ agent who can justify adopting an intention with an air-tight causally-linked sequence of actions and a ‘fairweather’ agent who drops their intention the moment a causal complication is introduced. While adopting and dropping intentions is essential for goal-oriented agents, restricting agents to these two classes does not capture the dynamic nature of intentions. Intention revision models explainable behavior changes, which is especially important for the complex interactions between character intentions in interactive narratives. We define an intention revision model with two parts. We first prescribe when agents should reconsider existing intentions with logic from BDI agent design. Second, we define how agents decide to revise intentions using persistent goals (P-GOAL) as characterized by Cohen and Levesque (1990). Using the QUEST (Graesser, Lang, and Roberts 1991) cognitive model of question answering, we describe an evaluation assessing the explainability of intention revision.