Bayesian Models for Keyhole Plan Recognition in an Adventure Game

• Authors: David W. Albrecht;Ingrid Zukerman;An E. Nicholson
• Venue: User Modeling and User-Adapted Interaction
• Year: 1998

We present an approach to keyhole plan recognition which uses a dynamicbelief (Bayesian) network to represent features of the domain that areneeded to identify users‘ plans and goals. The application domain is aMulti-User Dungeon adventure game with thousands of possible actions andlocations. We propose several network structures which represent therelations in the domain to varying extents, and compare their predictivepower for predicting a user‘s current goal, next action and next location.The conditional probability distributions for each network are learnedduring a training phase, which dynamically builds these probabilities fromobservations of user behaviour. This approach allows the use of incomplete,sparse and noisy data during both training and testing. We then apply simpleabstraction and learning techniques in order to speed up the performance ofthe most promising dynamic belief networks without a significant change inthe accuracy of goal predictions. Our experimental results in theapplication domain show a high degree of predictive accuracy. This indicatesthat dynamic belief networks in general show promise for predicting avariety of behaviours in domains which have similar features to those of ourdomain, while reduced models, obtained by means of learning and abstraction,show promise for efficient goal prediction in such domains.