Plan Recognition as an Aid in Virtual Worlds

While the development of virtual worlds provides players flexible, free spaces and attractive user-interfaces, the collaboration between computers and players has not become more efficient. Developments in human-computer collaboration may come from more efficient understanding between computers and humans. This paper presents a method to use plan recognition as an aid to improve the human-computer collaboration in a virtual world. An often-used data mining method, association rules mining, is employed to create a user behavior model, which represents a player’s basic characteristics. Based on this model and the current observations from players, a plan recognition algorithm tries to predict players ’ action sequence. By using these predictions, the virtual world can organize its local resource more efficiently and react to players ’ requests timelier. Moreover, some main problems in probability-based plan recognition are discussed