Bayesian networks: Real-time applicable decision mechanisms for intelligent agents in interactive drama

The growing use of intelligent agents in virtual realities calls for agents that can interact with users in a fluent and believable way. Intelligent agents are expected to react in an increasingly human like manner, showing distinct characteristics and to express emotion. Which calls for increasingly larger datasets for the agents to make informed and plausible decisions. Intelligent agents need to be able to process large knowledge bases and make decisions based on a wide range of relevant factors in real-time in order for the user to find them believable. This is very important for autonomous agents in emergent interactive drama because they need to respond fluently to user interactions in a manner that resembles that of a real actor. Bayesian networks (BNs) are particularly suitable to implement intelligent agents decision mechanism because they support transient emotions and decision making and accommodate for conflicts between the virtual agents goals. In their basic form, BN reasoning algorithms do not scale well, since the cost of updating values in a BN is NP-hard in the worst case. We propose efficient and scalable BN techniques using relevance reasoning, that are suitable for the needs of our autonomous agents in directed emergent drama and allow for real-time decision making.