VEsNA-Pro: Exploiting BDI Agents with Propensities for Emergent Narrative

In many real applications including entertainment, training, serious games, immersing users and players in a virtual reality environment and making their interaction with virtual humans engaging and believable is the key for the application success and effectiveness. Pro-AgentSpeak(L) extends the AgentSpeak(L) language for Belief-Desire-Intention (BDI) agents with the agents’ temperament, i.e. those personality traits and attitudes that constitute their propensities. Unlike beliefs, that model the agents’ epistemic state, propensities are related with their personality and emotional state. In order for these traits and emotional attitudes to drive deliberation, plans are annotated with weighted propensities and plan selection is guided by compatibility with the agent’s current propensities; execution, in turn, updates mutable propensities through post-effects. Two case studies designed with the Untold Games video game company show how propensities enable lightweight yet expressive agents’ behavioural diversity and properly address emergence of realistic and engaging narrative.