Jeff Orkin, Deb Roy

Abstract

We envision a future in which conversational virtual agents collaborate with humans in games and training simulations. A representation of common ground for everyday scenarios is essential for these agents if they are to be effective collaborators and communicators. Effective collaborators can infer a partner’s goals and predict future actions. Effective communicators can infer the meaning of utterances based on semantic context. This article introduces a computational model of common ground called a Plan Network, a statistical model that encodes context-sensitive expected patterns of behavior and language, with dependencies on social roles and object affordances. We describe a methodology for unsupervised learning of a Plan Network using a multiplayer video game, visualization of this network, and evaluation of the learned model with respect to human judgment of typical behavior. Specifically, we describe learning the Restaurant Plan Network from data collected from over 5,000 gameplay sessions of a minimal investment multiplayer online (MIMO) role-playing game called The Restaurant Game. Our results demonstrate a kind of social common sense for virtual agents, and have implications for automatic authoring of content in the future.