Social puppets: Towards modular social animation for agents and avatars

State-of-the-art computer graphics can give autonomous agents a compelling appearance as animated virtual characters. Typically the agents are directly responsible for controlling their graphical representation, but this places too much burden on the agents that already deal with difficult high-level tasks such as dialog planning. This paper presents work, done in the context of an interactive language and culture training system, on a new kind of engine that fits between the high level cognitive agent models and the animated graphics that represent them. This is a social engine that generates socially appropriate nonverbal behavior based on rules reflecting social norms. Similar to modular physics engines, the social engine introduces a re-usable component that can heighten believability of animated agents in games and simulations with relatively little effort.