A New Approach to Social Behavior Simulation: The Mask Model

This paper proposes a new perspective, based on the concept of social masks, for the simulation of a realistic NPC (Non-Player Character) behavior. The Mask Model goal is to support AI techniques for autonomous agents by encouraging or discouraging behaviors according to the social environment and by providing knowledge about possible reactions to the agent actions. In this approach, the NPC tendencies are controlled by the interactions of three overlapping mask layers: selfperception layer, social layer and interpersonal layer. The masks mould the tendencies, the feelings and the ethics of a NPC. By changing the links between characters and masks, a wide variety of different behaviors and story-lines may arise. The paper present an algorithm for the selection of the actions and an example implementation.