Characterizing and Modeling Agents in Digital Games

A promising approach in digital games is the possibility of customizing the game according to different demands. Artificial Intelligence algorithms play an important role in this direction, allowing the implementation of different behaviors for game agents. To accomplish this, it is necessary to model these agents in such way their behavior can be easily tunned to address different game features. In this paper, we discuss a generic representation to model virtual agents in digital games. Agents are modeled using a linear combination of different variables, which are used to represent specific game features. We perform experiments with FPS and Strategy games (COUNTER STRIKE and CIVILIZATION IV, respectively) and results show the effectiveness of this approach in characterizing and modeling agents. We were able to infer agents models by observing matches and also to generate different behaviors varying agent’s models.