Machiavellian agents: Player modelling to deceive and be deceived

Abstract

This paper explores several augmentations to the previously described Plantagenet model of computer game agents to give agents the deceptic capability to deceive the player and/or be deceived by the player. Augmented finite state machine controllers for agents in a simple role playing game are generated using an evolutionary algorithm. It is demonstrated that the proposed model is a practical option for generating populations of around 30 agents in which constraints on agent behaviour to ensure that actions are consistent with deception can be satisfied in substantial portions of the agent population.