Currently, the user model simply consists of a single stack containing representations of intentions. The intentions are those inferred by the system the user wants to achieve. The user model and the system model hold representations that are inferred dynamically during dialogue processing. They are used to represent current mental states of the system and the user.