User Intention Modelling and Intention Recognition in Games using World State Indicators

Abstract

The work presented in this thesis focuses on two main areas. First we develop a model of user intentions in games. That model defines various concepts related to player behaviour in (computer) games and interactive environments, such as actions, goals, plans and intentions. Additionally our model shows the relationship between those concepts and explains how they affect each other. The purpose of the model presented here is twofold. One the one hand it provides common definitions for research in the area of user behaviour modelling in games. On the other hand this model also forms the underlying basis for the remainder of our work presented here.

The second main area of focus is intention recognition in games. We propose a novel approach which is solely based on monitoring the changes in the game world state, instead of observing player actions. We evaluate current approaches to plan and intention recognition, their strengths and weaknesses. We further compare existing research on intention recognition to our approach and evaluate the performance of our prototype system iRecognise in the context of a case study using the board game RISK. A range of experiments that were carried out demonstrates that our proposed approach to intention recognition is valid and therefore verifies its intention recognition capabilities in the context of games.