Games are one of the the most popular applications of Artificial Intelligence (AI) techniques. From trivial games like Pac-Man to modern complex MMORPG (Massively multiplayer online role-playing game), e.g., World of Warcraft, we can see AI techniques have been widely applied in computer games or video games in the past 30 years. A non-player character (NPC) is the major target a player interacts with in a game, and we usually call an NPC a “game bot” or simply a “bot” in an FPS (First person shooter) game. The intelligence of a bot often decides the durability, difficulty, and joyfulness of an FPS game. A poor bot design makes the players feel bored, and dramatically decreases the attraction of a game, so appropriate intelligence is definitely a must for the game bots of FPS games.
Game AI has some special characteristics to differ it from Academe AI. Since the major purpose of a computer game is to entertain people, a bot must be fast, effective, robust, and efficient, which are reckoned as the four requirements in game bot design, according to the best user gaming experiences. Therefore, no matter how good the AI techniques are, it would be unacceptable if they can’t provide the users with satisfactory experience of fun. This specific and strong user demand makes game AI design more and more important. The cutting edge technology, which can offer more sophisticated computer power, at the same time, makes game AI design more and more feasible.
Most game AI designers use scripting or rules to control bot behavior. Just like traditional AI applications, game AI mainly focuses on rational inference. But in the real world, human decision making is not only involving rationality, but also sense. In this thesis, we propose an architecture for the bot designers to include affective computing in a game bot by attributing emotions to the bot. The architecture is based upon a variety of theories, including the famous OCC model. It treats Joy, Fear, Anger, Disgust, and Distress as five basic emotions. It uses three emotional variables, namely, Desirability of an event, Blameworthiness or Praiseworthiness of an action, and Appealingness of an object, to evaluate how a specific emotion is affected by the environment. In addition to an ordinary Rational Action Engine (RAE), which controls the rational behavior of a bot, we add an Emotional Action Engine (EAE) in the architecture to make the emotional behavior possible. An Action Arbitrator is equipped to decide, given a specific time and space, whether a bot should follow the rational or emotional inference. The arbitrator simulates how humans solve the conflicts between reason and emotion.
Our testing bed is Quake II enhanced by FEAR (Flexible Embodied Animat ‘Rchitecture), on which we have followed the architecture to create a rational bot, whose behavior is solely controlled by the RAE. The RAE is implemented as a rational FSM (Finite State Machine) containing four states: Exploring, Offense, Defense, and Gathering. It decides by logical inference how proper actions and knowledge are used for a bot to win the game. We have also created a second bot that includes both RAE and EAE. The EAE is implemented as an emotional FSM containing five states: Joviality, Fright, Fury, Loathing, and Chaos. Each emotional state represents a basic emotion, and decides what actions should be taken when a specific emotion is strong enough. Our evaluation shows, compared with the behavior of a rational bot, an emotional bot can exert more realistic feedback from the player’s point of view and exhibit richer behavior in a battlefield. This means an emotional bot can bring more fun to the human players in the simulation and improve the game play experience. The most important thing is that these newly add-on emotion-relevant functions won’t break the four requirements in the game bot design.

Games are one of the the most popular applications of Artificial Intelligence (AI) techniques. From trivial games like Pac-Man to modern complex MMORPG (Massively multiplayer online role-playing game), e.g., World of Warcraft, we can see AI techniques have been widely applied in computer games or video games in the past 30 years. A non-player character (NPC) is the major target a player interacts with in a game, and we usually call an NPC a “game bot” or simply a “bot” in an FPS (First person shooter) game. The intelligence of a bot often decides the durability, difficulty, and joyfulness of an FPS game. A poor bot design makes the players feel bored, and dramatically decreases the attraction of a game, so appropriate intelligence is definitely a must for the game bots of FPS games.
Game AI has some special characteristics to differ it from Academe AI. Since the major purpose of a computer game is to entertain people, a bot must be fast, effective, robust, and efficient, which are reckoned as the four requirements in game bot design, according to the best user gaming experiences. Therefore, no matter how good the AI techniques are, it would be unacceptable if they can’t provide the users with satisfactory experience of fun. This specific and strong user demand makes game AI design more and more important. The cutting edge technology, which can offer more sophisticated computer power, at the same time, makes game AI design more and more feasible.
Most game AI designers use scripting or rules to control bot behavior. Just like traditional AI applications, game AI mainly focuses on rational inference. But in the real world, human decision making is not only involving rationality, but also sense. In this thesis, we propose an architecture for the bot designers to include affective computing in a game bot by attributing emotions to the bot. The architecture is based upon a variety of theories, including the famous OCC model. It treats Joy, Fear, Anger, Disgust, and Distress as five basic emotions. It uses three emotional variables, namely, Desirability of an event, Blameworthiness or Praiseworthiness of an action, and Appealingness of an object, to evaluate how a specific emotion is affected by the environment. In addition to an ordinary Rational Action Engine (RAE), which controls the rational behavior of a bot, we add an Emotional Action Engine (EAE) in the architecture to make the emotional behavior possible. An Action Arbitrator is equipped to decide, given a specific time and space, whether a bot should follow the rational or emotional inference. The arbitrator simulates how humans solve the conflicts between reason and emotion.
Our testing bed is Quake II enhanced by FEAR (Flexible Embodied Animat ‘Rchitecture), on which we have followed the architecture to create a rational bot, whose behavior is solely controlled by the RAE. The RAE is implemented as a rational FSM (Finite State Machine) containing four states: Exploring, Offense, Defense, and Gathering. It decides by logical inference how proper actions and knowledge are used for a bot to win the game. We have also created a second bot that includes both RAE and EAE. The EAE is implemented as an emotional FSM containing five states: Joviality, Fright, Fury, Loathing, and Chaos. Each emotional state represents a basic emotion, and decides what actions should be taken when a specific emotion is strong enough. Our evaluation shows, compared with the behavior of a rational bot, an emotional bot can exert more realistic feedback from the player’s point of view and exhibit richer behavior in a battlefield. This means an emotional bot can bring more fun to the human players in the simulation and improve the game play experience. The most important thing is that these newly add-on emotion-relevant functions won’t break the four requirements in the game bot design.

[1] Ekman, P., Emotions Revealed: Understanding Faces and Feelings, Phoenix Press, (2004).
[2] Smith, C. A. and Ellsworth, P.C., Attitudes and Social Cognition, in Journal of Personality and Social Psychology, American Psychologists Association, Washington, vol. 48, no. 4, pp. 813-838, (1985).
[3] Ortony, A.; Clore, G. L.; and Collins, A., The Cognitive Structure of Emotions, Cambridge University Press, (1988).
[4] Christoph Bartneck, Integrating the OCC Model of Emotions in Embodied Characters. Department of Industrial Design Eindhoven University of Eindhoven, (2002).
[5] George Katsionis, Maria Virvou, Adapting OCC theory for affect perception in educational software, Department of Informatics University of Piraeus. (2005).
[6] Baillie, P., The Synthesis of Emotions in Artificial Intelligences: An Affective Agent Architecture for Intuitive Reasoning in Artificial Intelligences, Ph.D. University of Southern Queensland, (2002).
[7] Gomi, T. Vardalas, J. Koh-Ichi, Ide Appl. AI Syst. Inc., Elements of artificial emotion, in Robot and Human Communication, RO-MAN'95 TOKYO, Proceedings., 4th IEEE International Workshop in Tokyo, Japan, pp. 265-268, (1995).
[8] Toda, M., The Urge Theory of Emotion and Cognition. English translation in preparation by Professor Toda, School of Computer and Cognitive Sciences, Chukyo University, Technical Report No. 94-1-01. (1998).
[9] Jaime Griesemer and Chris Butcher, Bungie/MGS, The Integration of AI and Level Design in Halo, Game Developer Conference 2002. (2002)
[10] Quake II official site, http://www.idsoftware.com/games/quake/quake2/
[11] FEAR official site in sourceforge.net, http://fear.sourceforge.net/
[12] Sonnemans, Joep and Nico Frijda, The Determinants Of Subjective Emotional Intensity. Cognition Emotion 9, 483-506. (1995).
[13] Elliott, C. and Siegle, G., Variables Influencing the Intensity of Simulated Affective States, In AAAI technical report for the Spring Symposium on Reasoning about Mental States. Stanford University, Palo Alto, CA, USA. (1993).
[14] Elliott, C., The Affective Reasoner: A Process Model of Emotions in a Multi-agent System, Ph.D. Dissertation, Technical Report No.32. Northwestern University, The Institute for the Learning Sciences. (1992).
[15] Left Brain : Right Brain by Dan Eden, http://www.viewzone.com/bicam.html
[16] Klaus R. Scherer, Angela Schorr, and Tom Johnstone. Appraisal Processes In Emotion: Theory, Methods, Research. Oxford University Press. (2001).
[17] The Root of Human Conflict: Emotion vs. Reason by Mark D. Hamill, http://members.cox.net/mhamill1/mywriting/emotion.htm
[18] id Software official site, http://www.idsoftware.com/
[19] John Laird's Artificial Intelligence and Computer Games Research, http://ai.eecs.umich.edu/people/laird/gamesresearch.html
[20] Soar project, University of Michigan, http://sitemaker.umich.edu/soar/home
[21] Project EmoBot, Eastern Washington University, Department of Computer Science. http://penguin.ewu.edu/~ainoue/EmoBot/index.htm
[22] Rabin, S, etc.. AI Game Programming Wisdom (Game Development Series), published by Charles River Media. (2002).
[23] AI Game Programming Wisdom series official site, http://www.aiwisdom.com/
[24] Champandard’s game AI web site, http://aigamedev.com/
[25] AI Game Development: Synthetic Creatures with Learning and Reactive Behavior by Alex J. Champandard, New Riders Publishing, (2004).
[26] The Sims 2 official site, http://thesims2.ea.com/