In this section you can find an overview of all the papers I read for my research concerning the use of cognitive modelling tools to generate rule based behaviour in computer games.

Creating Human-like Synthetic Characters with Multiple skill levels: A case study using the Soar Quakebot

Laird, J. E., Arbor, A., & Duchi, J. C. (2001). Creating Human-like Synthetic Characters with Multiple Skill Levels : A Case Study using the Soar Quakebot. Methodology, 54-58.

Using the already existing SOAR Quakebot (more on that later), a preliminary attempt was made to develop and evaluate an AI controlled character with multiple skill levels and human-like behavior. It was only a very basic study, exploring the most basic of parameters, but already there were strong indications that human like behavior from NPC's using, in this case, SOAR is possible. This is actually a very interesting paper for me, as it clearly describes some parameters (Decision time, Aggressiveness, Complexity of tactics, Levels of expertise) that I can implement and measure. Obviously I cannot reproduce their study, due to a lack of test subjects, but it does provide a good starting point. On top of this, it describes a good methodology to measure the effect of these parameters. The goal was to see if it was possible to recreate human-like behavior in an NPC and to provide a methodology to measure the humanness of said NPC. It is shown that both goals are achievable.

SOAR: An architecture for general intelligenve

Laird, JE, & Newell, A. (1987). Soar: An architecture for general intelligence. Artificial intelligence, 33(1987), 1-64. Retrieved from Link

In this paper the general architecture of SOAR is described. What systems are used, how the internals work, how and what of human behavior is simulated.

A Gentle introduction to SOAR: 2006 Update

Lehman, J. F., Laird, J., & Rosenbloom, P. (2006). A Gentle Introduction to SOAR: An Architecture For Human Cognition: 2006 Update. Science, 4(0413013), 1-37.

This paper describes more or less the same as the SOAR paper mentioned above, but then updated to the situation in 2006. In the meanwhile SOAR underwent some updates, most importantly for my work: it is possible to use SOAR from python.

Deconstructing ACT-R

Stewart, T. (2006). Deconstructing ACT-R. Proceedings of the Seventh International. Retrieved from http://actr.psy.cmu.edu/papers/641/stewartPaper.pdf

A couple of years ago, the authors of this paper decided to rewrite the entire core of act-r in python. In doing this they had to breakdown ACT-R piece by piece to gain a decent understanding of the systems present, so that they could be rewritten. This paper is the result of that work. Because of this it gives an excellent insight into how act-r works and it is a nice introduction into the subject.

Programmed Graph Rewriting with Time for Simulation-Based Design

Syriani, E., & Vangheluwe, H. (2008). Programmed Graph Rewriting with Time for Simulation-Based Design. Proceedings of the 1st international conference on Theory and Practice of Model Transformation, 91-106. In this Paper pac-man is implemented using modelling tools. This was done to efficiently test the playability of a given game, in this case pac man. I should be able to reproduce these results, but using Act-R and SOAR to generate the behavior of the NPC's present.