Recently there has been growing interest in the interplay between traditional cognitive processing and emotional processing. In contrast with past beliefs that emotions do little more than interfere with cognition, it is now generally accepted that emotionality is beneficial and necessary for healthy living. Jáuregui [1] goes so far as to claim that the emotional system is a genetically implanted information system which supplies the agent with orders on what to do when, together with the urgency of required actions. He implies that it is emotionality which controls humans, not intellectual cognition.
Two fundamental questions about emotions are ``What is an emotion?'' [2] and more recently, ``What are the functions of emotion?'' In this study I concentrate on functionality and how it has evolved. While there has been recent work identifying the functions of emotions (eg. [3, pp. 67-70]), there hasn't been any satisfactory method to observe the evolution of emotion over time, and how its functionality develops.
This leaves us with philosophical theorising (eg. [4,5]) and efforts to laterally map observations of emotions in animals to observations of emotions in humans [6,3]. Neither of these are fully adequate for understanding the functions of our complex emotions, the former lacks the strength of a controlled experiment, while the latter leaves too many missing links. The use of artificial life and genetic adaptation techniques is proposed as a solution to both these problems; such simulation provides a bed for highly controlled experiments, while being fast enough to allow full longitudinal investigation of evolution.
Computer scientists have tended to ignore emotion, although recently they have concentrated on creating agents which act emotional and to some extent understand emotion. This has been seen as useful for two main reasons. On one hand, we would like artificial actors in interactive entertainment to be believable, which includes acting emotional. On the other, we would like computers to be emotionally intelligent, able to understand and interpret our emotions correctly. Unfortunately, there has little work on creating artificial agents that are emotional. Reilly [7], for example, freely sacrificed accuracy of emotion for a simpler representation. While simulation of social interactions has been successful (see [8,9,10]), it is not known whether anyone has yet utilised simulation as an avenue of emotional research and experimentation.
In section 2, background material relating to emotion and affective computing will be discussed, and the concepts and structures relating to finite state automata and genetic algorithms will be presented. Section 3 will describe the artificial life simulation environment, detailing the agents and their genetic structure, as well as the ways in which they interact with their world. Section 4 discusses the statistics gathered from artificial life simulations and section 5 describes the experimental model and design. Results will be presented and discussed in section 6 and the conclusions drawn from this investigation outlined in section 7.