A bottom-up approach to emulating emotions using neuromodulation in agents

Abstract

A bottom-up approach to emulating emotions is expounded in this thesis. This is intended to be useful in research where a phenomenon is to be emulated but the nature of it can not easily be defined. This approach not only advocates emulating the underlying mechanisms that are proposed to give rise to emotion in natural agents, but also advocates applying an open-mind as to what the phenomenon actually is. There is evidence to suggest that neuromodulation is inherently responsible for giving rise to emotions in natural agents and that emotions consequently modulate the behaviour of the agent. The functionality provided by neuromodulation, when applied to agents with self-organising biologically plausible neural networks, is isolated and studied. In research efforts such as this the definition should emerge from the evidence rather than postulate that the definition, derived from limited information, is correct and should be implemented. An implementation of a working definition only tells us that the definition can be implemented. It does not tell us whether that working definition is itself correct and matches the phenomenon in the real world. If this model of emotions was assumed to be true and implemented in an agent, there would be a danger of precluding implementations that could offer alternative theories as to the relevance of neuromodulation to emotions. By isolating and studying different mechanisms such as neuromodulation that are thought to give rise to emotions, theories can arise as to what emotions are and the functionality that they provide. The application of this approach concludes with a theory as to how some emotions can operate via the use of neuromodulators. The theory is explained using the concepts of dynamical systems, free-energy and entropy.