Emotion modelling with human belief revision in computer games

Abstract

Emotion modelling is receiving more and more attention from various fields, e.g. cognitive science, psychology, computer science and neuroscience. Most of these fields share the common research consensus that emotion can be beneficial to human's mental activities. This thesis is also grounded on the same consensus and makes further validations based on the following two hypotheses: One is emotional agents in games should behave more like human beings than emotionless agents; the other is that agents having full emotional architecture should obtain better playing performance than agents with only partial architecture. Based on theoretical support, the author further hypothesizes that peoples' long term belief can be one of the sources to release complex emotions.

The experiment result suggests the emotional agents did perform significantly better than emotionless ones, but it was unable to significantly reflect the advantages from fully structured emotional agents over the ones of the partial architecture.