Modeling emotions with EEG-data in StateCraft
MetadataShow full item record
Emotions have been shown to play an important part in human decision making, and emotions in Artificial Intelligence have been shown to affect agent performance and believability. The aim of this thesis is to use EEG-data to model players' emotions. The emotion model was incorporated into the existing Emotion module in the computer game known as StateCraft. Using artificial neural networks as a tool, two different models of the players' emotions were created, a general model and a country specific model, resulting in four different configurations of the Emotion module. Simulations of these four different configurations of the Emotion module were conducted. Statistical analysis of the simulation data shows that the agents perform worse overall with emotions than without. The country specific model appears to perform better than the general model in the simulations. Analysis also indicates that the four new EEG-based configurations perform worse overall than the existing Emotion module which is based on game states . The EEG-based emotions promote more risky behavior, and for some countries that can have a very negative effect on performance.