Emotional Agents for Shooter Games: Understanding How Players' Emotional Profiles Influence Game Playouts

Abstract

Gameplay testing still suffers from inefficient feedback assimilation, not only due to the subjective nature of the retrieved information but also due to the amount of time required to retrieve it. This work attempts to ameliorate that issue by automating the testing process without losing all emotional data. We aim to achieve this by modeling an agent to replicate expected human behaviors. The models are created based on previously collected data from actual gameplay sessions, translated into Arousal and Valence values, through fuzzy clustering. With this proof of concept, we expect to emulate human behaviors in a satisfactory manner and evaluate the usefulness of this method as a quality assurance tool.