Towards emotional and socially realistic game companions with personality

This thesis presents a fully integrated and modular framework for social and emotional game companion that focuses on realistic social interaction between the player and Non-Player Characters (NPCs) in a game environment. Moreover, this thesis proposes integrated computational models for a formulation of action selection rules based on the game companion’s personality, the relation between player and game companion and the perceived (non-)verbal actions to enhance the NPC’s believability and the player’s interaction in a game context. The rules were derived from data collections of both human-human and human-machine interactions. This thesis argues that NPC with such capability will accommodate a new experience when playing games. To illustrates the effectiveness of the proposed framework and computational models, the author implemented the framework and computational models into two game scenarios: The Smile Game where a human player who abides in the real world played with an ECA who inhabits a virtual world. The second scenario was The Skyrim Game, where an avatar represented the human player, who played with a virtual NPC in a virtual world. With a total of 117 participants in 217 interactions with the system, the results show that the players evidently perceived the NPCs’ personality in accordance with the one set to them. The NPCs’ ability to display emotions appropriately also provides the feeling of immersion in games to the players. Moreover, the ability to forge relationships naturally with the game companions gives the sentimental feeling towards the game companions. The main contribution of the work presented to the field of Intelligence Virtual Agents specifically in the domain of computer games, in this thesis is A fully integrated and modular framework for social-affect-aware game design complete with the computational models for social interaction with game companions. Several games have been built in several scenarios with simple social interactions between players and the game companions using the framework and computational models proposed to explore the opportunities of the framework and models. Finally, this thesis also presents some scenarios for data collections to construct interaction rules for game companions