«
Yu Chen
Software Development
Extracting AI Personality from AI Behaviors Using Particle Swarm Optimization and Artificial Psychosocial Framework
My thesis focused on extracting artificial intelligence ( AI ) personalities from predefined behaviors based on Particle Swarm Optimization ( PSO ) algorithm and Artificial Psychosocial Framework ( APF ). My goal was to build a pipeline that could identify valence values from input behaviors and then extract a personality output from the valence values . The two identification processes use PSO to find optimal results in the solution space . My artifact simulates a 2D turn-based roleplaying game ( RPG ) that uses APF as the base system to generate AI emotions and memories . The memory system inside APF is one of the two main systems used to help identify personality output , and I implemented another system called the personality system to interpret valence values from a personality .
The reason I chose this project was to further explore AI development and create a new concept and process that has not
been studied before . Applying personality to AIs is common in the game industry , but systematically generating personalities is a challenge . Also , extending the current APF and implementing a support system for the APF can help the people who are currently working on it do more interesting research in the APF and even change how AI is developed in video games .
I spent one year designing and implementing the system and artifact . I learned new concepts such as AI algorithms , APF , and AI design . This thesis can be applied to many video games that require complicated AIs such as RPG , first-person shooter ( FPS ), and real-time strategy ( RTS ) games . By finding a NPC ’ s personality , we can develop AIs that have unique behavior patterns and consistent behaviors . Also , introducing personality in AI development can avoid extra work of redesigning AI behaviors when adding or changing features .
72 SOFTWARE DEVELOPMENT