GPTs in Mafia-like Game Simulation

Abstract

In this research, we explore the potential of Generative AI models, focusing on their application in role-playing simulations through Spyfall, a renowned mafia-style game. By leveraging GPT-4's advanced capabilities, the study aimed to showcase the model's potential in understanding, decision-making, and interaction across scenarios. Comparative analyses between GPT-4 and its predecessor, GPT-3.5-turbo, demonstrated GPT-4's enhanced adaptability to the environment, with significant improvements in posing questions and forming responses. However, challenges such as the model's limitations in judging and suspecting actions of other players emerged. Reflections on AI's future capability and directions were also discussed. The findings suggest that although GPT-4 exhibits promising advancements over earlier models, there is potential for further development through expanding data and training techniques. The findings also underscore the importance of maintaining an inclusive and unbiased approach throughout this process, suggesting immense potential for Generative AI and its application.