A Social Network Generator for Games Evaluated Against a Real NPC Network with GPT-Generated Node Attributes

This study presents a novel approach for generating social networks for non-player characters (NPCs) in video games through an automated graph generation method. By leveraging existing methodologies and enhancing them in our generator, the aim is to produce NPC networks that closely mimic the structure and properties of agent social networks observed in video games. The method integrates attribute-aware graph generation to ensure realistic and complex network interactions among NPCs. A comparative analysis with an NPC network of a video game which has ChatGPT-generated node attributes demonstrates the effectiveness of our generator in creating more interconnected and dynamic NPC social structures. The findings suggest that our generator is a robust tool for researchers and game developers to simulate NPC interactions, thus enhancing the narrative and interactive depth of video games.