Advancing Realistic Non-Playable Characters Conversations in Juragan Fauna with GPT-3.5

This research explores the application of the GPT-3.5-based Natural Language Processing (NLP) model to enhance interactions between Non-Playable Characters (NPCs) and players in the simulation game Juragan Fauna. The study aims to evaluate the model's effectiveness in improving the accuracy and speed of NPC responses during player interactions. Using the Unity game engine, GPT-3.5 was integrated into various NPC roles, allowing for dynamic and contextually relevant responses to player inquiries. Performance evaluation focused on two key aspects: response accuracy and response speed. The results show that the model achieved an average response accuracy of 80.4%, demonstrating its ability to deliver relevant answers based on NPC roles and game context. Additionally, the average response time of 0.96 seconds ensured smooth, uninterrupted gameplay. Players experienced a more immersive and engaging environment, with NPCs responding to open-ended questions in real time. However, challenges such as the dependency on stable internet connections and the high operational costs associated with using API-based models were identified. The study provides several recommendations for developers, including optimizing NPC response lengths to balance response speed and cost-efficiency. This research highlights the potential of GPT-3.5 to revolutionize NPC interactions in simulation games, paving the way for more advanced and realistic gameplay experiences.