The Art of virtual storytelling: leveraging LLM for integrating synthetic player narratives in online gaming environments

Abstract

In contemporary online gaming, generating immersive experiences that captivate players is pivotal. The evolution of online gaming has fostered the integration of Synthetic Player entities, enabling dynamic and engaging interactions within the virtual gaming sphere. This thesis concentrates on equipping synthetic players, namely bots, with advanced capabilities to articulate their in-game experiences and thoughts and engage in virtual discussions mimicking human-like social communication. Harnessing Large Language Models (LLMs) such as ChatGPT\cite{openai2024gpt4}, the research delves into crafting character narratives driven by game state information and events. The study outlines the application of heuristics to identify and incorporate compelling moments, weaving engaging and contextually rich narrations within the digital game-play landscape. Furthermore, the research investigates the extraction of nuanced game features, including sentiments, combat interactions, spatial positioning on the map of the game, strategic insights, and collaborative efforts enhancing player engagement and strategic decision-making, to convey a comprehensive and captivating gaming experience. By utilizing LLM-driven text generation, the study highlights the strategic portrayal of in-game emotions and situational descriptions, ensuring a seamless, balanced blend between player achievements and contextual game-play dynamics. The research accentuates the significance of character-driven narratives, amplifying player engagement and connection to the game-play, thus fostering enriched gaming experiences distinguished by captivating and character-driven storytelling.