Generating Clue-Driven Investigative Game Narratives with Large Language Models

Fun-Aware Modeling and Narrative Generation Using LLMs for Role-Playing Games

Creating engaging narratives for role-playing games (RPGs) remains a challenge. Traditional handcrafted and branching story systems offer limited adaptability, while large language models (LLMs) typically optimize linguistic quality rather than player enjoyment. To address this gap, this study proposes a fun-aware LLM-based narrative generation framework that treats fun as a dynamic, multi-dimensional signal guiding iterative story generation. The framework evaluates narrative segments using six interpretable features: surprise with coherence, agency density, tension oscillation, narrative payoff ratio, tonal variability, and rule-bending moments. A human-subject study involving 24 participants showed that fun-aware narratives significantly improved perceived enjoyment, player agency, novelty, and replayability compared with baseline LLM-generated narratives while maintaining narrative coherence. Correlation analysis further revealed strong agreement between computational fun estimates and human enjoyment ratings (<inline-formula> <tex-math notation="LaTeX">$r = 0.71$ </tex-math></inline-formula>), supporting the validity of the proposed metrics. To further evaluate the framework, 80 paired RPG narratives comprising 944 narrative segments were generated under controlled conditions. Compared with baseline narratives, the fun-aware framework produced higher composite fun scores (<inline-formula> <tex-math notation="LaTeX">$M = 5.82$ </tex-math></inline-formula> vs. 4.71), greater event novelty (<inline-formula> <tex-math notation="LaTeX">$M = 6.77$ </tex-math></inline-formula> vs. 4.91), and richer branching structures (<inline-formula> <tex-math notation="LaTeX">$M = 2.34$ </tex-math></inline-formula> vs. 1.58 choices per segment). These findings suggest that modeling fun as an explicit design objective can improve both player-perceived engagement and computational measures of narrative diversity and interactivity without compromising narrative coherence.

Quest of Aivengarde: Comparative Study of Player Experience Across LLM Dialogue Systems

Non-player character (NPC) dialogue plays a crucial role in games. Narrative-driven video games in particular depend on NPCs to help shape the player experience by contributing to narrative, immersion, and player agency. Current technologies allow large language models (LLMs) to create dynamic, context-sensitive dialogue for NPCs, yet their impact on player experience remains underexplored. Quest of Aivengarde is a custom-built role-playing game developed as a research testbed for comparing four dialogue system designs: a static control version and three LLM-driven variants that rephrase, hybridize, or fully generate NPC dialogue. Building on previous pilot and demo studies, this paper integrates the system design and full empirical evaluation to examine how levels of generative agency affect interaction quality, immersion, and player engagement. Using mixed-methods analysis of behavioral logs and post-game surveys from 64 participants, results suggest that fully open-ended LLM dialogue fosters longer, more natural conversations and is particularly engaging for casual players. We discuss design trade-offs between narrative control and conversational freedom and propose a practical framework for selecting and combining LLM-driven dialogue approaches in future game development and research.

Transdiegetic Sound in Narrative-Driven Strategy Games

Game sound, as a feedback mechanism, functions along a diegetic continuum that bridges the boundary between player, game systems, and narrative. In narrative-driven strategy games, this continuum becomes especially important, as audio must communicate diegetically relevant information about story and character, while also supporting complex decision making. This raises key questions about how transdiegetic sound relates to design, narrative, and player experience for this genre. Specifically, how transdiegetic sound is incorporated into feedback systems in narrative-driven strategy games, and what kind of information they are used to communicate. In this paper, we examine the use of transdiegetic sound as form of feedback by analyzing 11 commercial narrative-focused strategy game titles. We map how such audio cues are designed for both narrative and gameplay functions, the specifics of how they traverse the diegetic boundaries, and discuss how these designs may shape player experience. Building on related work on transdiegetic sound in games, this study documents recurring design patterns and cues, and explores their communicative role within the context of narrative-driven strategy games.

Orchestrating Emergent Storytelling with Embodied Multi-Agent Systems

Narrative Gold Digging: Sifting and Narrating Stories from a Procedural Simulation

This paper reports on an experiment in formalizing and implementing “story-sifting” heuristics to automatically identify likely “interesting stories” from a large mass of events transpiring from the Chroniqueur procedural social simulation. Following preliminary experimentation with various concepts inspired by narratology, we proceeded to formally evaluate “unexpectedness” and “dramatic situations” as key heuristics to sample six stories from a simulation run and rendered each as two short stories, one written by a human author and the other by a LLM (in order to help distinguish the effect of narrative from story). These were scored by 24 participants for narrative appreciation. Inconclusive results about heuristics making any significant difference prompted self-reflection on what makes emergent narratives interesting (in a distinct way from traditional fiction), pointing both at “story effects” (some selected event chains constituting better story material) and “narrative effect” (better narrativization being able to cope with poor story material).

How LLMs Build Fictional Worlds: Setting and Narrative Space in AI-Generated Creative Storytelling

In this paper, we analyze how Large Language Models (LLMs) employ worldbuilding strategies, focusing on setting as one measurable dimension of storyworld construction. We compare 1,000 AI-generated stories per model in English and German with human-authored fiction from Project Gutenberg. Building on prior work, we operationalize setting through five types of narrative space:"action","perceived,""visual,""descriptive"and"no space", identified using fine-tuned BERT classifiers for German and English. We generate narratives using GPT 4.1, LlaMA 3.3, Mistral 3.2, and Gemma 3 and compare their spatial distributions to a human-authored baseline. We find that human-authored texts predominantly employ"action space,"grounding narratives in embodied character-environment interaction, whereas LLMs systematically overproduce"perceived space,"emphasizing atmosphere and affect. This divergence remains stable across narrative time. Overall, our findings show that LLMs exhibit worldbuilding patterns that differ consistently from human-authored fiction in ways that are both model-specific and language-sensitive.