In Social Deduction Games (SDGs) such as Avalon, Mafia, and Werewolf, players conceal their identities and deliberately mislead others, making hidden-role inference a central and demanding task. Accurate role identification, which forms the basis of an agent's belief state, is therefore the keystone for both human and AI performance. We introduce CSP4SDG, a probabilistic, constraint–satisfaction framework that analyses gameplay objectively. Game events and dialogue are mapped to four linguistically agnostic constraint classes—evidence, phenomena, assertions, and hypotheses. Hard constraints prune impossible role assignments, while weighted soft constraints score the remainder; information-gain weighting links each hypothesis to its expected value under entropy reduction, and a simple closed-form scoring rule guarantees that truthful assertions converge to classical hard logic with minimum error. The resulting posterior over roles is fully interpretable and updates in real time. Experiments on three public datasets show that CSP4SDG (i) outperforms LLM-based baselines in every inference scenario, and (ii) boosts LLMs when supplied as an auxiliary "reasoning tool." Our study validates that principled probabilistic reasoning with information theory is a scalable alternative—or complement—to heavy-weight neural models for SDGs.

Open-world missions often rely on repeated formulas, yet designers lack systematic ways to examine pacing, variation, and experiential balance across large portfolios. We introduce the Mission Action Quality Vector (MAQV), a six-dimensional framework—covering combat, exploration, narrative, emotion, problem-solving, and uniqueness—paired with an action block grammar representing missions as gameplay sequences. Using about 2200 missions from 20 AAA titles, we apply LLM-assisted parsing to convert community walkthroughs into structured action sequences and score them with MAQV. An interactive dashboard enables designers to reveal underlying mission formulas. In a mixed-methods study with experienced players and designers, we validate the pipeline’s fidelity and the tool’s usability, and use thematic analysis to identify recurring design trade-offs, pacing grammars, and systematic differences by quest type and franchise evolution. Our work offers a reproducible analytical workflow, a data-driven visualization tool, and reflective insights to support more balanced, varied mission design at scale.

Large scale control of nonplayer agents is central to modern games, while production systems still struggle to balance several competing goals: locally smooth, natural behavior, and globally coordinated variety across space and time. Prior approaches rely on handcrafted rules or purely stochastic triggers, which either converge to mechanical synchrony or devolve into uncorrelated noise that is hard to tune. Continuous noise signals such as Perlin noise are well suited to this gap because they provide spatially and temporally coherent randomness, and they are already widely used for terrain, biomes, and other procedural assets. We adapt these signals for the first time to large scale AI control and present a general framework that treats continuous noise fields as an AI coordinator. The framework combines three layers of control: behavior parameterization for movement at the agent level, action time scheduling for when behaviors start and stop, and spawn or event type and feature generation for what appears and where. We instantiate the framework reproducibly and evaluate Perlin noise as a representative coordinator across multiple maps, scales, and seeds against random, filtered, deterministic, neighborhood constrained, and physics inspired baselines. Experiments show that coordinated noise fields provide stable activation statistics without lockstep, strong spatial coverage and regional balance, better diversity with controllable polarization, and competitive runtime. We hope this work motivates a broader exploration of coordinated noise in game AI as a practical path to combine efficiency, controllability, and quality.

Roguelike and auto-battler games derive much of their replayability from randomized progression, rewards, and encounter sequencing, yet this randomness rarely extends to genuinely diverse game mechanics and content such as skills or combat units. Existing procedural systems for such content usually depend on fixed rule spaces, hand-authored templates, or proxy spell engines, which limit their ability to produce outputs that are semantically meaningful, valuable in play, and sufficiently expressive. We present Neural Spore Genesis, a browser-based roguelike auto-battler in which players buy elemental cells, arrange them on a 5 × 5 grid, and compile that spatial layout into a combat skill for auto-battle. To support this task, we build a high-expressivity structured skill framework covering a broad range of 2D auto-battler combat behaviors, and evaluate local open-weight language models against deterministic, template-based, random, and calibrated sampler baselines. Across 6 generator conditions and a 105-configuration corpus evaluated with 11 metrics, local LLMs achieve near-perfect element coherence while maintaining high intra-config diversity, a combination the baselines do not match. This work offers a practical evaluation environment for intent-aware procedural mechanic generation and supports future research on semantically grounded content systems for high randomness games.

Single-player deduction games are a canonical form of hidden-information reasoning. Agents iteratively issue actions (queries) and receive deterministic feedback, thereby shrinking the information set of feasible secret codes. Classical search techniques-such as Information-Set Monte-Carlo Tree Search (ISMCTS) or the entropy-driven Information-Set Entropy Search (ISES)-handle these games by sampling or by fully enumerating states, but both methods encounter difficulties when the combinatorial space explodes. This paper introduces a constraintpropagation variant of ISES that models the information set as a constraint-satisfaction problem (CSP) and applies the AC-3 arcconsistency algorithm after every observation. By aggressively pruning unsupported variable values before entropy evaluation, the method eliminates a large number of impossible states and accelerates inference without sacrificing optimality. Using several single-player deduction games from the Deduction Game Framework as case studies, we show that constraint propagation significantly enhances the efficiency of ISES.

Visual guidance plays a crucial role in level design, while prior work has largely relied on qualitative observations. This study presents a novel quantitative framework for evaluating visual guidance during level transitions in 3D role-playing games. By integrating analyses of depth maps derived from raycast grids with high-resolution RGB image sequences from our primary dataset in Dark Souls III, we quantify metrics such as luminance dynamics, chromatic complexity, and spatial depth distribution. This bimodal analysis separates geometric factors (from depth) and perceptual factors (from color), thereby clarifying how specific visual cues consistent with design principles such as spatial funneling and chromatic contrast-influence player navigation and immersion. Our initial empirical findings, derived from strictly quantitative and numerical analyses, suggest that the synergy between geometric constraints and perceptual cues provides an effective framework for both validating design principles and identifying navigation pitfalls in level transitions. These results not only provide a formal, data-driven understanding of level design but also offer actionable insights for creating more intuitive virtual environments and establishing evaluation criteria for future procedural level generation.

Action windows—spatiotemporal regions enabling player's safe execution of key in-game actions—are foundational to game task planning, yet their automated generation remains underexplored. In stealth games, for example, level designers carefully create guard patrols and environment layouts. However, critical tasks such as planning assassination routes for high-value targets (VIPs) still depend heavily on manual tuning. This work formalizes VIP task planning as the problem of automatically generating a path through a predefined environment with guard patrols, such that VIP's path contains player's safe action windows that are temporally and spatially dispersed, while maintaining coherence and meaningful interactions with environmental elements. We introduce two approaches: (1) an evolutionary optimization approach that is efficient in generating diverse routes by balancing multiple objectives, and (2) a constraint-driven safe-block search method that guarantees optimal sequences under strict design thresholds. Initial experiments validate that the evolutionary method generates diverse, high-dispersion routes with rapid runtimes, whereas the safe-block approach enforces hard constraints with predictable performance. Both methods integrate directly with existing level and patrol data, offering scalable solutions for automated stealth mission generation.

Procedural Content Generation (PCG) has long enabled efficient and varied game level creation.However, integrating high-level design intentions and game mechanics into complex 3D environments remains challenging.This paper introduces a comprehensive framework that transforms narrative-level descriptions into playable 3D game levels.First, Large Language Models (LLMs) parse natural language descriptions of game environments into a structured Game Level Description Language (GLDL), capturing essential spatial constraints.Next, we model level generation as a Facility Layout Optimization problem, ensuring that facility placements and configurations adhere to specified design criteria.Through comprehensive experiments, including automated constraint evaluations and agent-based simulations, our approach ensures both the feasibility and stability of the constraints extracted from textual descriptions.We confirm that the resulting game levels remain interactive, reasonable, and controllable to their original specifications.