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.
The use of personality to predict user preferences or select a job candidate are practical examples of how personality can be used in a variety of fields. But to assess someone's personality is not easy. The most commonly used techniques are personality questionnaires, but they are subject to social desirability bias and false responses, limiting the accuracy of personality assessment. Implicit techniques have also been used, but they require great amounts of user interactions with the system, and most of them only focus on broader personality dimensions. To overcome those limitations, we developed a short-duration mobile serious game, Coin Catcher, as a concept proof to implicitly measure the granular personality trait morality in less than 5 min. The game uses concepts related to normative morality, proposing dilemmas that the player must solve in order to progress in the game. Experiments with real users were conducted ($n=96$), showing morality can be implicitly determined by a simple game in a short period of time, without the bias associated with personality questionnaires. Additionally, correlations with other personality traits, such as altruism, cooperation, modesty, sympathy, and anger were also found, showing the game has the potential to measure several granular personality traits that better characterize a person.
This paper examines the impact of AI-generated versus human-written dialogue in a murder-mystery game on player experience. We manipulated both the game dialogue type (AI-generated vs. human-written dialogue) and the game type description (AI-labelled vs human-written-labelled) to assess effects on overall player experience (PX), immersion, and perceived autonomy. Results show that player expectations, modulated by the game type description, had a greater impact on overall PX than the actual game type. Players who thought that they played a game with a human-written narrative reported more positive overall PX, higher immersion and perceived autonomy regardless of which version they actually played. However, interacting with the AI-generated dialogue version led to higher perceived autonomy, suggesting that AI-driven dialogue can enhance perceived freedom without necessarily improving the overall PX. These findings show that player expectations, in response to game descriptions, play a central role in the experiences of players. This work contributes to understanding the influence of narratives and the extent to which AI features increase player engagement.
This paper presents a gamified Virtual Reality (VR) application integrated with a second-screen Web interface to support exposure therapy for children with sensory food aversion. The system allows therapists to configure therapeutic sessions, monitor user actions in real time, and adapt the virtual environment according to the patient's responses. The Web interface records session data, including behavioral performance and, when available, biosignals such as EEG and ECG, building a historical profile for each patient. The solution was developed through a co-design process with therapists to ensure usability and alignment with therapeutic practice. Preliminary results show the system's potential to improve engagement, support therapist intervention, and enable future developments, such as the integration of real-time neurofeedback to adapt the experience based on physiological data.
Modern racing games often take place on race tracks with huge surroundings, including forests, fields, cities, or mountains. While not being important to the racing itself, the surroundings must be present, dense, and detailed to enhance the gameplay feel, and must be performance-optimized for a smooth gaming experience. Creating these surroundings manually is a very laborious task, even for a skilled level designer. Also, manual design makes it prohibitively expensive to modify the track layout later. This makes the surroundings an ideal candidate for procedural content generation, which is nowadays extensively used in many areas of video game level design. In this paper, we present a method for controllable track surroundings generation. It generates a one square kilometer map in a matter of minutes, orders of magnitude faster than the manual process. It features several different biomes, various types of assets and vegetation, and can be further extended using custom assets. The generated race tracks and their surroundings are easily adjustable, making the level design iterative and easier to perfect.
As artificial intelligence (AI) continues to transform industries across society, it is having a profound impact on healthcare and biomedical research. To prepare students for this evolving landscape, there is a growing need for learning experiences that build foundational AI literacy and connect to real-world career pathways. However, most middle school students lack access to engaging and personally meaningful opportunities in AI education and career exploration. NarrativeCentered Learning (NCL) offers powerful affordances for contextualizing AI literacy through engaging storylines and role-based problem-solving. In parallel, Social Cognitive Career Theory (SCCT) emphasizes how students' beliefs about their abilities, expectations about outcomes, and personal goals influence the development of their academic and career trajectories. This paper introduces a design framework that integrates NCL and SCCT to foster both AI literacy and health-related career interest. We apply this framework to the design of a narrative game in which students take on the role of a medical intern investigating virtual patient cases using AI tools. We report findings from a usability study with 25 middle school students who played the game's first episode and participated in structured focus groups. Student feedback suggests that the game supported engagement, sparked curiosity, and encouraged emerging career interest. These findings offer preliminary support for the framework and inform the design of career-connected AI learning.
Competitive environments, whether in traditional sports or competitive online gaming, often foster intense emotions and harsh language. In Competitive Online Multiplayer Games, where players are not physically present together, verbal toxicity primarily manifests through voice and text chat. As online gaming remains a predominantly male-dominated space, much of this toxicity disproportionately targets non-male players. Behind the veil of anonymity, non-male players frequently face gender discrimination, hate speech, and unwanted sexual advances. To effectively address these issues, game developers must first understand how such harassment manifests and how it is perceived by players. This work examines gender discrimination in Competitive Online Multiplayer Games through two online surveys. The first survey gathered reports from 61 non-male players who had experienced harassment, resulting in 171 coded statements describing gender discrimination. A second survey presented these statements to 281 players across all genders, who rated their perceived severity and authenticity. Our results indicate that gender discrimination is generally perceived as equally severe across all gender groups. However, notable differences are visible in how players respond to harassment, with those from marginalized groups exhibiting higher levels of rumination. This highlights the compounded impact of gender-based toxicity in online gaming. These insights provide valuable direction for game developers seeking to create more inclusive and supportive gaming environments.
Game-playing agents for abstract board games almost universally employ state-space search for thinking ahead. Ideally, when new search enhancements are introduced, their effectiveness is investigated in different game-playing domains to better highlight their relative strengths and weaknesses. We present a highly configurable open-source framework for generating synthetic state-spaces for combinatorial games to better facilitate an in-depth exploration of new search enhancements.
We introduce Pixels2Play-0.1 (P2P0.1), a foundation model that learns to play a wide range of 3 D video games with recognizable human-like behavior. Motivated by emerging consumer and developer use cases-AI teammates, controllable NPCs, personalized live-streamers, assistive testers-we argue that an agent must rely on the same pixel stream available to players and generalize to new titles with minimal gamespecific engineering. P2P0.1 is trained end-to-end with behavior cloning: labeled demonstrations collected from instrumented human game-play are complemented by unlabeled public videos, to which we impute actions via an inverse-dynamics model. A decoder-only transformer with auto-regressive action output handles the large action space while remaining latency-friendly on a single consumer GPU. We report qualitative results showing competent play across simple Roblox and classic MS-DOS titles, ablations on unlabeled data, and outline the scaling and evaluation steps required to reach expert-level, text-conditioned control.
Video games represent a unique type of complex software project, despite differing in aspects such as methodologies, programming languages, and design patterns. Similar to traditional software, games can be released as open-source software (OSS) projects, thus fostering collaboration and knowledge sharing among developers. However, one of the primary challenges in this domain is the difficulty in finding source code for game development, which is often scattered across various repositories and platforms. Moreover, the lack of proper guidelines to document game projects is still missing, thus worsening this issue. In this paper, we envision a set of initial guidelines leveraging a mining-based methodology considering two different OSS platforms, i.e. itch.io and GitHub. First, we collect data from 765 open-source games from the itch.io platform and map the retrieved games to the corresponding repositories on GitHub, searching for documentation and source code. We further refine the list of games by manually analyzing the repositories, focusing on the quality of the documentation and the presence of source code, ending up with 613 games. On top of this gold set, we elicited a set of ten reproducibility guidelines specifically tailored for games. Our results show that the majority of the games do not have source code available, and the documentation quality is generally low, apart from games with high-rated GitHub projects. In addition, we provide a set of takeaways that can be further investigated by extending the provided guidelines. We believe that our dataset and methodology can be used as a starting point for future research in this domain, providing insights into the challenges and opportunities for assisting newcomers to game development using open-source projects.
Esports analytics has gained traction in recent years, leveraging machine learning (ML) to predict in-game events and enhance strategic decision making. This study develops structured datasets from Counter-Strike: Global Offensive (CS: GO) competitive matches to address two predictive tasks: round result prediction and player death prediction. Using data from professional Dust 2 matches in 2022, we extract key playerand team-level features such as health, spatial positioning, and economy-related metrics. Various ML models, including Logistic Regression, Decision Trees, and XGBoost, are evaluated and benchmarked against random guessing and majority-class baselines. Results show that XGBoost consistently outperforms other models, effectively capturing gameplay dynamics and providing accurate predictions. These findings offer valuable insights for esports strategy optimization, coaching, real-time decision support, and applications in live-game analysis and betting. The dataset and methodology also establish a foundation for future esports research and predictive modeling across other game environments.
As one of the largest entertainment sectors globally, the gaming industry has already surpassed the music industry in revenue and is expected to continue its steady growth in the years ahead. Beyond leisure, gaming has become a competitive arena, particularly with the rise of esports, where games like CounterStrike (CS) draws millions of players and viewers worldwide. With the expansion of online gaming and virtual environments, new security challenges emerge that require advanced solutions. One promising approach is biometric identification based on ingame behavioral data, which is non-invasive and challenging to replicate. Although research has demonstrated effective player identification in Virtual Reality and turn-based games like chess, games like $C S$ introduce unique challenges due to their dynamic high-speed interactions. This study explores behavioral biometrics to identify esports players based on their in-game behaviors. We use a multiclass Random Forest Classifier to analyze in-game movement, positioning, and weapon choices to identify individual players accurately. Our models achieve up to 98 % accuracy based on a combined dataset of 360 players across 320 matches. This data is divided into eight separate datasets for each map. These findings contribute to behavior-based player biometrics and identification, with applications in player analytics, team strategy optimization, and security in competitive gaming.
Accurately modeling and understanding player experience is crucial for designing engaging puzzle games. To achieve this, a common approach involves collecting diverse user data to train predictive playtesting models that mimic player behavior. However, existing data-driven methods often lack the ability to capture the full range of player strategies and require extensive feature engineering and network architecture modeling. This limitation becomes particularly evident when new game mechanics or features are introduced, which necessitate continual adjustments to the models. To address these challenges, we propose a more generalized representation that reduces — or even eliminates — the need for ongoing feature-engineering maintenance. Specifically, we investigate two general-purpose network architectures: (a) a transformer-based model (BERT) and (b) a graph attention model (GAT), both of which are designed to effectively capture the relational structure of Candy Crush Saga (CCS) game boards. Our experiments compare these approaches to Convolutional Neural Networks (CNN) baselines, revealing better performance on challenging board configurations and underscoring the benefits of our generalizable representation.
Training generalist agents capable of performing well across diverse environments is a significant goal of reinforcement learning (RL). Current state-of-the-art methods for multi-game RL rely on offline datasets, and often discard the policy used to gather trajectories despite its potential to provide a rich learning signal. In this paper, we revisit policy distillation (PD) for multi-game RL and introduce a new framework called Bootstrap Your Own Teacher (BYOT) that extends policy distillation to the online-RL setting. BYOT alternates between two phases: (i) game-specific finetuning and (ii) distilling bootstrapped teachers back into a shared multi-game policy. By directly regulating the multi-game learning dynamics in policy space, BYOT balances training without explicit gradient adjustments or reward normalization, whilst being highly parameter efficient. Our framework is empirically validated for both online and offline multi-game learning on the Atari-40 benchmark. BYOT outperforms all prior online Atari-40 multigame agents, achieving an IQM human-normalized-score (HNS) of 152.7 %. Moreover, by adopting state-of-the-art PPO teacher agents—contrasting the widely-used datasets from weaker DQN agents—and policy distillation, we more than triple the leading IQM-HNS in offline settings to 369.5 %, whilst using significantly fewer parameters. Overall, our results emphasise the power of distillation in multi-game settings.
The current video game industry faces soaring development costs and intensified market competition, underscoring the critical demand for more systematic game design methodologies. While existing approaches (e.g. MDA framework) provide analytical tools at the theoretical level, they lack sufficient operational guidance during the transition from conceptual design to prototype development. This study proposes Gameplay Evolution (GE), a novel methodology inspired by inheritance, variation, selection, and adaptation mechanisms from evolutionary biology theory. Through a standardized five-stage operational process, including evolutionary direction research, gameplay genes identification, iterative refinement through inheritance and variation, randomized selection validation, and adaptive propagation, GE systematically bridges the transitional gaps in traditional methods. To validate its practicality, this research integrates historical case analyses and experiments across specific game genres. Results indicate that GE enables designers to efficiently generate gameplay variants with enhanced playability while optimizing market adaptability.
Recording and broadcasting Go games in real time remains a challenging task, as traditional methods rely on electronic boards or manual transcription, both of which have limitations in cost, convenience, and accuracy. This paper presents an approach for automatically detecting board states and reconstructing move sequences using computer vision and machine learning techniques. A Convolutional Neural Network (CNN) and a Random Forest (RF) classifier were trained to recognize individual board intersections, demonstrating robust performance across varying lighting conditions and board configurations. To determine the move order when multiple stones are detected simultaneously due to disrupted images, KataGo was used to evaluate all possible variations of the new moves, based on score lead differences. The results indicate that the actual game sequence tends to have a lower mean score fluctuation compared to randomized move orders, supporting the hypothesis that KataGo can provide reasonable reconstructions of move sequences. Evaluations using recorded games from strong amateur and professional players demonstrated promising accuracy in move detection and sequence inference. The implementation is open source, serving as a foundation for further research and development toward automated Go game transcription and real-time broadcasting.
In esports broadcasting, human observers are tasked with providing viewers with a satisfying view of the event. Existing approaches focus primarily on detecting events and often fail to address how the viewing camera should transition after an event is detected. In this study, we defined the viewing process as event detection and viewport tracking, which is the process of following the detected event. In addition, we focus more on viewport tracking and propose a ConvLSTM-based encoder-decoder model based on the viewing data of professional observers. The model aims to automatically predict how observers follow events by learning their temporal patterns and spatial characteristics, and to build a more natural and effective esports automatic broadcasting system. The viewport tracking model was evaluated using Intersection over Union(IoU) and achieved a performance of 0.6689. This result represents a novel attempt to model the viewing sequence after event detection, which offers the potential to enhance the naturalness of automated broadcasting systems.
Rehabilitation for patients with muscle dystrophy is based on performing a lifelong program of regular exercise in coordination with the therapist. However, traditional physiotherapy can be repetitive and boring, which can lead to frustration. In recent years, Serious Games for Exercise Rehabilitation (SGER) have shown promise in making these repetitive tasks more engaging. This paper targets how Virtual Reality (VR) and Augmented Reality (AR) can be used to create fun and motivating SGER. The development process was supported by a rehabilitation team in a neuromuscular research-clinical center responsible for muscle dystrophy treatment. Therefore, medical knowledge comes from clinicians, including a doctor, with the goal of designing immersive environments that encourage therapeutic movements by playing. Based on VR and AR, 15 mini-games were designed to replace traditional rehabilitation exercises, focusing on improving upper limb movement and adaptability to different types of muscle dystrophy. These games were tested with 15 patients in the presence of a doctor to acquire feedback. The results show that the patients were satisfied with the games specifically chosen for their type of disease. With future improvements, those games can be improved and integrated into the therapy plan. This work provides a foundation for the use of game-based systems in realworld rehabilitation.
and testing AI agents in game design and gameplayrelated tasks, specifically within Pokiemon. In this competition, players must both strategically battle to maximize their win rate and anticipate their opponents' teams to improve their chances of victory before the battles. In this paper, we expand the existing competition by introducing a new track centered on a novel game balance issue, Rules Balance. Rather than modifying game units to encourage specific usage rates of Pokemon or teams during a competitive season, the focus is on reconfiguring the Pokiemon battle rules to incentivize non-damaging moves (moves with delayed rewards) by rational agents. This AI assistant tool relieves game designers of the costly and time-consuming task of testing numerous parameter combinations to achieve desired move usage patterns. This allows designers to allocate more time to the creative process of developing new game mechanics. We present and evaluate an initial agent based on a genetic search that competes against a tree-search battle agent. While the battle agent is motivated to devise the most efficient plan of action to win, the balance agent adjusts the rule parameters to encourage the rational agent to use specific desired moves.
With the rapid development of the gaming industry and the increasing popularity of multiplayer online games, realtime voice communication has become a crucial tool for team collaboration and tactical exchanges. Automatic Speech Recognition (ASR) technology plays a vital role in modern gaming by converting voice commands into text, enabling efficient communication among players. However, existing general-purpose ASR systems face significant challenges in gaming scenarios due to the unique characteristics of in-game communication, such as short phrases, rapid speech, game-specific jargon, and environmental noise. These limitations often lead to frequent recognition errors, increasing communication costs and negatively impacting the overall gaming experience. Furthermore, the scarcity of domainspecific ASR training data exacerbates these issues, hindering system optimization. To address the challenges of ASR systems in gaming scenarios, this study proposes the GO-AEC (Gaming-Oriented ASR Error Correction) framework. The framework leverages the generative capabilities of large language models (LLMs) and employs Retrieval-Augmented Generation (RAG) techniques with a gamespecific knowledge base to better adapt to gaming environments. Additionally, we introduce a data augmentation strategy that combines LLMs with text-to-speech (TTS) techniques to enhance the diversity and robustness of game-specific datasets. The GO-AEC framework consists of three key modules: the data augmentation module, the N-best hypothesis-based LLM correction module, and the dynamic knowledge base module powered by RAG. Experimental results demonstrate that, compared to baseline methods, the proposed framework reduces the character error rate (CER) by 6.22 % and the sentence error rate (SER) by 29.71 %. These findings indicate that the GO-AEC framework effectively addresses the challenges of ASR error correction in gaming scenarios.
