LEE- Oyun ve Etkileşim Teknolojileri-Yüksek Lisans
Bu koleksiyon için kalıcı URI
Gözat
Son Başvurular
1 - 5 / 23
-
ÖgeEvaluating performance of large language models in bluff-based card games: A comparative study(Graduate School, 2025-06-27)The aim of this study is to investigate and compare the decision-making performance of multiple agent strategies in a modified bluff-based card game under imperfect information. The study is focused on whether a large language model (LLM) that is prompted through in-context learning (ICL), can generate effective action recommendations when comparisons are made to other agents that use traditional rule-based strategies and reinforcement learning in a game where deception plays a critical role. The central hypothesis is that successful bluff or challenge actions can be adaptively recommended by reasoning-driven agents using LLM by interpreting structured sets of information despite having no training on game-specific reward signals. Our research question is "Can LLM-driven strategy suggest optimal actions by making predictions and inferences about opponents in a bluff-based card game?" This study is expected to contribute to growing research on the application of LLMs in real-time decision-making domains by analyzing the performance of the agents in terms of the success rate of actions and wins. To do that, the model's ability to reason over game states, player histories, and probabilistic cues to select actions in a bluff-based setting is focused on in the evaluation. The game environment is a modified version of the traditional Bluff (also known as Cheat or BS) card game. The modified version of the Bluff game uses a special 24-card deck which consists of 6 suits and 4 cards and is played by three players (agents). At the start of the game, cards are evenly distributed among all players, the first player is chosen randomly and turns continue in a clockwise direction in each game. Cards are kept hidden by players from others during the game. A fixed table rank is used throughout the game. During the gameplay, two actions can be performed by players: challenge and move. When a move is made by players, cards are piled face down in the middle and a certain amount of cards with the table rank and rank value is claimed to be played. This feature allows opponents to be bluffed by players. A bluff move is defined as one in which at least one card does not match the declared rank, while a truthful move is one in which all cards match the declared rank. To simplify the action selection, the challenge phase is limited to one player, as the challenge action may only be performed by the next-moving player. Therefore, the previous move may be challenged by a player if it is believed that it does not match the required rank. If the last played move is a bluff, then all cards in the pile are taken by the previous player; otherwise, all cards in the pile are taken by the player who performs the challenge. The first player who discards all the cards in their hand is defined as the winner of the game. The game is proceeded in turns until one player wins. To create a game simulation framework, the gameplay and rules are first modeled mathematically by using set notation based on our game design. Then, the frequency of action selections along with their corresponding outcomes is derived, and dynamic reward-penalty for each action and game conclusion is defined. Next, the game is modeled which is played by five different agents. Actions are selected by each agent based on its internal strategy logic, and rewards are distributed based on the success of bluffs and challenges. In this methodology, five distinct agents are implemented: Random Agent: This agent selects actions uniformly at random without considering game context. Serves as a performance baseline. State Dependent Agent: This agent uses handcrafted rules based on the current state, such as pile size, card distribution. The agent with minimal logic and contextual awareness as calculating the probability of occurrence of the matching cards in the opponent's hand is modeled. Bayesian Agent: Employs hierarchical priors from historical game data and current game state to evaluate best action for bluffing and challenging. Adapts action preferences based on prior success rates. DQN Agent: A deep reinforcement learning model trained to maximize long-term reward. It maps observed states to derive optimal actions using Q-learning, updating its policy across turns in episodes. To support reinforcement learning, we defined both episode-level rewards (e.g., winning or losing the game) and turn-based rewards for each action type (move and challenge) to guide the learning process of the agent effectively. We design two different learning configurations as Baseline-Oriented Training which is competing against baseline agents to learn stable behavior in known scenarios and self-play training which provides learning by playing against versions of itself to promote generalization and adaptability. LLM Agent: A GPT-based language model (GPT-4o) receives a structured prompt describing the rules and instructions, current state, past behaviors, and probabilistic game summaries such as winning rate. In addition, a simple chain of thoughts by instructions and strategy guidelines to analyze opponent behaviors, predict possible outcomes of possible action scenarios and evaluate risk-reward to select an action is design and implemented. It reasons step-by-step internally but outputs only the final action. To ensure fair evaluation of the LLM agent, an opponent filtering mechanism is implemented, where the LLM is only tested against a selected pool of agents with varied but stable behaviors. This prevents high variance due to opponent unpredictability and allows consistent measurement of reasoning-based performance. The conducted simulation results are analyzed by strategy type, focusing on bluff/challenge success, consistency, and win rate. Action success rates and cumulative performance are included in the metrics. The analysis is conducted by comparing static (Random, State-dependent, Bayesian), learning-based (DQN), and reasoning-based (LLM) agents to determine how performance in uncertain, deceptive environments is influenced by adaptive decision-making. By comparing action decisions of the LLM agent with those of the other strategies, the effectiveness of LLM-driven recommendations is evaluated in the study. The simulation framework is contributed to for understanding how LLMs perform in strategic decision-making tasks where uncertainty and deception are key components.
-
ÖgeA case study on the use of AI-generated music in video game development(ITU Graduate School, 2025)This thesis presents a comprehensive case study on the use of AI-generated music in video game development, examining the role of music in depth. The study included both an observational preliminary study examining the technical capabilities of AI-generated music tools and semi-structured interviews examining the experiences and perceptions of game development professionals. Starting from the historical development of game music and the basic principles of ludomusicology, this research examined the artistic, technical, professional, and ethical dimensions of integrating AI into the video game development field. The research was conducted in two stages. In the first stage, a pilot study was conducted to evaluate the capabilities of existing AI music production tools, Udio and Suno.AI, to produce music for different game genres (puzzle, action-adventure, horror). By observing the performance of the produced music in terms of structural, emotional, technical, and genre harmony, preliminary ideas were obtained on integrating these tools into professional development processes. In the second stage, semi-structured interviews were conducted with eleven people working in the gaming industry. The thematic analysis method was used to analyze the experiences of video game developers and musicians regarding these tools, the purposes for which they prefer them, how they use them in professional production processes, their perceptions on the artistic, technical and creative process, their concerns about issues such as originality, copyright and ethics, and their thoughts on how their long-term impact on the gaming industry will be. According to the findings, AI music tools were primarily used by developers in the early stages, such as idea generation, prototyping, and testing. It was stated that these tools provided advantages, especially for independent teams, in terms of speed, accessibility, and cost. However, due to structural limitations in controlling the music, difficulties in producing cyclical music, and uncertainties about copyrights and originality, participants were hesitant to incorporate this music into their final products. Participants generally evaluated AI as an auxiliary, supportive, and complementary tool, stating that no system can replace human creativity. This thesis contributed to the literature by providing a developer-focused perspective on how AI-assisted music production integrates into the game development process. In addition to technical capabilities, the study's consideration of developers' interactions and user experience with AI provided data on human-AI collaboration. The limited number of tools, types and participants, and the rapid change in AI technologies limited the study. Future research should explore a broader range of AI music tools and game genres, examine how players emotionally connect with AI-generated music, evaluate its impact on immersion compared to traditionally composed music, assess its ability to produce adaptive, interactive, and dynamic soundtracks, investigate integration with interactive audio engines, and develop frameworks for ethical use, legal clarity, and hybrid human-AI collaboration in music production.
-
ÖgeAnalysis on the relationship between human psychology and in-game preferences in terms of bartle's player type(Graduate School, 2025-06-25)Video games have evolved beyond mere entertainment into digital spaces where individuals explore identity and act independently of social norms. With increasing realism, games enable players to reflect their personality traits through their in-game behaviors. This thesis investigates the relationship between individuals' real-life personality traits and their player types in digital games. It focuses on how demographic variables such as age, gender, education, socioeconomic status, and gaming habits influence in-game behaviors. The study employs a quantitative methodology using a questionnaire including the Bartle Player Type Model and the Five Factor Model. The questionnaire was approved by the Istanbul Technical University Ethics Committee and distributed through snowball sampling via social media, resulting in 417 responses. After data cleaning, 411 valid responses were analyzed using correlation, ANOVA, and Tukey's Honestly Significant Difference tests, visualized through Python. The findings revealed no significant differences between those who played massively multiplayer online role-playing games and those who did not, suggesting the applicability of the Bartle model beyond its original context. Consistent correlations were observed across all categories, most notably between Openness and Explorer, and Agreeableness and Socializer. Demographic subgroups generally yielded similar results, though the most notable differences appeared in gender. Female Socializers showed higher Conscientiousness, whereas male Socializers displayed higher Agreeableness. Within the massively multiplayer online role-playing games group, women exhibited stronger Socializer traits than men, whereas men showed stronger Explorer traits than women. Further analysis identified the most variation in the Killer type, while the Achiever type showed no significant differences. Age and education level showed no notable impact. Individuals with higher Agreeableness tended to spend more time playing. Interestingly, Socializer types were more common among participants with lower socioeconomic status, while Explorer traits were more frequent in those with limited massively multiplayer online role-playing games experience. Openness was higher among younger players with lower socioeconomic backgrounds and less weekly game time. The study concludes that both personality traits and demographic characteristics influence player types and gaming motivations. Players with similar personalities exhibited different in-game behaviors depending on demographic context. This highlights the shaping role of demographic factors in the relationship between personality and player type. Notable patterns include gender-based motivational differences: women often viewed games as an escape and engaged more in social interaction despite being highly responsible in real life, whereas men integrated gaming with their real-life personalities, displaying helpful and cooperative behavior. Socioeconomic status also influenced motivation as lower-status individuals played to socialize, while those with higher status played for competition or escape. Additionally, longer massively multiplayer online role-playing games playtime correlated with a shift from exploration to social interaction, indicating the need for mechanics that sustain Explorer-type players. Overall, this research contributes to the understanding of individual differences in digital games, offering insights for player-centered game design and the role of games in identity formation and socialization. This demonstrates that games serve not only as a source of entertainment but also as a space for identity formation and escapism, particularly for disadvantaged demographic groups. It also suggests potential strategies for enhancing the long-term engagement of massively multiplayer online role-playing games. However, the study has limitations. Its limited time frame design prevents observing changes over time, and self-report data may be biased. Although the Bartle model is widely used, it was originally developed for players of Multi-User Dungeons which is the early forms of today's massively multiplayer online role-playing games and may not fully capture the behavior of other player types. The cultural homogeneity of participants limits generalizability. Future studies should address these limitations through longitudinal, behavioral, and cross-cultural approaches.
-
ÖgeIntegrating RNBO into unity: a framework for procedural and adaptive game audio(Graduate School, 2025-06-27)In modern game development, audio plays a critical role in shaping player experience, reinforcing atmosphere, and supporting gameplay. While the industry relies heavily on middleware solutions like FMOD or Wwise to handle sound design and implementation, these systems are sample-based. Relying on pre-recorded assets limits creativity and hinders responsiveness to real-time changes within the game. As games became more complex, the need for procedural, interactive, and real-time audio systems has grown significantly. However, implementing such systems often requires a deep understanding of DSP and programming. This creates a barrier between creative sound design and technical implementation. Max/MSP, developed by Cycling '74, is a visual programming language for creating audio. RNBO is a patching environment inside Max, and its made for exporting software in Max. This thesis explores how RNBO can be integrated into Unity. RNBO allows users to design real-time audio processors visually and export them as in C++ format. This makes it a promising tool for procedural sound. The project examines the RNBO Unity Audio Plugin, which allows RNBO patches to run as natively inside Unity. Although the plugin provides a working base, it also presents several limitations that restrict usability for non-programmers. The main limitations are the need to hardcode parameters in C# and compiling the plugin via command line. To address these limitations, the thesis proposes a custom pipeline that simplifies and improves the process of integrating RNBO patches into Unity. The system introduces a tool that automatically parses RNBO patch metadata and exposes it inside Unity. This eliminates the need for manual parameter declaration in code. This workflow enables sound designers to bind gameplay variables to RNBO parameters directly inside Unity's editor interface. By removing the hardcoded step of identifying and scripting parameter IDs, the process becomes significantly more maintainable. The methodology involves creating a simplified RNBO patch with core parameters (e.g., frequency and gain), exporting it as C++ code, compiling it into a Unity plugin, and testing its integration through both code and editor-driven interaction. The patch is used as a case study to demonstrate the workflow from RNBO design to real-time game interaction. Following this, the proposed tool is implemented to show how metadata parsing can improve accessibility and reduce technical complexity. Findings suggest that while the RNBO Unity Audio Plugin provides a strong technical foundation, its workflow is not optimized for rapid experimentation or non-programmer access. The proposed tooling closes this gap by enabling a more designer-friendly solution. It enhances usability without compromising from the low-latency, platform-independent benefits of native audio plugins. In conclusion, this thesis contributes to procedural and real-time audio in games by presenting a hybrid approach that combines visual sound design with modern game engine integration. It demonstrates that RNBO can function not only as a powerful audio tool, but also as a bridge between creative audio workflows and game development. With further development, the system has the potential to become an alternative or complement to existing middleware, offering more dynamic and expressive sound design options for video games.
-
ÖgeThe role of environmental design methods in player experience: a comparative analysis of procedurally generated and handcrafted game levels(Graduate School, 2025-06-25)Technological advancements in game design have enabled games to become more interactive, contextual, and personalized. These innovations allow for the creation of greater and more spectacular video game environments (everything in a game that is around a player), and they also lead to the development of technologies that assist in optimizing the game production process. The development steps have become more complex in order to meet the expectations of the games, which has increased the demand for new game content. In addition to handcrafted game content, procedurally generated game content which is algorithmic development with minimal human input has begun to be used in games. Upon reviewing the literature, it is evident that further research is required to understand the player experience in procedurally generated game levels. Player experience research in this discipline primarily focuses on 2D games and 2D game content generation. This study is an attempt to enhance research on procedural generation in 3D games and its impact on player experience. One of the primary goals of this study is to understand the differences between procedurally generated levels and handcrafted levels and to explore the effects, if any, of these levels on the players' experience. A mixed methods approach was utilized. The questionnaires used in game studies were examined in the literature, and afterwards, a survey was produced by combining questions regarding player experience. The online survey asked participants to rate the games using a 7-point Likert scale. The study was conducted with a study group of 39 individuals who have prior experience playing the games utilized in the research. Following the completion of this part of the survey, further semi-structured questions were asked. The study involves players sharing their gaming experiences deeply . Participants were requested to provide how the game environment, including open and closed spaces and interactable objects, influences their sense of immersion, enjoyment, and overall satisfaction with the game. When selecting three games for the research, it is ensured that one of them is procedural, the other one is handcrafted, and the last one combines both approaches. The first one is No Man's Sky which developers used procedural generation to create enormous and endless landscapes and achieve an optimal production balance for game content. The second is Outer Wilds, in which developers methodically designed each planet's to precisely moderate the flow of the game and ensure that players experience essential story elements. The third is Starfield which has over 1000 planets, some of which are hand-crafted and unique to each player, most are procedurally generated. As a result, the aim is to uncover the impact of the relation between environmental design methods and player experience, providing guidance for designers and artists in determining methods for game development. In the survey findings, the queries in the first section were assessed on a scale of 1 to 7, and their results were averaged. The second section carefully examined the semi-structured questions that were given to the participants, and the results were analyzed. The findings indicate that the game environment and its creation process significantly influence player experience, however they do not primarily determine the overall gaming experience.
- 1 (current)
- 2
- 3
- 4
- 5