Science AI Journal
Computer Science · Research topic

Open research questions in Artificial Intelligence in Games

63 unresolved questions extracted from the limitations and future-work sections of 503 Artificial Intelligence in Games papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • Traditional video production is a complex, time-consuming, and resource-intensive process that requires significant technical expertise. The system needs to be able to understand natural language descriptions and generate coherent video clips. The system needs to be scalable and able to handle large amounts of data.

    AI -Powered Web Application for Automated Short Video Generation · 2026 · DOI

  • Traditional video production is a complex, time-consuming, and resource-intensive process that requires significant technical expertise. There is a need for automated video creation systems that can reduce the time and effort required for video production.

    AI -Powered Web Application for Automated Short Video Generation · 2026 · DOI

  • The large state space of Ludo. The stochastic uncertainty introduced by the dice roll. The need for effective state encodings to capture the complexities of the game.

    OPTIMIZING DEEP REINFORCEMENT LEARNING FOR BOARD GAMES: A COMPARATIVE STUDY OF DQN AND PPO ON LUDO WITH STRUCTURED STATE ENCODING · 2026 · DOI

  • The paper identifies the challenge of scaling imperfect-information games due to the lack of a formal task model and evaluation metric for hand abstraction. The paper notes the challenge of developing hand abstraction algorithms that can effectively incorporate historical information.

    Beyond Outcome-Based Imperfect-Recall: Higher-Resolution Abstractions for Imperfect-Information Games · 2026 · DOI

  • Solo game development is hard due to the need for a single person to perform a wide range of tasks. AI can be difficult to integrate into game development. There is a need for novel approaches to using AI in game development.

    A Comparative Study of AI Tools and Techniques for Assisting Solo Artists and Programmers in Developing Video Games from Scratch · 2026 · DOI

  • The industry is facing a fundamental dichotomy between economic crisis and technological revolution. The adaptation of archaic cultural codes to the needs of a digital society is a key challenge.

    The Application of Artificial Intelligence Technologies and Novel Prospects for the Representation of Historical and Mythological Content in the U.S. Video Game Industry · 2026 · DOI

  • The need for a far more economical representation that would also be supported by RocksDB and re-writing of the code that generated, became a priority, - Performance of both proSQLite and the Prolog code became limiting issues.

    Chess_db: A framework for working with large chess game datasets. · 2026 · DOI

  • Future research should investigate the effects of acquisitions on employees and social networks - Future research should explore the application of the three processes of pivoting in different contexts

    Playing your cards wisely: Strategy implementation through the three processes of pivoting · 2025 · DOI

  • The paper identifies a gap in the understanding of strategic pivots in established organizations - The paper highlights the challenges faced by established firms in executing strategic pivots

    Playing your cards wisely: Strategy implementation through the three processes of pivoting · 2025 · DOI

  • However, its current computational model has an important limitation: it does not account for deliberate waiting and speed modulation, behaviors ubiquitous in natural environments.

    Dimension lifting in mental space for adaptive behavior in highly dynamic situations · 2026 · DOI

  • The lack of engagement and replayability in current mobile dungeon crawlers. The use of simple swipe or tap interfaces, static level layouts, and predictable enemy behaviours in current mobile dungeon crawlers.

    RISE OF SHADOWS: DEVELOPMENT OF A 2D ANDROID DUNGEON CRAWLER WITH AI DRIVEN GAMEPLAY MECHANICS AND PROCEDURAL GENERATION · 2026 · DOI

  • Applying the methods to other board games with similar characteristics. Extending the research to four-player settings. Exploring other state encodings and algorithms to further improve performance.

    OPTIMIZING DEEP REINFORCEMENT LEARNING FOR BOARD GAMES: A COMPARATIVE STUDY OF DQN AND PPO ON LUDO WITH STRUCTURED STATE ENCODING · 2026 · DOI

  • A falta de estudos sobre a aplicação de IA em diálogos de NPCs em jogos. A necessidade de desenvolver abordagens mais eficientes para gerar diálogos de NPCs.

    A APLICAÇÃO DA INTELIGÊNCIA ARTIFICIAL EM DIÁLOGOS DE PERSONAGENS NÃO JOGÁVEIS EM UM JOGO DE RPG · 2026 · DOI

  • To explore the application of the approach to other domains. To investigate the use of neural network architectures in the system. To evaluate the system on other datasets and in different contexts.

    Computational and Machine Applications of a Narrative Theory Based Symbolic Ontology: Movie Search and Recommendation · 2026 · DOI

  • Prior approaches have limitations in capturing narrative effects. There is a need for a more nuanced understanding of narrative comprehension and its role in media recommendation.

    Computational and Machine Applications of a Narrative Theory Based Symbolic Ontology: Movie Search and Recommendation · 2026 · DOI

  • The development of automated scenario generation methodologies. The development of specialized algorithms tailored to HLSMAC's unique challenges.

    HLSMAC: A New StarCraft Multi-Agent Challenge for High-Level Strategic Decision-Making · 2026 · DOI

  • Current MARL benchmarks have limitations that hinder progress in the field. Most existing benchmarks emphasize micromanagement over strategic decision-making.

    HLSMAC: A New StarCraft Multi-Agent Challenge for High-Level Strategic Decision-Making · 2026 · DOI

  • The paper suggests that developing signal abstraction algorithms capable of exceeding the PAOI bound and targeting the FROI bound is a critical direction for future work. The paper notes that applying the techniques to other domains that involve imperfect information and sequential decision-making is a potential area for future research.

    Beyond Outcome-Based Imperfect-Recall: Higher-Resolution Abstractions for Imperfect-Information Games · 2026 · DOI

  • Developing AI agents that can adapt to varying strategic landscapes without retraining is a central challenge in multi-agent learning. No AI system has yet achieved superhuman performance in competitive Pokémon VGC battles.

    VGC-Bench: Towards Mastering Diverse Team Strategies in Competitive Pokémon · 2026 · DOI

  • In this work, we introduced VGC-Bench, a benchmark designed to evaluate the generalization capabilities of AI agents in the challenging and combinatorially complex environment of Pokémon VGC. Our benchmark includes standardized evaluation protocols, a modular training pipeline, curated human gameplay data, and implementations of a broad set of baseline methods ranging from behavior cloning and reinforcement learning with game-theoretic algorithms like self-play, fictitious play, and double oracle, the strongest of which was able to win against a past World Championships competitor. We also contributed substantial improvements to the Poké-env library, including integration with PettingZoo and environment support for VGC formats, thereby enabling easy adoption by the broader research community. Through extensive experiments, we demonstrated that while current algorithms can attain strong performance in the singleteam setting, they degrade significantly when scaling to multi-team generalization – highlighting a key open challenge in multi-agent learning. By providing a reproducible benchmark and uncovering a difficult generalization frontier, our work establishes a foundation for future progress in robust multi-agent policy learning. Looking forward, we identify five major research directions enabled by VGC-Bench: (1) Generalization to 𝑛 Teams, 𝑛 > 1: Our current agents achieve strong performance in the single-team setting but struggle as the number of teams in training increases. A natural extension is to develop agents that can generalize across team matchups without having performance degrade, and ultimately perform at a superhuman level across arbitrary teams without needing to retrain. Researchers can use the experiments in Tables 3 and 4 to assess progress on this front. REFERENCES 2025. VGCPastes Repository and Tour & Stats Gallery. Google Sheets. Christopher Berner, Greg Brockman, Brooke Chan, Vicki Cheung, Przemysław Dębiak, Christy Dennison, David Farhi, Quirin Fischer, Shariq Hashme, Chris Hesse, et al. 2019. Dota 2 with large scale deep reinforcement learning. arXiv preprint arXiv:1912.06680 (2019). Noam Brown, Anton Bakhtin, Adam Lerer, and Qucheng Gong. 2020. Combining deep reinforcement learning and search for imperfect-information games. Advances in neural information processing systems 33 (2020), 17057–17069. Noam Brown, Tuomas Sandholm, and Strategic Machine. 2017. Libratus: The Superhuman AI for No-Limit Poker.. In IJCAI. 5226–5228. Murray Campbell, A Joseph Hoane Jr, and Feng-hsiung Hsu. 2002. Deep blue. Artificial intelligence 134, 1-2 (2002), 57–83. Saurabh Charde. 2019. Predicting pokémon battle winner using machine learning. (2019). Adam Gleave, Mohammad Taufeeque, Juan Rocamonde, Erik Jenner, Steven H. Wang, Sam Toyer, Maximilian Ernestus, Nora Belrose, Scott Emmons, and Stuart Russell. 2022. imitation: Clean Imitation Learning Implementations.

    VGC-Bench: Towards Mastering Diverse Team Strategies in Competitive Pokémon · 2026 · DOI

  • Integrating more principled causal discovery techniques or controlled interaction settings to bridge the gap between functional dependencies and verified causal mechanisms.

    SOM: Structured Opponent Modeling for LLM-based Agents via Structural Causal Model · 2026 · DOI

  • Existing approaches often entangle opponent modeling with prediction, limiting adaptability in dynamic interactions. There is a need for a framework that can explicitly decouple opponent model construction and opponent prediction.

    SOM: Structured Opponent Modeling for LLM-based Agents via Structural Causal Model · 2026 · DOI

  • The question of how to reliably collect behavioral data at rarely observed information sets. The understanding of when and how the strategy method introduces behavioral distortions.

    Strategy method effects in centipede games: An optimal design approach · 2026 · DOI

  • Future research should explore the long-term impact of AI-assisted game development on the game industry. Future research should investigate the use of AI in replacing human creativity. Future research should evaluate the potential of AI-assisted game development to create games with societal impact.

    A Comparative Study of AI Tools and Techniques for Assisting Solo Artists and Programmers in Developing Video Games from Scratch · 2026 · DOI

  • The lack of variation in the Skinner box could produce results that are specific to a narrow set of conditions, thus giving rise to a failure to reproduce an effect when task characteristics change even modestly. The use of video game engines may not be suitable for all types of behavioral research, and the cost and complexity of creating immersive environments may be a limitation.

    Video game engines as the new “virtual” Skinner box · 2026 · DOI

Most-cited papers in Artificial Intelligence in Games

Most recent work

Find a gap in your own Artificial Intelligence in Games sub-topic

This page shows what the Artificial Intelligence in Games literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Computer Science

Online Learning and Analytics
851 open questions
Digital literacy in education
518 open questions
AI in Service Interactions
517 open questions
Music Technology and Sound Studies
423 open questions
Teaching and Learning Programming
384 open questions
Quantum Computing Algorithms and Architecture
368 open questions
Virtual Reality Applications and Impacts
317 open questions
Intelligent Tutoring Systems and Adaptive Learning
298 open questions
63 open questions have been extracted from the limitations and future-work passages of 503 Artificial Intelligence in Games papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.