580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can AI achieve general competence across diverse strategy video games without relying on game-specific knowledge?
This research theme addresses the challenge of developing artificial intelligence agents that can competently play a broad range of strategy video games, including games they have not encountered before, without relying on handcrafted heuristics or domain-specific rules. It matters because unlike specialized AI tailored for a single game, such as chess or Go, general AI methods for strategy games offer pathways towards more flexible, adaptive, and robust decision-making agents. This line of research fosters progress in general AI capabilities and provides scalable solutions to complex, high-dimensional, and partially observable gaming environments.
Key finding: This work formulates the concept of General Video Game Playing (GVGP), extending from traditional General Game Playing, by proposing a framework where an AI agent must play unknown video games using only the current state and... Read more
Key finding: This paper reports on the General Video Game AI (GVGAI) competition and framework, highlighting the difficulties in creating AI agents that must perform in real-time with only object-oriented game state information (rather... Read more
Key finding: This paper introduces 'Tribes', a new multi-player, stochastic, partially observable turn-based strategy game framework that includes complex elements such as technology trees, build orders, economy management, and multi-unit... Read more
Key finding: This ongoing research develops generalized, geometry-independent strategic features for use in Monte Carlo Tree Search (MCTS) playout biasing across diverse board games. These features capture spatial patterns that... Read more
2. What methodologies enhance strategic decision training and learning in complex multiplayer and collaborative network environments through serious and educational games?
This research theme explores how video games modeled as serious or educational environments can support training in strategic decision-making, particularly in collaborative networks and historical contexts. It investigates the design, pedagogical foundations, and integration of educational content within game mechanics to create immersive 'soft failure' environments where players can safely explore strategic outcomes. This theme is significant for bridging experiential learning with actionable knowledge transfer in management, history, and military education through interactive simulation.
Key finding: This paper proposes a serious game framework that enables professionals to practice strategic decision-making in a virtual collaborative network environment, emphasizing experiential learning via simulation of... Read more
Key finding: This work constructs a theoretical basis linking video game design and instructional pedagogy, highlighting the intrinsic motivational properties of games and challenges in integrating educational content without diminishing... Read more
Key finding: This study presents 'World War Web', a web-based multiplayer strategy game designed for history education that integrates seamlessly with Learning Management Systems (LMS) through standards-compliant question and assessment... Read more
Key finding: This article analyzes the evolution from manual physical wargaming to complex digital Grand Strategy video games, illustrating how the latter provides expansive historical narratives and detailed political-military contexts... Read more
3. What are the player-centric design considerations and expectations for AI opponents in strategy video games to ensure enjoyable and meaningful gameplay experiences?
This theme focuses on understanding from the player perspective what qualities constitute a worthy AI opponent in strategy video games. It investigates player expectations in terms of challenge, fairness, realism, and engagement, and how AI behavior impacts player enjoyment over long gameplay sessions. Insights here help inform AI design that aligns with human preferences, balancing difficulty, predictability, and believability, rather than purely maximizing computational strength or strategic optimality.
Key finding: Analyzing extensive player forum discussions on Civilization VI and Endless Space 2, this grounded theory study identifies core player expectations for AI opponents: maintaining tension through balanced challenge, fairness to... Read more