We address the challenges of evaluating the fidelity of AI agents that are attempting to produce human-like behaviors in games. To create a believable and engaging game play experience, designers must ensure that their non-player characters (NPCs) behave in a human-like manner. Today, with the wide popularity of massively-multi-player online games, this goal may seem less important. However, if we can reliably produce human-like NPCs, this can open up an entirely new genre of game play. In this paper, we focus on emulating human behaviors in strategic game settings, and focus on a Social Ultimatum Game as the testbed for developing and evaluating a set of metrics for comparing various autonomous agents to human behavior collected from live experiments.

Interactive narrative environments offer significant potential for creating engaging narrative experiences that are tailored to individual users. Increasingly, applications in education, training, and entertainment are leveraging narrative to create rich interactive experiences in virtual storyworlds. A key challenge posed by these environments is devising accurate models of director agents' strategies that determine the most appropriate director action to perform for crafting customized story experiences. A promising approach is developing an empirically informed model of director agents' decision-making strategies. In this paper, we propose a framework for learning models of director agent decision-making strategies by observing human-human interactions in an interactive narrative-centered learning environment. The results are encouraging and suggest that creating empirically driven models of director agent decision-making is a promising approach to interactive narrative.

We describe a heuristic search technique for multi-agent pursuit-evasion games in partially observable Euclidean space where a team of trackers attempt to minimize their uncertainty about an evasive target. Agents' movement and observation capabilities are restricted by polygonal obstacles, while each agent's knowledge of the other agents is limited to direct observation or periodic updates from team members. Our polynomial-time algorithm is able to generate strategies for games in continuous two-dimensional Euclidean space, an improvement over past algorithms that were only applicable to simple gridworld domains. We demonstrate that our algorithm is tolerant of interruptions in communication between agents, continuing to generate good strategies despite long periods of time where agents are unable to communicate directly. Experiments also show that our technique generates effective strategies quickly, with decision times of less than a second for reasonably sized domains with six or more agents.

This paper looks at the challenges and solutions for intelligent mobile agents existing in virtual environments. Representing and using a map of these very large-scale, dynamic environments is a key challenge in providing an autonomous agent for large, online worlds. We look at a method for improving the generation of probabilistic roadmaps by observing and using the movements of other avatars in an environment. We also look at methods for segmenting and scaling the map in very large environments. Finally we extend this to look at generating and updating maps in dynamic environments.

The problem of optimal planning under uncertainty in collaborative multi-agent domains is known to be deeply intractable but still demands a solution. This thesis will explore principled approximation methods that yield tractable approaches to planning for AI assistants, which allow them to understand the intentions of humans and help them achieve their goals. AI assistants are ubiquitous in video games, mak- ing them attractive domains for applying these planning techniques. However, games are also challenging domains, typically having very large state spaces and long planning horizons. The approaches in this thesis will leverage recent advances in Monte-Carlo search, approximation of stochastic dynamics by deterministic dynamics, and hierarchical action representation, to handle domains that are too complex for existing state of the art planners. These planning techniques will be demonstrated across a range of video game domains.

This paper proposes a method for the development of AI for autonomous agents in game worlds modeled by fixed regular grids. This approach uses context behaviors driven by distance maps to support multiple agents in a dynamic environment. This paper introduces the notion of hierarchical distance maps which allow for higher-level goals to be easily specified by designers. We also discuss potential applications of our approach in the design and development of agents and behaviors in the block world genre.

This article describes a technical approach to movement-based interactions between a human interactor and an intelligent agent based on the theatrical Viewpoints movement framework.

We present CLASSQL, a multi-agent model for playing real-time strategy games, where learning and control of our own team's units is decentralized; each agent uses its own reinforcement learning process to learn and control units of the same class. Coordination between these agents occurs as a result of a common reward function shared by all agents and synergistic relations in a carefully crafted state and action model for each class.

In this paper we describe a method of modeling play styles as deviations from approximations of game theoretically rational actions. These deviations are interpreted as containing information about player skill and player decision making style. We hypothesize that this information is useful for differentiating between players and for understanding why human player behavior is attributed intentionality which we argue is a prerequisite for believability. To investigate these hypotheses we describe an experiment comparing 400 games in the Mario AI Benchmark testbed, played by humans, with equivalent games played by an approximately game theoretically rationally playing AI agent. The player actions' deviations from the rational agent's actions are subjected to feature extraction, and the resulting features are used to cluster play sessions into expressions of different play styles. We discuss how these styles differ, and how believable agent behavior might be approached by using these styles as an outset for a planning agent. Finally, we discuss the implications of making assumptions about rational game play and the problematic aspects of inferring player intentions from behavior.

This article describes a technical approach to movement-based interactions between a human interactor and an intelligent agent based on the theatrical Viewpoints movement framework. The installation creates a liminal virtual / real space for the human and AI to interact by the use of digital projection for the AI visualization and shadow play to represent the human. Observations from a recent public demonstration of the system and future directions of work are also discussed.