Putting the "Human" into "Human Behavior Models"

Intelligent Agents are often developed for the purposes of mimicking the behavior of a person performing a task. These “Human Behavior Models” (HBMs) (Pew and Mavor 1998) are often used to perform complex tasks such as flying airplanes in simulation, interacting as non-playercharacters in games, participating in teammates in simulation-based training environments, etc. These models’ level of “human-ness” can vary greatly across models and uses, at multiple levels. For the early work in this area, interest was primarily in the purely “rational” problem-solving aspects of human behavior: how does a human perform a task under normal conditions? These models are often guided by theory of human information processing, and typically included cognitive features such as memory, perception, and learning. But they also leave out many aspects of what we typically think of as “being human.”