Simulating Humans

Simulations designed for training or experimentation have (at least) two elements that are critical to their success. The majority of simulation research focuses on content validity -assuring that the simulation is accurate in terms of the domain. For example, ensuring that simulated productivity in a given situation approximates the productivity that would be observed in the real world should the corresponding context arise. However, for organizational simulation or any other system in which human decision making is critical, content validity is not sufficient to provide an adequate simulation experience. For example, in military or emergency training the emotional content is just as important as situational accuracy. A principal goal of such exercises is to give experience to decision makers in evaluating options and making choices in the face of a variety of stressful or emotional factors. To achieve this emotional state requires interaction plausibility with sufficient believability. That is, the interfaces with the systems and the people must reflect the tenor of what the interactions would be in the real world. For organizational simulation this task is particularly challenging because organizational structure implies particular interpersonal interactions, and these are the interactions that must rendered with sufficient fidelity. Specifically, the system must be able to render the output of simulated human characters in a manner that can provoke the appropriate interpersonal and/or emotional response of the user. In this chapter, we describe a variety of recent advancements in computer graphics and animation technologies that are directly relevant to this daunting task. Our exposition will cover all aspects of human modeling and animation. We will discuss prevalent techniques for simulating human behaviors as it supports organizational simulation and describe approaches and tools that are appropriate to do such forms of simulation. We will specifically focus on the rendering of "talking heads" -upper torso simulations of people speakmg arbitrary text and also containing a variety of emotional content. Recently there have been numerous developments in image-and video-based rendering that produce photorealistic video sequences of talking characters. These methods, combined with recent advances in computer vision analysis of videos of people, make plausible the prospect of highly effective output for organization simulation.