According to expert practitioners and researchers in the field of human behavior modeling ([Silverman et al., 2002; Pew and Mavor, 1998; Ritter et al., 2003]), a common central challenge now confronting designers of HBM (human-behavior-modeling) applications is to increase the realism of the synthetic agents' behavior and coping abilities. It is well accepted in the HBM (human-behavior-modeling) community that cognitively detailed, "thick" models are required to provide realism. These models require that synthetic agents be endowed with cognition and personality, physiology, and emotive components. (We will hereafter refer to these rich models as "cognitively detailed models" or "thick agents.") To make these models work, one must find ways to integrate scientific know-how from many disciplines, and to integrate concepts and insights from hitherto fragmented and partial models from the social sciences, particularly from psychology, cultural studies, and political science. One consequence of this kind of integration of multiple and heterogeneous concepts and models is that we frequently end up with a large feature space of parameters that then need to be filled in with data.