Agent based human behavior modeling: A knowledge engineering based systems methodology for integrating social science frameworks for modeling agents with cognition, personality and culture

The objective of my dissertation is to develop an integrative process for constructing, validating and refining agent-based models of human behavior, and hence, carrying out human behavior modeling (HBM) in a defensible manner. For lack of a better term and due to extensive involvement of Knowledge Engineering (KE) techniques and construction of the knowledge models, the process will be conveniently referred to as the KE-based process (KE process).; The hypothesis is that the KE process will help: (1) systematically transform empirical evidence, tacit knowledge, and expert knowledge into data for modeling; (2) reduce human errors and cognitive biases (e.g. confirmation bias); and (3) verify and validate the model as a whole, and the knowledge base in particular.; In order to test the hypothesis, I have employed the KE process to create several models of leaders, followers, crowd members, rebels, and agitators in conflict situations. Next, I have been able to validate my KE process under naturalistic conditions by testing, verifying and validating these models against real world datasets. During the course of the studies, I have integrated existing, but fragmented, frameworks in social science to bring about the behavior of people in social systems. I have demonstrated the process through three case studies. In the validation and discussion sections, I explore how and where correlations between simulated agents versus real characters are improved through the application of the KE process. The conclusions address lessons learned, tools created to help others use the KE process, and steps for future research.; The dissertation has demonstrated a process for HBM that could be employed to construct cognitively detailed synthetic agents to be used in such applications as immersive training, analytical support and domain knowledge acquisition. Specifically, the process helps integrate the social science models, structure the value trees and estimate the parameters of the agents, so that these agents make decisions and exhibit behaviors that are realistic.