Modeling and Analyzing Systemic Risk in Complex Sociotechnical Systems The Role of Teleology, Feedback, and Emergence

Recent systemic failures such as the BP Deepwater Horizon Oil Spill, Global Financial Crisis, and Northeast Blackout have reminded us, once again, of the fragility of complex sociotechnical systems. Although the failures occurred in very different domains and were triggered by different events, there are, however, certain common underlying mechanisms of abnormalities driving these systemic failures. Understanding these mechanisms is essential to avoid such disasters in the future. Moreover, these disasters happened in sociotechnical systems, where both social and technical elements can interact with each other and with the environment. The nonlinear interactions among these components can lead to an “emergent” behavior – i.e., the behavior of the whole is more than the sum of its parts – that can be difficult to anticipate and control. Abnormalities can propagate through the systems to cause systemic failures. To ensure the safe operation and production of such complex systems, we need to understand and model the associated systemic risk. Traditional emphasis of chemical engineering risk modeling is on the technical components of a chemical plant, such as equipment and processes. However, a chemical plant is more than a set of equipment and processes, with the human elements playing a critical role in decision-making. Industrial statistics show that about 70% of the accidents are caused by human errors. So, new modeling techniques that go beyond the classical equipment/process-oriented approaches to include the human elements (i.e., the “socio” part of the sociotechnical systems) are needed for analyzing systemic risk of complex sociotechnical systems. This thesis presents such an approach. This thesis presents a new knowledge modeling paradigm for systemic risk analysis that goes beyond chemical plants by unifying different perspectives. First, we develop a unifying teleological, control theoretic framework to model decision-making knowledge in a complex system. The framework allows us to identify systematically the common failure mechanisms behind systemic failures in different domains. We show how cause-and-effect knowledge can be incorporated into this framework by using signed directed graphs. We also develop an ontology-driven knowledge modeling component and show how this can support decision-making by using a case study in public health emergency. This is the first such attempt to develop an ontology for public health documents. Lastly, from a control-theoretic perspective, we address the question, “how do simple individual components of a system interact to produce a system behavior that cannot be explained by the behavior of just the individual components alone?” Through this effort, we attempt to bridge the knowledge gap between control theory and complexity science