Risk analysis of complex and uncertain systems (Q2518395): Difference between revisions

When I read a book about risk analysis I feel sorry for the author. The reason is that the topic is so important, and with such wide approaches and applications, that it is almost impossible to satisfy every reader. Yet the author should be commended for such an interesting book. The new results of this book include many real life applications especially in health management. This is important since in some cases only theoretical studies are given. Moreover the worldwide spread of H1N1 makes these results quite relevant. Also mentioning some limitations to quantitative risk assessment (QRA) is an important point in favor of this book. It is important to notice that diversity is basic to solve problems in most complex systems. Hence using one approach to risk assessment is expected to be less successful than using several different approaches. This is a main point this reviewer wants to convey. The book consists of sixteen chapters. Chapter one discusses an alternative to QRA namely concern-driven risk management. Chapter two introduces probabilistic risk assessment (PRA) in some complex engineered systems. Novel methods and game theoretic ideas are discussed. Chapter three introduces quantitative risk assessment (QRA) for public health risks. The hierarchical effects of decisions of some agents on others is investigated. Chapters 4 and 5 emphasize that sound risk assessment requires sound risk models. By sound it is meant correct causal relations and correct assumptions. The author warns against unsound shortcuts. This reviewer believes that here lies the basic dilemma of QRA. How can one be sure that his/her model is sound!! In complex systems causal relations are not always easy to derive. In fact causality in complex systems are mostly circular i.e. \(A B C \dots A\). A way out is to use different approaches to risk assessment so that (hopefully) they complement each other. In the rest of the book the author attempts to solve this dilemma within the framework of QRA as follows: Chapter 6 uses information theory and data-mining algorithms to detect (possibly nonlinear) causal relations in epidemiological data. Chapter 7 uses the methods of chapter 6 to test causal hypotheses using data, how to discover new relations and how to avoid one's prior beliefs. An application in antibiotic-resistant bacteria infection is presented. Chapters 8 to 12 consider how to use available knowledge and information about causal relations in complex systems to estimate upper bounds on the preventable fraction of disease that could be contained by removing specific exposures. The method is applied to antibiotic-resistant bacteria infection and smoking-related lung cancers. Chapters 10 to 13 use systems dynamics analysis and simulation to predict input-output relations using empirical data and the causal processes being simulated. Some useful techniques include concentrating on the steady state levels of sub-processes which are achieved relatively quickly and which affects other slower processes through time averaged values. Another useful technique is Markov inequality relating deterministic simulations of mean values to bounds on probable values of the underlying stochastic processes. Also a method to reduce some complex systems to smaller ones with similar equilibrium behavior is given. Chapter 14 presents value of information analysis (VOI). A case study for the mad cow disease (BSE) is presented in detail to show the effectiveness of using decision trees, sequential decision optimization and VOI. Chapter 15 studies game theory (competition) and hierarchical optimization models. An important application is in terrorism risk analysis. Chapter 16 surveys methods for predicting the resilience of complex networks against deliberate attacks and methods for improving systems resilience using mathematical optimization and phase transition modeling. A key idea is that scale free networks have large resilience against attacks characterized by eliminating small number of nodes provided that the other nodes have enough surplus routing capacity. But if those nodes have less than a critical amount of surplus capacity the network will breakdown. This is a kind of phase transition from resilient to vulnerable. More applications to the specialists are given throughout the book. The book is written for practitioners but I think it is quite useful for anyone interested in risk analysis and assessment. A slight weakness of the book is the absence of multi-objective optimization (where one deals with more than one objective simultaneously) and the absence of diversity of approaches (except chapter one). Concluding, this is one of the best books on quantitative risk assessment. I strongly recommended it to anyone interested in reducing risk in this highly volatile world. But I also strongly recommend that alternative methods e.g. concern-driven risk management should be used to complement QRA. Diversity is almost always needed to solve problems in complex systems. We have learned that the hard way in pest management (when only chemicals were used) and in the recurrence of diseases due to drug resistance when only a single drug was used. The following author's sentence summarize the situation ``This book seeks to show \dots how QRA can successfully be carried out and used today to IMPROVE risk management in a variety of important real-world applications''. The worldwide spread of H1N1 makes this book quite relevant and timely.