For more information on this publication, visit /pubs/rgs_dissertations/RGSD408.html Published 2018 by the RAND Corporation, Santa Monica, Calif.R is a registered trademark Limited Print and Electronic Distribution RightsThis document and trademark(s) contained herein are protected by law. This representation of RANDintellectual property is provided for noncommercial use only. Unauthorized posting of this publication online isprohibited. Permission is given to duplicate this document for personal use only, as long as it is unaltered andcomplete. Permission is required from RAND to reproduce, or reuse in another form, any of its researchdocuments for commercial use. For information on reprint and linking permissions, please visit /pubs/permissions.html.The RAND Corporation is a research organization that develops solutions to public policy challenges to helpmake communities throughout the world safer and more secure, healthier and more prosperous. RAND isnonprofit, nonpartisan, and committed to the public interest. RAND’s publications do not necessarily reflect the opinions of its research clients and sponsors.Support RANDMake a tax-deductible charitable contribution at /giving/contribute iiiAbstractThis dissertation considers the viability of applying Value of Information (VoI) methods incomplex systems for policy analysis, concluding that these methods can be applied, but thatdifferent methods are appropriate in different cases. VoI is value to a decisionmaker of thedifference information makes for a decision, or what a decisionmaker should pay forinformation. Compared to present practice, specific value of information techniques canpotentially improve policy analysis and decision making. To that end, a methodology for policyapplications of VoI is presented that can simplify practical application.The dissertation applied computational methods for VoI to two case studies in disparatedomains: biosurveillance and detection of violent extremists. Each of the case studies outlineshow Bayesian methods can be used to evaluate VoI, including a novel formulation forconsidering additional information sources in the systems. The case studies are used both toconsider the relative benefits and drawbacks of these approaches in policy decision making, andto identify potential challenges. Approaches for identifying and avoiding these challenges andfor selecting appropriate methods are integrated into the overall methodology, which is presentedin the final chapter. ivTable of ContentsAbstract ...... iiiTable of Contents ... ivFigures........ ixEquations..... xTables ......... xiSummary ... xiiAcknowledgments xivAbbreviations ....... xvi1 Introduction ....... 11.1 Importance of VoI for Policy . 11.2 What is Value of Information (VoI) . 2 Basic Formulations of VoI ...... 2 1.2.1Decision to Treat ......... 3 1.2.2Decision Trees . 4 1.2.3Perfect Information ..... 4 1.2.4Imperfect Information . 5 1.2.5 1.3 Considerations on Computing VoI and Related Quantities ... 7 Extensions of VoI for Continuous Information ........... 7 1.3.1Real Options .... 8 1.3.2Algebraic Tools for Parametric Uncertainties . 8 1.3.3Related Methods and Applications ... 12 1.3.4 1.4 Applying VoI and Related Methods in Policy Analyses ..... 131.5 Research Approach .. 14 Methodological Research Questions . 15 1.5.1Dissertation Structure and Outline .... 15 1.5.2Heterogeneous Information Fusion ... 15 1.5.3Biosurveillance .......... 17 1.5.4 2 An Overview of Policy Decisions for VoI Questions . 192.1 Taxonomy for Decision making ...... 19 Conceptual Model for Decisions ....... 19 2.1.1Refining the Conceptual Model ........ 20 2.1.2 2.2 The Role of Models in Decision Processes .. 22 Problem Formulation23 2.2.1Criteria and Metric Selection24 2.2.2Alternative Generation .......... 26 2.2.3Screening ....... 27 2.2.4Assessment .... 28 2.2.5v Concluding a Decision Process ......... 28 2.2.6Necessity of Utility Theory in VoI Decisions ........... 29 2.2.7Implicit Utilities and Necessity ......... 30 2.2.8 2.3 Models and Decision Problems in General .. 31 Example: Flood Control Systems ...... 33 2.3.1Iteration and Model Selection Problems ....... 34 2.3.2Model Selection for Information-Sensitive Decisions .......... 34 2.3.3 2.4 Requirements for VoI Evaluation .... 35 Information Models ... 35 2.4.1 2.5 Computational Model Characteristics .......... 37 Causal Models ........... 38 2.5.1Generative Models .... 39 2.5.2Representational Clarity ........ 40 2.5.3 2.6 Tradeoffs in Model Selection ........... 432.7 Model Types44 Decision Trees and Related Methods44 2.7.1Decision Tree Related Methods ........ 48 2.7.2Iterated (Markov) Processes, Simulation-based Models, and Machine Learning ..... 48 2.7.3Discrete Bayesian Networks . 49 2.7.4Probabilistic Programming and Bayesian Statistical Models53 2.7.5 2.8 Stage Setting with an Illustrative Case of Terrorist Threat Decisions ......... 56 Illustrative Decision Process . 57 2.8.1Modeling Alternatives ........... 58 2.8.2Extending the Model . 61 2.8.3 2.9 Model Review Conclusion ... 623 Threat Fusion Case Study ....... 643.1 Background .. 64 Current Intelligence Process .. 65 3.1.1Systems in Use for Extremist Threat Evaluation ...... 65 3.1.2Intelligence Sources .. 67 3.1.3Decision process and model .. 70 3.1.4 3.2 Threat Fusion Model Components ... 70 Model Components ... 70 3.2.1Propensity for Terrorism (PFT) Model ......... 71 3.2.2Heterogeneous Information Fusion Model in General .......... 74 3.2.3Threat Fusion Data Source Model .... 77 3.2.4Utility Model . 94 3.2.5Overall Model Considerations .......... 98 3.2.6Model Requirements Evaluation ....... 99 3.2.7 3.3 Example Cases and Results .. 99 Domestic Gang Violence .... 100 3.3.1Radicalization .......... 105 3.3.2 。。。。。。