CREATE Research Archive Research Project Summaries 2010 Modeling Terrorist Beliefs and Motivations Richard John University of Southern California, richardj@usc.edu Heather Rosoff University of Southern California, rosoff@usc.edu Follow this and additional works at: http://research.create.usc.edu/project_summaries Recommended Citation John, Richard and Rosoff, Heather, "Modeling Terrorist Beliefs and Motivations" (2010). Research Project Summaries. Paper 15. http://research.create.usc.edu/project_summaries/15 This Article is brought to you for free and open access by CREATE Research Archive. It has been accepted for inclusion in Research Project Summaries by an authorized administrator of CREATE Research Archive. For more information, please contact gribben@usc.edu.
Modeling Terrorist Beliefs and Motivations Richard John and Heather Rosoff, University of Southern California richardj@usc.edu, rosoff@usc.edu 1. Modeling Terrorist Beliefs and Motivations Overview... 1 2. Research Accomplishments... 2 3. Research Products... 6 3.1. Presentations Conferences... 6 4. Education and Outreach Products 6 1. Modeling Terrorist Beliefs and Motivations Overview The objective of this research is to further develop our pioneering approach to adversary threat assessment through the construction of random utility models of terrorist preferences. This work builds on previous research efforts (Rosoff, 2009; Rosoff & John, 2009) that have used decision analysis models and elicitation methods to: (1) construct of a value tree for a terrorist leader or organization using value focused thinking (VFT), (2) construct a random multi-attribute utility model (RMAUM) capturing trade-offs among conflicting objectives and single-attribute utility functions representing risk attitudes of terrorist leaders, (3) construct probability distributions capturing key uncertainties for terrorist leaders (e.g., attack success) and uncertainties in the utility function parameters provided by adversary experts. Understanding the objectives and motivations that drive terrorist group behavior is critical. Current methods for terrorism risk assessment focus on target vulnerability, terrorist capability and resources, and attack consequence. What many researchers have yet to consider is the influence of terrorist group values and beliefs in deciphering the root cause of their militant behavior. This understanding has the potential to contribute to probabilistic estimates of terrorist threats. During Year 6 we further developed the adversary preference modeling (APM) methodology by (1) evaluating how much the proxy terrorist utility model diverges from using a utility model for the US coupled with a zero-sum assumption, (2) assessing how anti-terrorism resource allocations impact terrorist preferences for different attack strategies and estimated relative likelihoods of various attack strategies, and (3) validating the APM through a case study with political/social/advocacy groups with an action oriented agenda that is driven by specific motives, values, and objectives. To address the first question, we developed (dis-)utility models for evaluating the same attack strategies from the perspective of the US, and compared the resulting risk profiles to those of a terrorist leader. To address the second question, we performed sensitivity analysis over parameters in the terrorist utility model that should be sensitive to different allocations of antiterrorist resources. Three anti-terrorist resources were considered (1) tagging of original radiological and biological sources terms, (2) hardening targets for better protection from attack, and (3) installing screening devices to improve detection of radiological and explosive materials. Each anti-terrorist resource was evaluated in terms of its impact upon proxy terrorist s attribute definitions, probability estimates and ultimately, estimates for the likelihood of various attack strategies. Lastly to address the third question, we opted to test the validity of our APM This research was supported by the United States Department of Homeland Security through the National Center for Risk and Economic Analysis of Terrorism Events (CREATE) under award number 2010-ST-061-RE0001. However, any opinions, findings, and conclusions or recommendations in this document are those of the authors and do not necessarily reflect views of the United States Department of Homeland Security, or the University of Southern California, or CREATE.
methodology through the development of a model for a domestic political action group. Research progress and accomplishments made throughout Year 6 are described in the subsequent section. 2. Research Accomplishments A significant portion of Year 6 work was a continuation of Year 5 work related to the first two research questions described above. With respect to our evaluation of the zero-sum assumption, we demonstrate that quite different preferences and implied relative likelihoods result from our proxy decision analysis of terrorist preferences compared to a more standard economic evaluation of expected losses from the perspective of the US. It is clear from our analysis that terrorist organizations (and their leaders) have very different fundamental objectives that are not captured by a zero-sum assumption applied to an evaluation model constructed from the US perspective. Furthermore, even in cases where there terrorist objects are similar to the mirror (opposite) objectives of the US, there are remarkable differences in attribute scales, attribute weights, and attribute utility functions capturing attitudes toward risk. These differences in values contribute to large differences in the evaluation of alternative attack strategies based on the proxy terrorist decision analytic model and the model assuming US values and a zero-sum game. With respect to our sensitivity analysis of anti-terrorism initiatives, we demonstrate that preferences and relative likelihoods are altered depending upon the anti-terrorism resource under consideration and the proxy terrorist s perceived risk posed by that resource. For example, as seen in Figures 1 and 2, the analysis shows that the tagging countermeasure, did not impact Proxy 1 s attack preferences. However, Proxy 2 s likelihood estimates decreased for those alternatives directly impacted by the anti-terrorism resource (Port nuclear device, dirty bomb, smallpox attack and anthrax release). In doing so, the desirability of Proxy 2 s most preferred attack, the no attack alternative, increased. Probability of Attack 0.92 0.92 With Tag Countermeasure No Countermeasure 0.06 0.06 0.03 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Figure 1: Proxy 1 Relative Likelihood for Tagging Countermeasure Page 2 of 6
With Tag Countermeasure No Countermeasure Probability of Attack 0.59 0.49 0.14 0.14 0.11 0.10 0.10 0.07 0.02 0.07 0.04 0.05 0.010.03 0.000.01 0.00 0.01 Figure 2: Proxy 2 Relative Likelihood for Tagging Countermeasure Interestingly, while anti-terrorism resources alter attribute scales and uncertainties and attack probability estimates, attack strategies preferences could remain unaffected depending on the values assigned to attribute weights and attribute utility functions capturing attitudes toward risk for each alternative. Through the exploration of different anti-terrorism resources impact on proxy terrorist preferences, insight is gained as to the extent to which government actions might influence terrorist decision making. With respect to APM model validation, we took a two-phased approach; the first of which was completed in Year 6. During Year 6 we built a multi-attribute utility model based on PETA s presentation of itself and what could be found online from second-hand sources. Phase 2 is expected to be conducted throughout Year 7. Phase 2 is anticipated to involve interviewing a PETA stakeholder in order to demonstrate the accuracy and validity of the originally constructed second-hand model, and hence validate the APM modeling approach. A general search of online resources was conducted, leading to a first set of objectives and alternatives. Second-hand sources attempting to describe what PETA is and does agreed with PETA s self-publicized accounts on these objectives, although also including scandals and financial donations to activist groups that are willing to commit property damage that PETA understandably did not publicize. PETA s overall goal is improving animals rights. This is characterized by the four core values expressed in their motto - Animals are not ours to eat, wear, experiment on, or use for entertainment, as well as an emphasis on pet spay and neutering, to reduce the numbers of unwanted animals. The four motto objectives are folded into the Minimize Cruelty to Animals sub-objective to avoid model over-specification. PETA must also consider growing its organizational power. Maximize Organizational Power considers income, members and minimizing cost, as well as maximizing new volunteers and influence. Lastly, Perception of Group refers to the group s delicate balancing act of radical positioning and working with the system, and outreach. The twin perception objectives are an attempt to model the tension of Page 3 of 6
PETA s choice of positioning in the political spectrum; they want to bring the discussion to where they are, and have staked out a more radical position than the culture holds regarding treatment of animals. However, they want to be taken seriously; if they are perceived as too far out of the spectrum, then achieving successes inside the system via the traditional methods becomes impossible. The two metrics selected to characterize Perception of Group, maximize media hits and minimize negative press, cover the idea of getting attention but only the right kind of attention. Figure 3 is a graphic depiction of PETA s objectives hierarchy. Improve Animal Rights Maximize Organizational Power Max Incoming Money Max New Members Max New Volunteers Maximize Influence Network Minimize Costs Minimize Cruelty to Animals Max Changes to Industry Practices Max Educational Materials Max New Laws Perception of Group Max Media hits Minimize Negative Press Figure 3. PETA Objectives Hierarchy Alternatives as well as metrics and scales were identified for each of the attributes. In addition, estimates of attitudes toward risk, trade-offs among various objectives, and uncertainties about Page 4 of 6
the success of particular alternatives are included. The following are the ten attribute scales, ordered by the median values of the normalized weights assessed (Figure 4). Weight Max New Members 17.9 Max New Laws 16.1 Max Changes to Industry Practices 14.3 Max Media hits 12.5 Max New Volunteers 10.7 Max Incoming Money 8.9 Max Educational Materials Distributed 8.0 Maximize Influence Network 5.4 Minimize Costs 4.5 Minimize Negative Press 1.8 Figure 4. Weights for Adversary Value Expert Preference Set The aforementioned assessments were used to construct a random multi-attribute utility model (RMAUM). The RMAUM was estimated using a simulation model implemented within Excel and @Risk 10,000 iterations. As seen in Figure 5, we found that the alternative with the highest mean utility estimate is Undercover Investigation (0.62), followed by New Campaign (0.49), New Ad and Publicity Study (both with a utility of 0.44) and Purchase Stock (0.36). Figure 5. CDF Utility Distributions Page 5 of 6
The risk profiles were then used to obtain attack probabilities over a fixed time period, given an assumption that one of the nine options (including no attack) will be selected. As seen in Table 1, 91% of the time Undercover Investigation was the most preferred alternative, and the only other alternative that achieved a meaningful preference ranking was a New Campaign, at 7%. Alternatives P(max j (u(x j )) = x i New Ad 0.01 New Campaign 0.07 Publicity Stunt 0.01 Purchase Stock 0.00 Undercover Investigation 0.91 Table 1. Probability of Alternative The second phase of this project and a submitted paper for publication on the overall project is planned for Year 7. 3. Research Products 3.1. Presentations Conferences 1. John, R. S. & Rosoff, H. Decision analysis by proxy for the rational terrorist, Poster presented at 30th annual meeting of the Society of Judgment and Decision Making, Boston, MA., November 20-23, 2009. 2. John, R. S. & Rosoff, H. Value focused thinking for evaluating alternative counter-terrorism policies. Homeland Security and Counterinsurgency invited research presentation at the annual meeting for the Institute for Operations Research and Management Science (INFORMS), San Diego, California, October 11-14, 2009. 4. Education and Outreach Products Education and Outreach Initiatives (Please detail below) # # of Postdoctoral Researchers 1 # of students supported (funded by CREATE) 1 Funded by CREATE Heather Rosoff, Postdoctoral Researcher. University of Southern California (USC) School of Policy, Planning and Development. Kacie Shelton, Master of Systems Architecting & Engineering Candidate, Expected August 2011. USC s Viterbi School of Engineering. Page 6 of 6