NSF IED Workshop Report

WORKSHOP REPORT ON DETECTING AND COUNTERING IEDS AND RELATED THREATS Principal Investigator: Dr. Judee Burgoon, Center for the Management of Information, University of Arizona Co-Principal Investigator: Dr. Vasundara V. Varadan, Microwave & Optics Laboratory for Imaging & Characterization, University of Arkansas Significance of IEDs and Related Threats Improvised explosive devices (IEDs) represent one of the most significant and deadly threats faced by armed service personnel and innocent civilians in current conflicts. Related threats refer to terrorists and suicide bombers who have caused large numbers of civilian deaths in non-war settings such as 9/11 and the commuter train bombings in the United Kingdom, Spain and more recently in India. The two threats are somewhat related but the perpetrators as well as the victims and the locations are quite different and require different approaches to detect and counter. In a war theater such as Iraq or Afghanistan, the threat posed by IEDs is more insidious and in uncontrolled and widely extended locations. Threats at airports, borders, train and train stations as well as civil structures present a more controlled ingress/egress scenario but with untrained and innocent victims who vastly outnumber law enforcement personnel unlike the war theater. Use of IEDs has resulted in nearly half of the hostile fire deaths in Iraq and Afghanistan, and their use is only becoming more frequent (Barry, Hastings, & Thomas, 2006). IEDs as an unconventional warfare tactic are not new; they have been in use for decades. Since their inception, the United States and other governments have invested significant resources to mitigate this threat. Past efforts have centered on developing technology or tools to detect or counter IEDs, but it is increasingly clear that the IED threat defies a completely technical solution1. Attention must be paid not just to the bombs but also to the bombers. The mounting death toll from 9/11 and from train and embassy bombings around the world (approaching 5000) underscores the severity of the threat and diversity of perpetrators, locales and forms of improvised devices. Homegrown terrorists who were recently arrested in the UK and possible cells in the US are different from the suicide bombers who are recruited in the Middle East. The production, distribution, and detonation of IEDs involve the coming together of engineered materials with human beings at particular times and places. To detect and prevent such acts, it is necessary to understand all parts of this configuration. Research findings from psychologists, cultural anthropologists, communication scholars, sociologists, political scientists, and geographers are as critical to successful detection, intervention, and prevention as understanding engineering technicalities and logistics. For example, social psychological and communication research on the detection of deception can be used to train security agents in civilian and military This report was supported by a National Science Foundation Grant 0631462. The views expressed are those of the principal investigators and workshop participants and not an official expression of the National Science Foundation. Claims made in this report are based on the workshop participants’ briefings or addresses unless otherwise cited. 1 contexts to spot suspicious conduct that may be a prelude to an IED incident. Systems for automating detection of suspicious and deceptive activity are likewise drawing upon research in criminal justice, communication, neuroscience, and social psychology to identify and automate detection of behavioral signatures. There are many other areas of current social and behavioral science research that can be just as valuable. . The NSF Workshop Objectives Believing that research on counter-IED efforts will advance more rapidly and successfully by joining technical and behavioral sciences, the Directorate of Social, Behavioral, and Economic Sciences, in conjunction with the Directorate of Engineering, supported a workshop to bring together scientists across the social, behavioral, information, and engineering spectra whose work is relevant to the challenges posed by IEDs and related threats. Participants were invited to share their substantive knowledge related to countering IEDs and to propose leading edge research that combines technical and social/behavioral perspectives. Although research on the root causes of hostile intentions among terrorists and insurgents is clearly germane, it was deemed as outside the scope of a two-day workshop. Instead, the workshop focused on more temporally proximal components in IED incidents, including recruitment of participants, resource acquisition, planning, assembly, implantation, and detonation of such devices. One central objective was to propose an agenda for future short-term and long-term research leading to better prediction, detection, and prevention of IED use. A second objective was to promote innovative collaborations across NSF directorates in anticipation of NSF-wide Broad Agency Announcements related to security threats. This report presents the resultant recommended research priorities and the process by which they were derived. It should be noted that several of the research themes overlap with one another, recognition that the researchable issues are interrelated and that progress on one front may have significant bearing on progress on another. A third objective was to address methodological considerations in conducting these lines of research. Participants were asked to identify what sorts of methods they and their discipline would propose applying to this problem and where in the behavioral process of transforming hostile intentions into detectable actions research can focus most profitably in the near term. For example, will there be more gained from centering on discovering behavioral patterns such as random walks in public places or in uncovering how communication networks are used to acquire material and to plan attacks? Can laboratory research simulate key aspects of the behavior leading up to IED placement and detonation or must field research be used? What engineering technology solutions can be proposed for detecting and countering threats at airports and similar civil facilities and what technologies, for a scenario such Iraq or Afghanistan? Rather than produce either/or answers, the purpose of this part of the exercise was to force a cost/benefit analysis of various lines of inquiry and focus on where the largest pay-offs are likely. A related methodological issue was whether and how to fuse technical sensor data with behavioral data to produce the best predictive models. 2 Participants Workshop participants included 23 invited experts from the social science disciplines of social psychology, sociology, communication, political science, and geography and from the engineering and information disciplines of information systems, computer science, computer engineering, electrical and mechanical engineering, and applied physics. Following are the participants and their institutional and disciplinary affiliations. It is their expertise, views and recommendations that are represented in this report: Maruthi Akella, Assistant Professor of Aerospace Engineering and Engineering Mechanics, University of Texas, Austin Charles F. Bond, Jr., Professor of Psychology, Texas Christian University Donald Brown, Professor and Chair of the Department of Systems and Information Engineering, University of Virginia Judee K. Burgoon, Professor of Communication, Family Studies and Human Development; Director, Human Communication Research, Center for the Management of Information; Associate Director, Media Interface Network Design Labs, University of Arizona Kathleen M. Carley, Professor of Computer Science at the Institute for Software Research International and Director of the Center for Computational Analysis of Social and Organizational Systems (CASOS), Carnegie Mellon University Jeffrey Cohn, Professor of Psychology, University of Pittsburgh/Carnegie Mellon University William Crano, Professor of Behavioral and Organizational Sciences, Claremont Graduate University Andrew Dollins, Research Scientist, Department of Defense Polygraph Institute William Donohue, Professor of Communication, Michigan State University Lawrence W. Hunter, Researcher, Applied Physics Laboratory, Johns Hopkins University Mathew Jensen, doctoral candidate in Management Information Systems, University of Arizona Arie Kruglanski, Distinguished University Professor of Psychology, University of Maryland Harry Martz, Director, Center for Nondestructive Characterization, Lawrence Livermore National Laboratory, University of California, Berkeley David L. Masters, Director N-STAR, Office of Naval Research Dale Murray, Research Consultant, Sandia National Laboratory Jay F. Nunamaker, Jr., Regents and Sodwedel Professor of Management Information Systems, Computer Science, and Communication, University of Arizona H. Dan O’Hair, Presidential Professor of Communication and Director, Institute for Communication Research, University of Oklahoma; President, National Communication Association Marc Sageman, Sageman Consulting, Washington, DC Gary Strangman, Director, Neural Systems Group, Massachusetts General Hospital, Harvard University Vasundara Varadan, Billingsley Chair and Distinguished Professor, Microwave and Optics Laboratory for Imaging & Characterization and Department of Electrical Engineering, University of Arkansas Michael Vlahos, Senior Staff, Applied Physics Laboratory, Johns Hopkins University Joseph B. Walther, Professor of Communication and Telecommunication, Information Studies & Media, Michigan State University 3 May Yuan, Professor of Geography; Director, Center for Spatial Analysis, University of Oklahoma A number of additional scientists and government representatives also attended the workshop. The complete list of all participants and attendees appears in Appendix A. Workshop Methodology The two-day workshop consisted of sessions devoted to framing the problem, substantive overviews of the broad range of perspectives and methodologies that are possible, and generation of several proposed research strands. All participants gave a formal brief covering their content, methodological expertise and/or the expertise represented by their discipline. Groups of related briefings were followed by discussions and recording of research ideas. A complete agenda appears in Appendix B. To maximize group brainstorming and input, an intranet was established using NSF laptop computers and a server furnished by the Center for the Management of Information (CMI). GroupSystems (computer-assisted groupware) was used to conduct electronic brainstorming, to collect and organize research themes, and to produce recommendations. Facilitators were Drs. Jay Nunamaker, Director of CMI, and Jim Lee, Associate Director of CMI. A number of different tools were used within the GroupSystems software. The Topic Commenter Tool was used to allow participants to record comments and questions during each presentation. After each set of related presentations, all workshop participants used the Categorizer tool to brainstorm research topics relevant to the preceding presentations and to place the research topics into general meaningful categories. The Group Outliner tool was used to record the research themes produced from subgroup discussions that took place periodically throughout the workshop. Finally, rankings of the research ideas were solicited and recorded using the Vote tool. All contributions and balloting within the GroupSystems software were anonymous so as to remove any inhibitions to contribute and encourage open and honest collaboration. Framing the IED Problem The magnitude of the IED problem is such that identifying researchable issues can be approached from innumerable vantage points. The workshop framed the problem according to the life cycle of IED incidents. Background briefings addressed the nature of the bombs themselves, the bombers and their socio-political milieu, the stream of activities from planning through detonation of an IED that constitute the life cycle of an IED incident, and common scenarios to which proposed research could be applicable. IED Devices and Their Detection An IED is a “homemade” bomb created to cause death or injury using commercial, military or homemade explosives that are hidden and set off using a variety of trigger mechanisms. IEDs are not new—they have been used for six decades by terrorists, insurgents, and others as a component of unconventional warfare. However, their frequency of occurrence and their visibility as a preferred form of weaponry among terrorists and insurgents has heightened attention to this very complex—and some would argue intractable—problem that they represent. IEDs as part of the terrorist’s arsenal of tactics are a form of “propaganda by deed,” a spectacle intended to get a lot of people watching and to inflict psychological repercussions beyond the 4 immediate victims (see Gearson, 2002). IEDs are increasingly being used in Iraq and Afghanistan and have caused nearly half of the hostile fire deaths in Iraq. The military spent $3.3 billion to defeat IEDS in 2005 alone—mostly on more armor and technology. IEDs such as aircraft piloted by suicide bombers and improvised explosives carried into trains and buildings have resulted in the death of several thousand innocent civilians. The U.S. Military has tried detecting IEDs with UAV-sensors, jamming radio frequency detonators and other technical solutions. Increased security measures at airports and increased vigilance in general may have foiled potential incidents in the US since 9/11. However, these technical solutions have not been very successful (the shoe bomber case) and most IEDs are detected as a result of experience, training and intelligence (Barry, Hastings, & Thomas, 2006). The primary components in an IED are the charges and the initiation systems. The charges for an IED are readily available from past armed conflicts. Thousands of tons of munitions and other explosives have been stockpiled, and common citizens in Iraq, Afghanistan, and other countries currently have access to these munitions and the explosives they contain. Radioactive materials for producing dirty bombs are available in most hospitals. In Iraq, the charge for a typical IED is an artillery shell with an ignition mechanism in the compartment containing the explosives. More recent arrests in the UK have exposed the potential threat from seemingly harmless liquid components that can be mixed to make explosives on board a plane. Each type of threat requires a different technology for detection or perhaps a variety of detection techniques that are complementary. While the charges of the IEDs have remained consistent, the initiation systems of IEDs have become more sophisticated as insurgents have adapted to counter-IED efforts. Triggering mechanisms once employed rudimentary devices such as clothespins with electrical contact points. These devices were triggered by manually pulling a cord attached to the clothespin which separated the contact points. Other simple devices being used as initiation systems included battery powered clocks and kitchen timers. Perpetrators who have succeeded in getting access into an aircraft or building have innumerable options available for detonation. Insurgents now are more commonly using remote initiation systems, with 72% being radiocontrolled. These systems make use of hardware from wireless devices such as wireless door bells, keyless car entry devices, and radio-controlled cars. While these devices offer greater flexibility and anonymity when triggering the IED, they still require the perpetrator to be relatively proximal to the IED. More sophisticated initiation systems can increasingly separate the triggering agent and IED by using high-power cordless and cellular telephones. Additionally, sophisticated initiation systems are being combined to include a remote trigger (such as a cellular telephone) with a timing mechanism in daisy chain arrangements. The next wave of initiation systems appears to be inexpensive remote triggering made especially for IEDs. These devices provide very reliable triggering capability. Implantation of devices has also become increasingly sophisticated and devious, moving beyond simply burying them in the road. For example, bombs may be hidden in a variety of innocuous5 looking household items, camouflaged to look like surrounding objects such as rocks or construction rubble, concealed behind other roadside objects, hidden in brush, or encased in concrete, hidden in clothing or in toothpaste tubes, baby bottles and the like. An even more insidious delivery system now being employed is to place the devices in an automobile being driven by a kidnapping victim who is released and made to drive to a check-point or other target where the bomb is then detonated. Even as new detection methods are developed, the ‘bad guys’ develop more ingenious methods to introduce new type of threats in increasingly new ways. It is important to continue development of new technologies for detecting the new types of threats. The wide array of IED materials and their delivery systems has required the development of an equally large variety of detection devices, each with their own strengths and weaknesses. Nonintrusive devices have been developed for bulk and trace detection of explosives, chemical and biological materials, special nuclear materials, dirty bombs, and other weaponry. Imaging techniques such as dielectrometry with microwaves and millimeter waves cannot detect explosives directly, cannot penetrate metal, but may be more successful than metal detectors in detecting non-metallic objects concealed under clothing. Such detection devices require human operators for data analysis and decision-making. In many situations it may be sufficient to detect concealed objects and not necessary to identify them. It is difficult to imagine an innocent reason for concealing an object under one’s clothing, so it is not always necessary to identify exactly what object is hidden under clothing. The vast majority of passengers have nothing concealed and will proceed through without interruption, saving time but preventing non-metallic threats at the same time. Scanning with synthetic aperture radar is limited by antenna apertures, and moving antennas to increase apertures is impractical but electronic beam steering or the use of large antenna arrays is immensely feasible. Holographic methods may be too time-consuming and expensive but could be used on a selective basis when something abnormal has been detected. Computed tomography and X-ray backscatter imaging reduce the demand for human operators, who need only respond to alarms, but current equipment is expensive and requires suspected people, packages or cargo to pass through the screening equipment (see, e.g., Figure 1, which shows a dual energy package search using CT imaging.). Nuclear-based techniques (e.g., thermal neutron activation, pulsed fast thermal neutron activation, nuclear quadruple resonance) are better suited to detecting explosives and can be used for large cargo as well as small packages but are unsafe for humans and can be very expensive, especially if used with cargo that must be taken apart, pieces inspected, and re-assembled. Many technologies currently have unacceptably high false alarm or false negative rates. Technologies that can be used in situations with controlled ingress/egress are very different from what defense personnel need to use to detect roadside threats. Separate from detection devices have been attempts to counter their deployment or detonation. For example, tens of thousands of planes in Iraq have been used to jam the radio-based detonators in Iraq, but such approaches can be extremely costly, resulting in a highly asymmetrical warfare environment in which terrorist and insurgent expenditures are minuscule in relation to the costs of detecting or preventing them. Moreover, devices produce low signals with non-unique signatures that must be detected against large background noise. 6 Figure 1. Dual energy package search with CT imaging techniques. The net result is that, notwithstanding advances in detection methods and technologies, detecting and countering current devices through technical means alone cannot be the only weapon in the counter threat arsenal. The problem is compounded by the fact that the threats vary dramatically in type and size. Although current trends involve using conventional explosives, chemical and biological agents and nuclear bombs could become the instruments of choice in the future, as demonstrated in the sarin gas release in Tokyo’s underground (Gearson, 2006). Thus, understanding not just the bombs but also the bombers is key to the success of detecting and countering IEDs. The Life Cycle of IED Events In general, the use of an IED consists of six main phases—1) Determination to engage in IED activities, 2) Planning and acquisition of resources, 3) Construction, 4) Implementation (transfer, planting the device, staging the activation), 5) Detonation and Observation, 6) Evaluation of Effectiveness and Response. Each of these phases may be carried out by different people. Figure 2 roughly outlines the IED lifecycle and the research focus of engineering and social and behavioral sciences. Most counter-IED funding and projects have concentrated on the detection of IEDs and detonation prevention through an engineering perspective. Technologies that fall into the detection and detonation prevention category may include jamming devices, blast blankets, Laser-Induced Breakdown Spectroscopy systems (LIBS), and unmanned ground vehicles (UGV). While technology has prevented numerous incidents, clearly much more can and must be done to detect potentially problematic incidents earlier in the IED lifecycle. As McFate (2005) notes, IEDS require knowledge, material, organization, and a permissive surrounding environment. Research from a social and behavioral science perspective can be utilized to thwart the IED process long before the IED is placed by understanding how bombers are recruited and indoctrinated, by detecting preparatory activities related to the planning and construction phase of bombings, by detecting suspicious activities presaging an imminent attack, and by understanding the social environment that permits or encourages such means. Social and 7 behavioral science research can also aid human intelligence efforts through better understanding of sympathizers who facilitate IED events, through refinement of methods for intelligencegathering, and through development of tools or methods for questioning suspects of impending or completed attacks. Figure 2. IED Lifecycle and research focus. Yet another way to consider how social-behavioral sciences can contribute to countering IEDs is to focus not on the “bad guys” but the “good guys” who are charged with countering IEDs, by analyzing how they can make effective use of the information gained from signal, image and sensor information. Figure 3 illustrates the interrelationships that could be examined. The current amount of data available to decision-makers and operators is overwhelming. Better understanding of human analysis, decision-making and planning processes could facilitate timely and accurate use of the signal, image and sensor data that are gathered. Fusion of sensor/image data analysis with human analysis is crucial. It could ensure that necessary and actionable information is pushed to operators in a timely manner. It could also contribute to the development of automated tools, for example, alerting those on the front lines to potential dangers or building computer agents that reason autonomously. Figure 3. Relationships among cognitive-social, information and physical domains. 8 Possible Scenarios The possible scenarios where IEDs and related threats have been used and where they could potentially be used are great in number and varied in context. Since the devices are not professionally manufactured, IEDs can range in shape and size and can be used in almost any situation where the intent is to cause damage or harm. Handmade suicide bombing vests and car bombs may be considered IEDs. The most accessible and frequent example of an IED occurs when perpetrators conceal explosive devices along a road and wait for a target to approach the hidden explosives. As a potential target nears the IED, an operative triggers the devices using a variety of mechanisms described above. This type of scenario occurs frequently in Iraq and Afghanistan. A less probable but potentially catastrophic scenario is that of a passenger boarding a plane with components of a potential IED device concealed under his/her clothes. Although roadside bombs are most commonly encountered, additional scenarios exist where IEDs have been used. One such example is the 1996 Olympic Games held in Atlanta, Georgia. In this case the bomb was deposited in a trash receptacle and triggered remotely. Other IEDs were encountered in the train bombings Madrid (2004) and India (2006) and the Bali bombing in 2002. In these cases, bombs were created and deposited in a crowded area and detonated. The impact of such an incident were it to take place in a crowded sports arena such as a Big Ten college football game with 100,000 plus fans crowded into a relatively small area is unimaginable. Another possible scenario is an Islamic fundamentalist who decides to become a suicide bomber using a primitive U235 (nuclear) bomb in a major city such as Tel Aviv that triggers retaliation, international interventions, massive annihilation, and a Dr. Strangelove finale. Future threats could also take place closer to American soil. South American and Mexican instabilities could result in insurgents and terrorists crossing U.S. borders to dramatize their positions or gain political leverage through bombings in the U.S. For these reasons, research on countering IEDs must include means of securing U.S. ports of entry by land, sea and air. This includes gaining accurate knowledge of what can be concealed and detected inside luggage, inside cargo, or on people. For example, as Figure 4 reveals, sample “threat vectors” for air transport are many and most involve humans, who are the main source of detectable behavioral anomalies. Perpetrators of IED Incidents The main thrust of the workshop was in considering the human side of IED incidents and specifically the bombers and their organizations. A full understanding would naturally require thorough treatment of the background of terrorism and political insurgencies, and distinguishing between externally initiated and “homegrown” terrorism (such as witnessed in the Oklahoma City bombings, the recently discovered Buffalo, New York terrorist cell, and the United Kingdom railway bombings)--topics that are well beyond the scope of this report. Instead, a brief synopsis is presented of the briefing material and issues that framed the research discussions. 9 Figure 4. Possible threat vectors in an airplane IED scenario. Experts in terrorism were asked to identify what we know and what we don’t know that should become the springboard for research. Shortcomings noted in extant knowledge were many. First, the knowledge base is largely anecdotal rather than drawn from statistically reliable and validated empirical data. This problem of anecdotal versus empirically-based research has also appeared in the larger issue of terrorist acts in general. Some researchers have attempted to compile large datasets of terrorist acts in hopes of uncovering clues about terrorist operations and the terrorists themselves (Crenshaw, 1995; Pape, 2005; Sageman, 2004). This data collection is difficult and extremely time consuming, however, it must be done if useful and reliable findings are desired. A new paradigm of evidence-based terrorism research is needed to address the IED problem. This includes strong theory and rigorously gathered and tested empirical evidence. To conduct systematic, empirically-based research, researchers must determine a reasonable control group which produces a productive comparison. Past research has examined social, psychological, and political similarities across various individual terrorists and terrorist acts (Pape, 2005; Sageman, 2004). However investigations concerning the differences between terrorists and the surrounding population have not been conducted. Without an effective control group against which terrorist actions can be compared, it is impossible to arrive at any specificity about individual terrorist characteristics. Research must consider what would actually constitute a legitimate control group for comparison purposes. Many individuals share the same characteristics yet few become terrorists. 10 Second, at the individual level, the quest for a terrorist personality has failed, perhaps due to overemphasis on individual histories and the fundamental attribution error of attributing behavior to stable individual characteristics rather than to situational factors. The clear lack of an effective control group has exacerbated this problem. Thus far, research has shown that the only similarity between individuals engaged in terrorist activities seems to be that such individuals spring from oppressed communities (Bond, 2004). It has been repeatedly suggested that bombers spring from poverty. However, the empirical results are mixed, indicating that bombers and those who support violent tactics often enjoy a higher socio-economic standing than those who oppose them, the 19 hijackers who committed the 9/11 attacks being a case in point of terrorists being neither poor nor uneducated (von Hippel, 2002). What attracts individuals to the current jihadist movements and willingness to annihilate Jews and other “infidels” through violent means (see, e.g., Figure 5) is far more complex than a single personality profile would imply. Behaviors of individuals must be understood within the full mix of socio-political and other contextual factors that govern human actions. Behaviors taken out of context are misleading. Figure 5. Protestors in London. Third, an overly cognitive model of terrorists has prevailed. As previously discussed, the findings of such an approach are suspect because of representativeness and weaknesses in comparison. Further, a focus on cognition at the expense of social or political factors will be insufficient and inaccurate. Pape (2005) has noted that the vast majority of suicide bombs have been directed at liberal democracies such as the U.S., France, and Turkey that are uniquely vulnerable to suicide bombings and other unconventional warfare mechanisms. The root causes of suicide bombings may be applicable to deployment of IEDs. Thus, a thorough exploration of the economic, political, and social context will be required (von Hippel, 2006). At the same time, a “top-down” analysis may be necessary but will be insufficient to predict who will engage in terrorist acts and why. Such an analysis neglects critical individual characteristics that heavily influence perceptions, motivations, and goals. For any analysis of the IED phenomenon to be truly useful, a full range of macro and micro factors must be effectively considered. Fourth, static models must be replaced with dynamic ones. The context of terrorism, and the terrorists themselves, are constantly in flux. Terrorist acts are the result of long social processes that eventually shift from nonviolent to violent means. Theories and statistical models must recognize these dynamics. The Madrid bombing is a case in point. As illustrated in Figure 6, 11 individuals and terrorist organizations that had few or no connections with one another prior to 1994 created a very dense network of relationships by 2003. Figure 6. Connections among individuals and organizations involved in Madrid bombing, 1994 and 2003. Fifth, a renewed focus is needed on social organizations and the process of organizing in the context of terrorism. Considerable research has been performed on groups and their function; however, the context of terrorism presents many new aspects that need to be considered. Terrorist groups may be organized as cellular networks (such as al Qaeda) or as matrix structured organizations tightly integrated into the community (such as Hamas). The structure of these groups needs to be understood in terms of both the network of connections among members but also the basis for interaction, the linkage to resources, and the distribution across locations. Simply assuming that the groups are small world networks or hierarchies will not make them so. While network forms have been studied, the terrorism context suggests that the traditional forms studied (random, small world, hierarchy) are simply not applicable. Additional work that looks at the features of network topologies and how they can be sampled is needed (Airoldi & Carley, 2005). At present, groups that employ IEDs vary in their degree of organization from relatively informal cells, composed of long term friends and acquaintances forming diffuse networks to highly disciplined organizations characterizing such groups as Hezbollah and Hamas of FARC. Because the employment of IED type devices is typically a group phenomenon, it is important to understand the group goals and the conditions under which a given goal is activated and rendered capable of driving actions such as the launching of the IEDs. For instance, an organization like Hezbollah may have two types of goals, one related to their Lebanese identity and the other to the religious Shia identity. Whereas the Lebanese identity may be related to goals of economic development and political power to influence internal affairs in that country, the Shia identity may be related to the struggle against the West, hence it is more likely to lead to the employment of various belligerent tactics including IEDs. Similarly, an organization like Hamas (or the Basque ETA, or the IRA) may have political goals of influence and governance that may be incompatible with the launching of IEDS. It is important to understand the interplay of militant versus political objectives incompatible with militancy as it may affect fluctuation in the employment of IEDs. 12 Though the terrorist group and its goals are of critical importance and need to be better understood through research, the level of the individual and the process of individual radicalization is also of considerable significance. The terrorist group draws on individuals’ readiness to be recruited into its ranks and be engaged in various risky activities (including suicide bombing). Thus, it is important to understand the individuals’ motivations and the way these are manipulated by the terrorist groups to enhance recruitment. Available data and social psychological theory suggests the hypothesis that two sources of motivation are jointly involved in creating the readiness to join a terrorist organization, and/or to embark on a terrorist mission. These have to do with (1) the individuals’ personal (due to loss or trauma) and social (due to discrimination and alienation from the local society in a diaspora) circumstances creating frustration and the sense of disenfranchisement. (2) The collectivistic goals expressed predominantly in the terrorists’ public statements. Psychological theory suggests that these two sources of motivation may operate conjunctively, the personal/social frustration creating the preparedness to buy into the ideology and accept its objectives and means of attainment (including the launching of IEDs). Of importance also, these two types of motivational factors have different implications for possible counterterrorist strategies. The reduction of personal/social frustrations may lead to military strategies particularly careful to avoid collateral types of damage, whereas undermining the collectivistic/transcendental goals and the pursuit of militant means to these goals may require effective rhetoric and communication programs. It is also important to understand better the efficacy of the various counterterrorist efforts that have been attempted in the past (e.g. by the Americans, the British, or the Israelis) whose aim was to reduce the terrorists’ motivations to launch IEDs of various sorts. These included military means such as targeted assassinations of terrorist leaders, political means such as negotiations and or withdrawals from territories, massive arrests, and means reducing the efficacy of IEDs (e.g. by reducing the perpetrators capacity to reach their targets). Research should uncover the relative effectiveness of these attempted means on (1) the groups’ decisions to launch IEDs at a given time period, (2) individuals readiness to join the terrorist group and carry out its activities. Various open source data exist on those questions and it is essential to submit them to the appropriate modeling and statistical analysis (e.g. time series analysis attempting to explain the occurrence of IED attacks as function of prior counterterrorist initiatives). Terrorist groups seem to have adopted an almost organic quality to self-organize and fulfill their needs. These self-organizing groups have been studied in contexts apart from terrorism and are usually centered on common interests or similar goals. As part of the organization, communication channels are established, group norms are determined, and group hierarchy is laid out. The self-organization and formation of a group identity must be further understood. Additionally, in order for groups to form and to construct and deploy IEDs, a number of specialized skills and support need to be amassed or trained for. Researchers need to understand these skills and the conditions under which they can be acquired. Necessary support of the group may come in the form of monetary compensation or supplies; however, other types of support are just as critical. Groups must have a firm concept of the group goals and group ideology that supports the use of IEDs. The group also needs the explicit or implicit support of the larger 13 community in which the group operates. This support may be tied to the group’s ideology or be forced through coercion and intimidation. Given the extant group research, numerous methods exist for disrupting group function and illegitimating group goals. These strategies must be explored in the context of terror groups to determine their effectiveness. Sixth, the symbolic nature of IEDs is poorly understood. While the IED itself is comprised of explosives and circuitry, it also has symbolic meaning as an act of defiance in the face of forces with superior technology. The image of equipment damaged by an IED demonstrates the ability of terrorists to humiliate and inflict damage to an otherwise superior adversary. As this image is spread throughout the population, it alters the perceptions of the abilities of both the terrorists and the adversary (Gearson, 2002). Critical to understanding and countering IEDs is identifying what the IED represents and how that image can be effectively countered. Associated with the image of the IED is the effect of media and mass communication. Successful IED operations are widely available through multiple channels (e.g., the Internet, news broadcasts, etc.). Effects of media on a population have been explored previously but mostly in Western cultures. Public reactions and media influence are poorly understood within the populations of interest; even more poorly understood is how the media might be used to counter IED use. Themes Recommended for Funded Research and Development Eleven key issues emerged from the briefing’s brainstorming sessions and breakout sessions that were identified as prospects for research from a social-behavioral, engineering, or combined perspective. These were refined as workshop participants considered the topics in light of their expertise and experience and cast in terms of researchable theme areas. The final 11 topics were then ranked on importance as follows (see Appendix C for the numeric results). 1. Development of corpora for research 2. Fusion--of separate events into a single model and of sensor data with socio-psycho-cultural indicators 3. Model prediction and validation 4. Detection of deception and hostile intent 5. Culture Influences 6. Disruptive technologies and strategies of deployment 7. Social, psychological characteristics of terrorists 8. Counter-IED training and skill 9. Communication and popular media 10. Human-computer interaction in IED detection 11. Technology transitions These recommended areas for future sponsored research are detailed next. Many of these themes overlap in their theoretical bases and methodological approaches as evidenced by the discussion among participants. Such overlap provides substantial opportunity for the development of interdisciplinary and triangulated research efforts that will contribute to the development of corpora for research (see next section). 14 Development of Corpora for Research The most critical area in need of support and funding is the development of open source databases and corpora that would be available to researchers. This objective was endorsed by all workshop participants. There was general consensus that any Broad Agency Announcement addressing the IED issue should include a special core/infrastructure task with guaranteed ongoing funding to develop and maintain a database for all other projects. Currently, no single data set includes all the relevant data. Rigorous scientific research is also stymied by the lack of access to classified materials, which means that research operating in parallel may not inform other efforts. Participants agreed that multiple corpora would be needed to accommodate the differing needs of different lines of inquiry. For example, the needs arising from social network analysis will differ greatly from automatic behavior identification. The core corpora-building task may need to include development of datasets. A bidder might be paid to collect this information or could be paid to sanitize existing classified data. If real data cannot be used/sanitized in some fashion, an alternative suggestion was the use of multi-agent and red-team simulations to generate event data. There exist multi-agent models, such as the RTE model, that could be used for this purpose. One type of corpus needs to accumulate data on known IED incidents. Basic statistics grounding each incident would be included, such as time, location (region, city, or road), attacker identities, vehicles involved, victims, and so forth. Such a database would also contain relevant information about people, relationships, ideas, and contexts; more specifically, it could include information about specific individuals suspected of involvement in IED incidents or terrorist organizations, classes of individuals suspected as being likely recruits, the social and communicative relationships among these individuals or classes of individuals, and their organization’s religious, moral, and political values. Contextual information might include information related to the social, political, economic, and technological environments in which these individuals are embedded. The corpora would need to be dynamic, updating information as relationships, ideas, and contexts change over time. This ability would allow researchers to study emerging social networks and how loosely tied groups of supporters organize themselves to become an effective terror cell. Automated and semi-automated tools to support data collection, documentation, and formatting specifically associated with this type of data are needed in order for the corpora to be maintainable at a reasonable cost. A key issue here is that much of the data exists as raw texts, photos, and video clips. This implies both massive storage costs as well as the need for automated or semi-automated techniques for converting such data into the quantitative form needed for many types of research. A second corpus could include actual or experimentally simulated communications and other suspicious conduct to test for the ability to detect such conduct from a combination of overt observation and automated detection from images and sensors. This could also include virtual data generated by computer based simulation systems or on-line player games. One of the critical problems with current research in counter-terrorism (and countering IEDs) is the lack of a valid control group. Many individuals have the means and motivation to become 15 terrorists, yet relatively few of these individuals actually engage in directed terrorist actions. Thus far, the factors that set apart those who actually become terrorists from those who have the opportunity and motivation and yet refuse terrorism are poorly understood. The corpora should contain both types of individuals: people with adequate motivation and means to become terrorists and actual terrorists. With such corpora, high-specificity models can be developed which may predict future participants in terrorist acts. The corpora should be built on ground truth and should be reasonably large. Sources for such data could include captured documents, trial accounts, validated testimony, and so forth. Corpora containing large numbers of individuals would increase the reliability and ecological validity of any findings. Other corpora that are needed are ones that include behavioral and sensor data on actual activities of persons of interest, gathered either in field settings or from experimentally generated data (Burgoon, Twitchell et al., in press). The types of research that might generate a given corpus are described more fully below. Of note here is the need to make videotapes, audiotapes, transcripts, psychophysiological charts, and other raw data available for further exploration on the order of the kinds of collections that have been assembled in the NIST meeting room corpus. It is also important that the development of data banks be organized in terms of the relevant theoretical questions that one would want answered by means of these data. Specifically, different types and categories of data are suited for addressing different questions of interest. Though event-types data are useful for examining the impact of counterterrorist initiatives on the frequency and type of IED events, social network analysis is useful to address the nature of various terrorist organizations, whereas population surveys, simulations and experiments can be useful to determine attitudinal and situational variables likely to promote the tendency to volunteer for terrorist activities. After all, the overriding purpose of terrorism related research is to reduce the incidence of terrorist activities (such as the IEDs) and this will be enabled only if certain critical, theoretical questions are answered in terms of solid empirical data. Thus, the “bottom up” collection of data and data bank development should be constrained by questions the answers to which are judged necessary to carry out desired counterterrorist activities. Finally, access to the corpora should be restricted to properly credentialed researchers who can judge findings that result from the corpora. Inclusion of a peer review mechanism would ensure that valid conclusions are drawn from the corpora and that sensitive information is protected. To develop such corpora for research, the following steps are recommended: 1. Develop a common ontology for effective data storage that facilitates sharing between researchers (and perhaps government agencies). Develop thesauri that may relate the corpora to additional resources. 2. Develop a set of theoretically guided questions to which the data based answers will be deemed useful for countering the employment of IEDs. 3. Investigate synthetic/simulated database generation and use red team techniques to test this data. 16 4. Create a wiki-style environment where researchers can upload their open-source data, participate in discussion forums, and post open issues or collaboration opportunities. 5. Develop procedures to “clean” data and possibly pay to "sanitize" or "de-identify" datasets that are currently classified or otherwise not freely available. 6. Organize a workshop (perhaps with initial grant winners and researchers funded by sources other than the NSF) on data requirements and how to include data in the corpora. 7. Create simulated data which may substitute for unknown or unavailable datasets. As a final caveat, development of these corpora can answer many important questions but it must be recognized that tracing incidents and the actions of individuals of interest is insufficient to answer questions about terrorism and its processes such as radicalization, mobilization, recruitment, motivation, and termination. Additional social science research is urgently needed to address these longer-term issues. Fusion Workshop participants universally agreed that development and competitive testing of data fusion methods was important. The adaptation of traditional data fusion techniques that address information assurance correlation problems offers a potentially high-payoff toward development of unified models for high success-rate IED prediction capabilities. Correlation of vulnerability information from disparate sources has long been an issue in information assurance research. Traditional data fusion research studies the correlation of information from sources as diverse as single-purpose sensors to all source media information. In a similar fashion, independent threads of data on events within close proximity of known vulnerabilities, existing socio-political configurations, terror-network communications, and suspicious financial transactions could be unified to determine if a concerted IED attack signature is indeed in place. Various techniques for data correlation such as expert reasoning and fuzzy logic algorithms can be explored with appropriate technical extensions to draw meaningful near real-time prediction capabilities. Two general types of fusion research were envisioned: of event indicators into a single model and of sensor data with psycho-socio-cultural data. As regards prediction and modeling in the former case, no single measure can provide 100% sensitivity (accurately detecting the target signal such as a hostile individual) and 100% specificity (accurately discriminating non-target signals such as a non-hostile individual). Rather, different measures contain partly independent information that must somehow be combined or fused together for effective models. Research in statistics and artificial intelligence has pursued this question in the past; however, these models are often subject to limiting assumptions about the data or require large amounts of data to develop robust models. Further, fusion of time series data poses a difficult problem. To be able to explore complex models of perpetrators of IEDs, more complex methods of data fusion must be developed. These methods must not be subject to limiting assumptions, must be trained using limited datasets, and must be robust and parsimonious when dealing with time series data. Fusion methods could be relatively domain-general research, in that several domains could benefit from any techniques that were developed. While it was deemed reasonable to develop 17 and test fusion approaches using most any sufficiently complex dataset or domain, competitive evaluation of fusion approaches in the IED domain might require a substantial IED-specific corpus incorporating ground-truth information. Thus, fusion capabilities for use in the IED domain will in part depend on the quality of available corpora. High fidelity simulation tools can also be used to generate virtual data for testing and contrasting detection and fusion capabilities as has been done in the epidemiological area (Carley et al, 2006). As regards fusion of behavioral and sensor data, potentially fruitful topics for research include (1) the development and unification of novel information-centric data fusion models appropriate to fusion of disparate sensor data and (2) fusion of complex multi-dimensional information-rich probabilistic representations of plausible threat scenarios. Other issues surrounding fusion include best practices for data reduction of a large number of multimodal indicators, selection of indicators or composites to be combined into predictor models, use of serial versus simultaneous fusion approaches, best statistical or data mining models for classification, approaches for incorporating behavioral dynamics and handling time series data, and capacity to achieve realtime fusion of indicators so that preventive measures can be taken. Relevant questions in this research area include: 1. How can we combine moderately accurate indicators into a single robust model of IED prediction? 2. How can machines learn without large amounts of training data and without assuming stationarity? 3. How can diverse data such as sensor, geospatial, social, and political data be combined with individual characteristics to increase sensitivity and specificity in a model of terrorist activity? 4. Can models be developed that accept input data at multiple time scales and provide continuously evolving predictions based on temporally evolving inputs? 5. Which modalities/sensors account for the most variance: a. individually, when single modalities are available? b. in combination, when multiple modalities are available? 6. Can models be developed that produce real-time or near-real-time prediction of hostile intentions by would-be bombers? Model Prediction and Validation This research area aims to develop models for the prediction of IED potential as well as validate the developed models for confidence and reliability. Model development to predict IED events is hampered by data availability. Therefore, the research area depends closely upon the data fusion research described above. While theoretically derived models can provide a general trajectory of event potential, empirical data are necessary not only for determination of model parameters but for model validation. As of now, only limited data are available for IED research and significant amount of IED data remains classified. Validation and performance estimation continues to be a difficult task due to small sample sizes and unrepresentative samples. Of particular concern is the lack of a representative control group in model performance evaluations. This is an area that needs considerable work. 18 Rule-based systems have potential to provide useful validation tools in determining the operating state (quality) of the IED prediction models. In the field of control theory for complex dynamical systems, rules/hypotheses built through Bayesian probabilistic data association methods are coupled with Kalman filtering techniques to enable the creation of gates that provide appropriate validation certifications on the available sensory data. For the task of validating the IED prediction models, similar rule-based systems could be studied to estimate sensor faults such as bias, drift and ultimately to predict spatio-temporal model degradations. One option for developing a specific control group is to conceptualize insurgent/sectarian perpetrator groups as typical street gangs. A great deal is known about gangs from around the globe. For example, consistent with research cited above about intra-terrorist group activity, gangs appear to vary by the nature of their social network ties. McGloin (2005) found that gangs in Newark, New Jersey are loosely organized with pockets of cohesion. Importantly, several gang members emerged as connectors who link groups for various purposes. Understanding this linking role serves to describe terrorist groups or gangs in terms of their structure, governmental processes, and task functions which is critical to predicting their criminal activities (Ruble & Turner, 2000). Many other studies have also focused on gang organizations building models to understand how they work. In a frequently cited study, Decker, Bynum and Weisel (1998) compared African American and Hispanic gangs in both San Diego and Chicago. They focused on how gangs conduct their entrepreneurial activities, how they are organized, how they infiltrate neighborhoods, hide in communities, and exert political pressure in local government. Finally, in a study comparing gang and nongang violence, Maxon, Gordon and Klein (1985) found that gang violence tended to be more public, involve automobiles, firearms, and result in injuries to victims. These studies suggest a great deal of similarity between street gangs and insurgent/sectarian gangs in Iraq and Afghanistan. Clearly, much is known about street gangs that can serve as a control group for terrorist activities in Iraq and other countries. One approach to developing this control group would be to study gang structures across several cultural groups with different levels of influence in their communities with the goal of creating a model that predicts specific criminal activity. This process would then guide the same kind of model building among terrorist gangs. How are they organized and funded, how do they function in their respective communities, how are they funded, and so forth? Thus, there is some promise in pursuing gang models as a control group. Another key component of the model involves mapping terrorist/gang activity. Spatially explicit methods, especially with the aid of Geographic Information Systems (GIS) technologies, have been broadly applied to address environmental and social issues. When used in IED research, these methods can be used to identify hot spots and predict the likelihood of IED emplacement. For example, risk modeling in the domains of severe weather, criminal events, and social and individual behaviors elicits hot spots of events and occurrences (such as hail damages and criminal analysis; see Figure 7 as illustration). If an IED database is available, which consists of past bombing events and convicted cases, GIS methods can be developed to identify IED hot spots and spatiotemporal trends of IED emplacement. Furthermore, the IED events can be contextualized with other geographic environments, such as transportation networks, city landscape, social groups, cultural settings, and others to delve into the boarder and richer geographic context that brings together bombing and bombers to improve hypothesis building 19 and enhance explanation power of the IED model to be developed. In addition to predictive models, the ability to identify similar cases and reveal critical junctures in the evolution of similar cases can provide important insights into the development of the current situation and highlight areas of concerns. Comparable research has been done in database modeling and information query support for storm tracking (McIntosh & Yuan 2005; Yuan, 2001). Such conceptual frameworks and spatiotemporal algorithms may be adaptable to IED cases. Figure 7. Illustration of spatial mapping technique. A key issue here is linking the social network information on terrorists engaged in IED activity with geo-spatial information on such activity. Even preliminary studies show that different terrorists and terrorist groups operate in different locales, and that different kinds of IED attacks appear to occur in different locations. However, to leverage this information, it is necessary to think of overlaying the socio-cultural space and the geo-physical space. Another aspect of the modeling and prediction theme concerns the timeliness of predictive models. IED threats require real time decisions. Decision making can be impeded or confounded by overloading the decision maker with different kinds of data. As illustration, the TSA agent at an airport line needs to make a split second decision on the potential threat posed by a would-be passenger. Currently, agents have to process most of the data in their head putting together visual cues, ticket/ID information and sensor data as may be available. A critical objective of building computer models to augment human judgment is to derive methods to produce real-time or nearreal time projections. Mathematical strategies used in control theory such as model reduction may be adapted to the IED scenario. Models must be developed to condense vast amounts of data into ‘byte’ size pieces of information that can be used in a timely manner. Eigenvalues and eigenmodes are often used to describe succinctly the dynamical response of an otherwise complex system in a more compact manner. Although eigenvalues and eigenmodes may be infinite in number, there is an ordering and a hierarchy in their strength and dominance. Model reduction is a further filtering and condensation of a data set so that control of the dynamical system can be achieved more efficiently and robustly. Such concepts may be adapted to countering IED threats. Relevant questions in this research area include: 1. Can IED emplacement be predictable? a. How can hot spots be identified? 20 2. 3. 4. 5. 6. 7. b. What are the existing spatiotemporal methodologies and conceptual frameworks applicable for modeling of IED emplacement and social support networks? What is an effective process for validating counter-IED models? a. What data should be used? b. What is an appropriate control group? c. How can competing predictive models be compared effectively? What is the appropriate validation methodology for complex system models (multiagent, simulation, etc.)? How can the performance of a predictive model be estimated using limited datasets? How can GIS and social network information be combined to enhance prevention and prediction efforts? How can pattern recognition and neural network methods be applied to model development? What model reduction techniques should be applied to fused data sets to reduce data overload and enable real-time decision making? Deception and Hostile Intent Detection Two general approaches are available for detecting deception and hostile intent. The first is to model and predict hostile intentions of humans from behavioral observation or physiological measures For a number of years work in deception and intent detection has been progressing with promising results (e.g., Bond & DePaulo, 2006; Burgoon & Qin, 2006; DePaulo et al., 2003; Meservy, Jensen, Kruse, Burgoon & Nunamaker, 2005; Vrij, 2000). Workshop participants recommended that this effort be expanded to counter the IED problem. Work in deception and intent detection has focused on automatically observing and drawing conclusions about an individual’s observable behavior or physiology (Burgoon, Adkins et al., 2005; Pollina, Dollins, Senter, Brown, Pavlidis, Levine, & Ryan, 2006; see Figure 8). Polygraph Eye Tracking Figure 8. Sample psychophysio logical measures. Laser Doppler Vibrometry 2.600000 Thermal Imaging 1.300000 LDV vel Muscle Tension Tremor 0.000000 4.00000 0 LD Vve l -1.300000 0.00000 0 V ol ts 34.00000 36.00000 38.00 000 seconds 40.00000 -4.000000 Cardiovascular 60.00000 62.00000 seconds 64.0000 0 66.00000 0.100000 -8.000000 LD Vve l 0.000000 0.100000 L D V v e l Heart Sounds Respiration 0.000000 -0.100000 -0. 100000 5.860302 -0. 200000 38.0000 0 40.00000 42.00000 seconds 44.00000 -0.200000 44.00000 V olts 0.000000 38.0 0000 LD Vve l 40.00000 42.00000 sec onds Voice 9/23/2006 Detecting and Countering IEDs -5.860302 V o lt s 11.720604 V olts Vo l ts 9 21 Physiological monitoring is primarily relevant in the context of interpersonal interaction and direct questioning of suspects, but some remote-sensing technologies are available and/or under development. Here, relevant indicators include heart and respiration rates, blood flow, electrodermal responses, electroencephalograph changes, eye movement, pupil movement and dilation, micro-muscle tremors in the voice and face, and other neurophysiological responses— using either contact-based or non-contact measurement devices that may be combined, as for example into a single booth (e.g., Figure 9). A useful adjunct domain of investigation—most relevant for the interpersonal-based techniques—includes development of appropriate questioning patterns, control questions, and examiner behavior to elicit true-positive responses and minimize false-positive ones. Ideally, neurophysiological measures (e.g., anomalous brain activity patterns) would be coupled with information from observable behaviors (e.g., facial micro-expressions) to more accurately identify a deceptive subject. This multi-modal approach would need to proceed in concert with the development of data fusion techniques (discussed above). Figure 9. Israeli-developed Cogito prototype that measures multiple psychophysiological indicators during questioning. Unlike facial images and fingerprints for person identification, there is no one behavioral indicator that is known to reliably signal deception or intention on its own. Deception and intention often manifest themselves in many facets of human behavior including visual phenomena (such as facial expression, gaze direction variation, and gesture), audio phenomena (such as vocal pitch and longer response latency), and physiological phenomena (such as autonomic arousal as indicated by pupil dilation) (DePaulo et al., 2003; Ekman, 2001; Zuckerman & Driver, 1985). Multimodal approaches to detection of deception and hostile intention are likely to prove most powerful. Channels of nonverbal communication that have been amenable to analysis in the past include vocalics (or paralanguage), which pertains to human vocalizations; proxemics, which pertains to distancing and spacing patterns among humans; and kinesics, which pertains to all forms of body movement, including facial, gestural, postural, and gait. Automatic and persistent surveillance equipment could be placed to monitor high traffic areas, public spaces, or meeting rooms to flag suspicious behavior and possible suspects. Videotapes could be collected of all people questioned at a customs check point, and analyses conducted of individuals whose answers to 22 questions are verified versus contradicted by inspections. The videotapes could be scored for various behaviors. People who “failed” the inspection might be asked to undergo further questioning in which psychophysiological measures are collected with polygraph, infrared, thermal, and EEG measures are collected. Thermal imaging data might also be collected. Such types of research would be applicable to a range of circumstances where surveillance occurs or where suspects and informants are questioned. Temporal features of nonverbal behavior are likely to be critical. Previous research has emphasized morphological features to the neglect of dynamics. In previous research, we found that spontaneous and deliberate facial expressions having the same appearance could be discriminated based on differences in their timing (Cohn & Schmidt, 2004). More recent work shows that amusement, embarrassment, and polite smiles can be discriminated based on a combination of morphological and dynamic features (Ambadar & Cohn, in preparation). Another critical research direction should examine which potential indicators can be automated, which require solutions that augment human intelligence, and which are best left to human judgments. For the foreseeable future, an interactive system would seem preferable to one that is fully automatic, but what forms it might take and with what efficacy is also a researchable question. The other general approach to detecting hostile intent is to utilize sensors to detect concealed weaponry on humans. This line of research would be especially applicable to assessing threats in crowded environments. Metal detectors are commonly used at the entrance to many public areas such as airports, government buildings, sports arenas and train stations. However, perpetrators of terrorism are always devising new means to accomplish their destructive goals. It is clear that most handguns will be detected by a metal detector, so we have the shoe bomber incident that may have become a major incident but for the alertness of passengers and the cabin crew. New methods are required to detect non-metallic threats concealed under clothing. This is presently a real Achilles heel in the detection strategy used by security staff. Hand baggage undergoes more scrutiny with X-ray inspection and there is no routine inspection of people except for the ability to detect metal. Even for metal, different airports have different thresholds for detecting metal as evidenced by the ability of passengers to walk through with varying amounts of metal jewelry. The metal jewelry could be easily substituted by a small blade or other weapon concealed under clothing. Use of microwaves, millimeter waves and TeraHertz radiation have been proposed for remote detection of concealed objects. This approach may be especially feasible in controlled ingress/egress scenarios where people have to walk through a checkpoint in some orderly fashion. Dielectroscopy or mapping the dielectric properties of an inhomogeneous object such as a person may be used as rapid coarse screening method to detect objects concealed under clothes and shoes. It is not necessary to precisely identify and image the concealed objects, it is only necessary to detect its presence. Clothes are nearly transparent to millimeter wave radiation and objects such as ceramic weapons, fertilizer, liquids and gels hidden under clothing will provide a very good dielectric contrast under clothes and hence will be easy to detect. Safety issues are expected to be minimal, since the power levels required will be comparable to that emitted by cell phones and the sources will be much further away from the body than a cell phone. 23 Relevant questions in this research area include: 1. What neurophysiological measures and observable behaviors are reliably associated with actual hostile intent and with preparing, planting, detonating an IED? 2. Which modalities/sensors are most helpful in identifying these behaviors? 3. What moderators influence behavioral signatures? (e.g., culture, language in use, gender, scene)? 4. How time-invariant or dynamic are behavioral patterns? 5. What high-dimensional models can describe behavioral patterns? 6. What observable behaviors are most easily recognized automatically? 7. What is the optimal frequency window for mapping objects concealed under clothing or in footwear? 8. How can behavioral data be fused with sensor data to detect the setting of an IED? 9. How can automatic extraction and analysis of multimodal behavioral data be used most effectively to assist human interviewers in intention detection? 10. What are potential safety issues and research on dielectric absorption by the human body at millimeter wave frequencies? 11. What are the potential privacy issues of utilizing sensors for assessing threats from humans? 12. What technologies are needed to enable deployment of sensors in public places? a. Can electronically steerable antennas be developed and applied to the IED scenario? b. Can compact and inexpensive TeraHertz sources be developed? 13. How can system configurations by optimized to lower false positives and false negatives while maintaining throughput? Cultural Influences Researching cultural influences is a complex and difficult problem that will require a great deal of systematic investigation. Cultural characteristics of perpetrators of IEDs and those who support them have long been forwarded as a contributing factor to IED use. However, the evidence to support this conjecture is largely anecdotal. Needed is statistically validated information on cultural differences in intentions, views of morality and mortality, attitudes toward use of lethal weapons, and behavioral indicators of intentions (see, e.g., Global Deception Detection Team, in press). Cultural influences also need to be understood within the context of the permissive majority in the surrounding population who allow IED bombings to take place (McFate, 2005). Of particular interest is research pursuing any culturally based means that can be used to decrease the popularity of IED events. Cultural differences also exist across and within terrorist groups. These differences impact the social and resource networks that form, the strategies for socialization of new members, the role of women in the groups, and so on. Both social network and multi-agent modeling of terrorist groups could take such cultural information into account; however, there is not only a dearth of such information but also the current technologies are limited in their ability to handle that information were it available. Of particular interest are social network metrics that are sensitive to cultural differences and multi-agent techniques that can take cultural differences at the strategy level into account. 24 Relevant questions in this research area include: 1. What are the cultural factors that influence values, behaviors and neurophysiology in perpetrators of IED incidents? 2. Does cultural homogeneity versus heterogeneity of interviewer-interviewee relationships influence information acquisition? 3. How can IEDs be understood in evolving conflict (warfare doctrine) context a. Coupling IEDs with other terror engagement mechanisms (kidnappings, beheadings) b. Under what conditions a given terrorist group will radicalize its tactics, or vice versa will attenuate its militancy. Do such conditions vary across terrorist organizations from different cultures 4. How can the Worldwide Web best be used to collect pertinent cross-cultural data? 5. What styles of networks are common in different cultures? 6. How do cultural differences in socialization and enculturation processes influence the social networks that form and the way they are used to generate action? Disruptive Technologies and Strategies of Deployment Workshop participants discussed how technologies could be used to obfuscate and reverse collaboration between insurgent groups. This area would require the understanding of group dynamics, norms, and accepted practices in order to be effective. Further, the disruptive efforts must allow for adapting group tactics. Suggested disruption targets may include inter and intra group communication, mass advertising of terrorist acts, and recruiting technologies mechanisms. Decision makers also need statistical models that identify the probable outcomes of a given decision tree. Critical examples were provided showing that a standard social network approach to disrupting terrorist groups was unlikely to be effective. The standard approach would be, given the network of terrorists, to locate those individuals with a high degree of centrality or betweenness and then isolate them. Several problems exist with this approach, among them: (1) the network data are often incomplete and rapidly changing, decreasing the robustness of these metrics, and (2) terrorists who are critical in a terrorist network are often not as “visible” in terms of their social connections but rather are critical due to access to specialized resources, locations and so on thus creating the need for dynamic network analysis tools that take into account multi-modal networks. Alternative tools that take into account multi-mode, multi-link data under levels of uncertainty, however, show promise (Carley 2003, 2004). Network concepts could also be applied to analyzing information networks among ordinary citizens to create focused awareness of potential threats and thus eliminate IED threats earlier in the IED lifecycle. An example of tapping into a network of ordinary citizens to raise awareness is currently ongoing in Great Britain. Shoreditch TV 2006 is an experiment in beaming closedcircuit surveillance TV into the homes of ordinary citizens who can monitor and alert police to any potential threats. A possible implementation of such a network infrastructure might involve 25 distributing large numbers of low-cost text-messaging devices to Iraqi citizens. Figure 10 displays an example of a small, inexpensive text-messaging device. These devices could be adopted and used primarily for unrelated communication (e.g., business dealings, personal communications) yet also enable citizens who come upon suspicious situations to transmit salient details via text or simply press a button which sends general location information (by cell tower) to a central information warehouse. Based on the number of messages and type of information reported, resources could be focused on investigating high-priority incidents. The transmissions would be confidential and secure and greatly lower possible barriers to reporting insurgent activities. The network infrastructure would enable large-scale anonymous reporting that is commonly used to capture at large criminals. Figure 10. An example of an inexpensive text message device (Blackberry, 2006) Certainly any such system could potentially be misused. Misuse might include orchestrating ambushes, riots, and conflicting leads. However, creating a low-cost communication infrastructure that utilizes ordinary citizens, motivated by appropriate incentives, might create a culture of information sharing that would lead to better awareness of IEDs and the activities of those who utilize them. This and similar information networks of citizens are researchable implementations that could lead to a better understanding of the IED lifecycle and the patterns and social interactions of people engaged in such behavior. Requirements that would need to be examined would include cost, anonymity, trust in the network, feasibility of using a geographic information system (GIS) or GPS, ease of use (e.g., icon versus text messages), security of data transmission (encryption), multifunctionality of the individual devices (more than just notification), and motivation to use the devices. Relevant questions in this research area include: 1. How can collaboration be "reverse engineered" and ingroup formation exploit schisms between and within groups to weaken or disrupt movement up the pyramid? Can, for example, outgroups be created or can alternative views of ingroups be bolstered? 2. How can we attack peer-peer organizational structure using scalable technology? 3. How do the intended audiences of terrorist organizations find the "advertising" of bomb blasts and related threats? Can such advertising be disrupted and if so, with what effects? 4. What new metrics are needed that take into account actor criticality from a resource, access to media, location perspective? 5. How can critical actors be identified given highly incomplete and time varying data? 26 Social, Psychological Characteristics of Terrorists Closely tied to the cultural influences research are the social and psychological characteristics of those who participate in IED events. A thorough understanding of these perpetrators may yield useful information which could be used to counter IEDs. Such countering efforts may come from hampering the recruiting process and from identifying active terrorists and those sympathetic to their cause. Along with understanding the terrorists themselves, it is also important to understand those who explicitly support IED bombings or implicitly allow IED bombings to take place. Understanding the social and psychological qualities of the IED perpetrators’ support structure and influence processes are additional fruitful avenues to explore. Research on ingroup/outgroup processes, for example may yield insights into how ingroup/outgroup relationships are formed and how ingroup members are influenced to support increasingly lethal means of achieving their ends. For example, ingroup members may recognize and tolerate varying viewpoints of other ingroup members but view outgroup members as a monolith, making for easy scapegoating and stereotyping of the enemy. Fundamental social science research on how to sway those who are attitudinally ambivalent, tentative, or on the periphery should guide what kinds of persuasion campaigns are needed to deter potential recruits from joining violent cells and organizations. Social psychological work has shown that sympathizers with extremists, who are in the minority within the ingroup, may exert influence on non-obvious, peripheral issues and eventually result in more support for initially rejected positions such as “kill all Americans.” These indirect forms of influence must be understood for their implications for ingroup members solidifying views of their membership and for the kinds of countermeasures and messages that might employ similar indirect methods to secure aid from non-hostile majority or minority group members and to separate radical minority members from their ingroups. It is important to understand the terrorists and those that support them not just as individuals but also as groups and organizations. For example, multiple groups claim credit for IED events. Each of these groups are likely to be structured differently, be located in different areas, have access to different resources, and have different recruitment activities. The interactions among terrorists and their linkages to different groups are avenues through which they influence others and can be influenced by others. A better understanding of these networks, how the shape of these networks impacts social influence, and what actions the individuals and groups take is needed. Experts agree that the launching of IEDs (including suicide attacks) is primarily a group phenomenon. It is therefore important to understand the conditions under which individuals’ tend to over-identify with their group and deemphasize their identity as individuals and their individualistic goals (of education, family, economic well being, etc.) that are by and large incompatible with militancy. Social psychological research is suggestive in this regard as it indicates that the emphasis on mortality or the need for cognitive closure tends to increase individuals’ tendency to become “group centric” and to mesh their personal motivations with group related objectives that often may be channeled in the direction of militancy toward some identified outgroup. Relevant questions in this research area include: 27 1. What are strategies used to recruit terrorists and what are methods to counter them? 2. Are there social and psychological motives behind IED use and can we develop social and psychological counter-motives? 3. How are such motives induced by situational events, e.g. perceived injustice and humiliation by an outgroup, a lack of means to pursue individualistic goals, how are they fueled through the blogosphere, the internet, social and religious institutions? 4. What social, psychological, and ideological factors should be included in a threat profile to more accurately identify current and future terrorists? 5. How can we influence those who implicitly allow IEDs to be manufactured and placed? 6. How are terrorist groups structured? 7. How does the structure of the group enable it to influence members to particular actions? 8. How does the way in which terrorist groups are linked together and to non-terrorist groups constrain or enable violent activity? Counter-IED Training and Skill The armed services have a number of specialized, highly trained teams whose job is to clear suspect roads of any IEDs or other threats. These teams may operate on a regular schedule for a well-traveled highway, but may also clear smaller roads by request. Within these groups is contained a considerable amount of knowledge and skill that could potentially be recorded and utilized in training programs and automatic detection efforts. Turnover and lack of knowledge specification has made the process of capturing this knowledge difficult. It is suspected that effective human IED detectors use heuristics to detect IEDs. If these heuristics can in some way be captured, effective methods of IED detection could be shared more broadly. At the same time, training related to detecting hostile intent and deception is needed to increase the sensitivity to such conduct from all those on the front lines where hostilities occur and those responsible for security screening efforts. In detecting IEDs or IED perpetrators, some soldiers have developed a set of heuristics that serve them well. These soldiers should be interviewed to identify these heuristics. Research should then be conducted on whether these heuristics can be incorporated into training programs or operationalized as metrics (perhaps as dynamic network metrics) so that they can be run on new data as it arrives. Relevant questions in this research area include: 1. Why are some human IED detectors so effective? a. What skills do they have that could be spread to others? b. Study reservists to identify skill at detecting and why they are skilled. 2. What competencies are needed to be skilled detectors? a. Does a specific background or prior experience affect IED detection 3. What do humans need to do to get better at detecting IEDs? 28 a. Can training be an effective way to improve IED detection or is ability based solely experience? 4. Can human heuristics be operationalized as rules or metrics so that they can be used in an automated or semi-automated detection system? 5. Can we study those IED detectors who are very effective, and contrast them with those who are not as skilled, to determine the skills and abilities that successful detectors have? a. Are there background characteristics, which might not obviously be associated with IED detection skills that empirically differentiate the very successful detectors from the mediocre ones? Communication and Popular Media IEDs have had a large impact on opinions surrounding insurgencies and terrorist actions. Much of this impact is due to communication and popular media channels (O’Hair & Heath, 2005). Studying these mass communication channels and the intended and unintended consequences of media coverage and mass dissemination of ideas and ideologies may yield useful insights into how insurgents and terrorists communicate among themselves, court new recruits, attract media attention, and persuade communities/networks of influence (Miller, Matusitz, O’Hair, Eckstein, in press; O’Hair, Heath, & Becker, 2005). Information diffusion methods and media popularity in areas that are sympathetic to terrorist causes will be particularly important to examine as they could potentially influence social and psychological characteristics (O’Hair, Bruce, & Shamas, in press). Previously unexplored avenues of research could be profitably addressed through: (1) integrating spatiotemporal models for predicting media use/influence (O’Hair, Yuan, & Siriko, in press), (2) developing dynamical modeling systems to account for changes in both terrorist/insurgents strategic employment of media and audience preferences for media (Corman & Schiefelbein, 2006), and (3) triangulating other components of IED research (incident analysis of hot spots, moderators of behavioral signatures, employment of corpora data base) to improve predictive models of media strategies and influence. Innovative communication research initiatives might also include proactive strategies involving interactive online messaging with suspected terrorists or sympathizers in an effort of influencing them as well as their online communities (Matusitz & O’Hair, in press). Messages of this fashion could include reference to media content that would serve persuasive efforts. Multi-agent models that take into account the underlying social and telecommunication networks can also be used to assess the impact of various interventions and new technologies on information flow. Key issues here are understanding the communicative power of various concepts and understanding how socio-demographic and cultural differences influence choice of interaction partner and telecommunication medium. Relevant questions in this research area include: 1. What is the role of messaging and selection of channels that make the IED creation and support processes effective? 2. Can we automate blog profile detectors to identify leaders, potential recruits, and interaction networks? 3. What media choices do those sympathetic to terrorist ideologies make? 29 4. What media/message mix is most likely to sway public opinion in targeted areas? 5. Can IED use be attacked indirectly through the media? a. Perhaps through public service announcements concerning unexploded ordinances or by advertising employment for explosive experts. b. Or through innocuous venues such as musical, dramatic, or comedic formats (info-tainment strategies). 6. Are there indirect associations to IEDs that might be used in broad media channels to alter the appeal of IEDs to their users or public acceptance of them in targeted areas? 7. How can media analysis and audience segmentation research be reconstituted to more accurately profile terrorists’ strategies, gauge community involvement, and improve detectable actions? 8. How can online communication strategies be manipulated to redirect terrorists and their online communities toward traditional media content? 9. How can we effectively triangulate spatiotemporal data, incident analysis, area/cultural profiles, and triggering events in developing a dynamic and predictive model of media/audience influence? Human Computer Interaction in IED Detection One longer-term issue in reducing the IED threat is how to combat the formation and development of hostile groups. In addition to the mass media promoting or romanticizing terrorist ideologies, new interactive media such as Internet discussions have additional capacities. Internet systems such as search engines, websites, and chat rooms can connect otherwise isolated individuals who share ideological leanings. Moreover, as suggested by the “social identification/de-individualization” model of computer-mediated communication, when members of various social groups encounter similar others online, the nature of Internet chat tends to exaggerate commonalities and diminish the dissimilarities that might otherwise be apparent face-to-face (Spears, Lea, & Lee, 1990). Additionally, the hyper-personal model of online communication explains how relationships that accrue online may become particularly involving and intense, surpassing parallel face-to-face connections (Walther, 1996). These capacities suggest that Internet communication among prospective hostile collaborators may exaggerate and intensify their attitudes and determination to do harm. Research assessing hate groups online found that postings in such groups do not explicitly advocate violence, since members appear aware that such advocacies are illegal. However, many online communities and address rosters are maintained which may constitute dormant forms of contacts, should there be some prompt for collective action (Douglas, McGarty, Bliuc, & Lala, 2005). In light of controversies last January regarding WWW search engine providers’ resistance to providing records to national security agencies, and bad press regarding telephone companies’ disclosure of individuals’ call records, it is inadvisable to attempt to profile and monitor individuals’ Internet communication records. However, many ideological groups’ online discussion spaces are open to any viewer to observe even though participants often operate as if it were a private space. Examining such publicly-accessible Internet chat spaces may provide a legal and effective means of observing the development of hostile ideologies and collaborators. 30 Researchers have long sought to merge humans and machines in an effort to mitigate the weakness of each and accentuate the strengths of both. Numerous tools have been introduced that may help in the detection of IEDs. These tools may include explosive sensors, attention directors in surveillance tasks, and network science tools for identifying emergent leaders and groups. However, all of these tools are merely extensions of human information gathering and are subject to some amount of uncertainty. Of critical concern is the method in which humans utilize information from detection and analysis tools in the decision-making process. Human perceptions of reliability, human biases, and the limits of human vigilance all combine to potentially thwart any machine-assisted counter-IED effort. These issues need to be carefully addressed in connection with any automated tools meant to extend the human information gathering ability. Important is the presentation of material in an easy to understand format so that it can be rapidly and usefully assimilated. The concern was repeatedly raised that new technologies that do not produce easily understood and interpretable results simply will not be used. Relevant questions in this research area include: 1. How do we assess the certainty of information and add certainty into analyses? 2. What procedures or processes are needed to encourage scientists and engineers to develop technologies that require minimal interpretation and are easily understood by the ultimate users? 3. How can we evaluate the effectiveness of systems that rely on human interpretation? 4. How do we overcome automation bias in human-machine counter-IED efforts? 5. Can human perceptions confound IED identification? a. How can such confounding be reversed? 6. What is the optimum method of organizing the data for operators? Technology Transitions The workshop attendees identified one final area of research, although this area received the lowest amount of interest and support from the attendees, based on the widely-held perception that it was outside the workshop’s scope of identifying directions for future basic research. This area focuses on transitioning technology from the laboratory to the field. There are two primary motivations for improving technology transition: (1) providing up-to-date, state-of-theart technology for operational use to improve performance in the field and to “keep up” with a rapidly adaptive adversary, and (2) providing a forum for objective evaluation of research contributions and a source of feedback into the research process. Although both motivations are important, the second is most related to basic research questions. Because understanding of the IED domain is currently incomplete, an accurate model of which factors contribute to IED incidents and how these factors contribute is currently unavailable. This limits validation opportunities for research results and proposed methods. Performance metrics gleaned from operational evaluation could account for effects of otherwise unmodeled variables and significantly enhance the ability to shape future research questions based on this feedback. 31 A major challenge to researching technology transitions is that many critical field needs are either held constant or overlooked in a laboratory setting. In most cases, field operations are situational and dynamic in that pre-conditions continue evolving and cannot be fully controlled as in a laboratory. Near-real time data may be considered as obsolete when real-time data is deemed necessary for decision making. Geospatial data are acquired before operations, often from multiple sources, and are not necessarily correspondent with the current situation. For example, a building may have changed ownership after compilation of the geospatial data. Similarly, human behaviors may depart from typical patterns found in laboratory experiments. Members of a bombing team may exhibit distinctive manners from the behavioral characteristics identified from a laboratory setting. Decisions, however, need to be made under high stress and instantaneously to distinguish suspects from the general public. While laboratory simulations can be quite realistic, the scalar effects (e.g., models of human interaction built on groups of two or three people versus analyzing hundreds of people at once) cannot be fully reflected in a laboratory. The end users (e.g., commanders, human intelligence officers, soldiers) need to have a high degree of situational awareness and to adapt to the environment in which the technology is deployed. As such, the technology transition can also be hampered by a wide range of human and decision factors that merit basic research attention. Relevant questions in this research area include: 1. What are the organizational barriers to effective use and rapid deployment of nonDOD-developed technology to support defense missions? 2. What are factors critical to a successful technological transition to mission critical situations? 3. What are scalar effects of laboratory technologies to field applications? a. How can scalar effects in space, time, and quantity be identified? b. How can methods be developed to overcome scalar effects and other barriers to technology transitions? 4. Are there characteristics of end users that promote or inhibit adoption of transitioned technologies? Summary Technology is being pushed past its limits by IEDs. Explosives and detonation devices are now widely available. When a terrorist is willing to die, the weapons at his/her disposal multiply even more. It is now much easier to create chaos and make people feel afraid than to create stability and make people feel secure. It is imperative to continue to research new sensor technologies, signal processing algorithms, fuse multimodal datasets and develop condensation and reduction methods to prevent data overload. Threats due to IEDs in the war front and threats to our civilian population and civil infrastructure require widely different solutions and research methods. It is also obvious that sensor technology, screening and scanning is all post facto, after perpetrators have already planned and coordinated their implementation strategy. Anticipation and prevention of such threats is the preferred route. It is now much easier to destroy infrastructure than rebuild it. It is clear that response to IEDs and terrorism by oppression, occupation by an overwhelming force or even annihilation of an entire country, are not practical or acceptable to the world community. The alternative is to better understand the underlying social factors so our efforts 32 can be focused more efficiently. There is no single magic solution to this very complex problem. The blending of socioeconomic, religious/cultural factors with the action and policies of various governments are just as important as development of improved sensor technologies. An example of a potentially fruitful blend of technical and social approaches is how, where, and when electronic eavesdropping and data mining can or should be used to detect and monitor social group relationships and contacts. With funded support in these areas, social science will become thoroughly involved in the counter-IED effort and will join engineering and other technical fields in search for effective solutions to IEDs. This workshop has made it very clear that socialbehavioral scientists and engineers should work closely to develop multimodal approaches to a difficult, continuously evolving problem. Interdisciplinary research is key and the National Science Foundation is an ideal agency platform to coordinate such research. References Airoldi, E. O. & Carley, K. M. (2005). Sampling algorithms for pure network topologies: a study on the stability and the separability of metric embeddings. SIGKKD Explorations (Special Issue on Link Mining), 7(2), 13-22, http://www.acm.org/sigs/sigkdd/explorations/issues/7-22005-12/2-Airoldi.pdf. Baker, T. B. (1988). Models of addiction: Introduction to the special issue. Journal of Abnormal Psychology, 97(2), 115-117. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173-1182. Barry, J., Hastings, M., & Thomas, E. (2006). Iraq's Real WMD. Newsweek, March 27, 2006 Issue. Blackberry, http://blackberry.com/products/blackberry/index.shtml Block, J., Block, J. H., & Keyes, S. (1988). Longitudinally foretelling drug usage in adolescence: Early childhood personality and environmental precursors. Child Development, 59(2), 336355. Bond, C. F., Jr., & DePaulo, B. M. (2006). Accuracy of deception judgments. Personality and Social Psychology Review, 10(3), 214-234. Bond, M. (2004). The making of a suicide bomber. New Scientist, May 15, 2004. 34-37. Burgoon, J. K., Adkins, M., Kruse, J., Jensen, M. L., Meservy, T. O., Twitchell, D. P., Deokar, A., Nunamaker, J. F., J.r, Lu, S., Tsechpenakis, G., Metaxas, D.,& Younger, R. E. (2005). An approach for intent identification by building on deception detection. Proceedings of the 38th Annual Hawai’i International Conference on System Sciences. Big Island, HI. 33 Burgoon, J. K., Twitchell, D. P., Jensen, M. L., Adkins, M., Kruse, J., Deokar, A., Lu, S., Metaxas, D., Nunamaker, J. F., J.r, & Younger, R. E. (In press). Detecting concealment of intent in transportation screening: A proof-of-concept. IEEE Intelligent Systems. Burgoon, J. K., & Qin, T. (2006). The dynamic nature of deceptive verbal communication. Journal of Language and Social Psychology, 26(1), 76-96. Campbell, S. B., & Cohn, J. F. (1991). Prevalence and correlates of postpartum depression in first-time mothers. Journal of Abnormal Psychology, 100(4), 594-599. Carley, K. M., Fridsma, D., Casman, E., Yahja, A., Altman, N., Chen, L., Kaminsky, B. and Nave, D., 2006. “BioWar: Scalable Agent-based Model of Bioattacks,” IEEE Transactions on Systems, Man and Cybernetics-Part A, 36 (2), pages 252- 265. Carley, K. M., (2004). Estimating vulnerabilities in large covert networks using multi-level data. In Proceedings of the NAACSOS 2004 Conference, Pittsburgh, PA. Carley, K. M., (2003). Dynamic network analysis. in Dynamic Social Network Modeling and Analysis: Workshop Summary and Papers. R. Breiger, K. Carley, and P. Pattison, (Eds.) Committee on Human Factors, National Research Council, National Research Council. Pp. 133-145, Washington, DC. Chrousos, G. P., & Gold, P. W. (1992). The concepts of stress and stress system disorders. Journal of the American Medial Association, 267(9), 1244-1252. Clark, D. A., Steer, R. A., & Beck, A. T. (1994). Common and specific dimensions of selfreported anxiety and depression: Implications for the cognitive and tripartite models. Journal of Abnormal Psychology, 103(4), 645-654F. Cole, P. M., Jenkins, P. A., & Shott, C. T. (1989). Spontaneous expressive control in blind and sighted children. Child Development, 60(3), 683-688. Corman, S., & Schiefelbein, J. (2006). Communication and media strategy in the Jihadi war of ideas (Report #0601). Tempe, AZ: Consortium for Strategic Communication. Crenshaw, M. (1995). Terrorism in context. University Park, PA: Pennsylvania State University Press. Cummings, J. S., Pellegrini, D. S., Notarius, C. I., & Cummings, M. E. (1989). Children's responses to angry adult behavior as a function of marital distress and history of interparent hostility. Child Development, 60(5), 1035-1043. Cummings, M. E., & Davies, P. T. (1993). Maternal depression and child development. Journal of Child Psychology and Psychiatry, 35(1), 75-112. 34 Cummings, M. E., Zahn-Waxler, C., & Radke-Yarrow, M. (1984). Developmental changes in children's reactions to anger in the home. Journal of Child Psychology and Psychiatry, 25(1), 63-74. Decker, S.H., Bynum, T., & Weisel, D. (1998). Gangs as organized crime groups. Justice Quarterly, 15(3), 395-425. DePaulo, B. M. (1992). Nonverbal behavior and self-presentation. Psychological Bulletin, 111(2), 203-243. DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, L., Charlton, K., & Cooper, H. (2003). Cues to deception. Psychological Bulletin, 129(1), 74-118. Douglas, K. M., McGarty, C., Bliuc, A., & Lala, G. (2005). Understanding cyberhate: Social competition and social creativity in online white supremicist groups. Social Science Computer Review, 23, 68-76. Ekman, P. (2001). Telling lies. New York, New York: Norton. Gearson, J. (2002). The nature of modern terrorism. Political Quarterly, 73(4), 7 - 24. Global Deception Research Team (in press). A world of lies. Journal of Cross Cultural Psychology. Linde, Y., Buzo, A., & Gray, R. M. (1980). An algorithm for vector quantizer design. IEEE Transactions on Communications, 28(1), 84-95. Matusitz, J., & O’Hair, D. (In press). The role of the internet in terrorism. In D. O’Hair, R. Heath, K. Ayotte, & J. Ledlow (Eds.), Terrorism: Communication and rhetorical perspectives. Cresskill, NJ: Hampton Press. Maxon, C.L., Gordon, M.A., & Klein, M.W. (1985). Differences between gang and nongang homicides. Criminology, 23, 209-222. McFate, M. (2005). Iraq: The social context of IEDs. Military Review, May-June Issue, 37 - 40. McGloin, J.M. (2005). Policy and intervention consideration of a network analysis of street gangs. Criminology & Public Policy, 4(3), 607-635. McIntosh, J. & Yuan, M. (2005). Assessing similarity of geographic processes and events. Transactions in GIS. 9(2), 223-245. Meservy, T.O., Jensen, M. L., Kruse, J., Twitchell, D. P., Tsechpenakis, G., Burgoon, J. K., Metaxas, D. N., and Nunamaker, J. F., "Deception Detection through Automatic, Unobtrusive Analysis of Non-verbal Behavior," IEEE Intelligent Systems, Sept./Oct., 2005, Vol. 20, No. 5 pp. 36-43. 35 Meservy, T. O., Jensen, M. L., Kruse, J., Burgoon, J. K., & Nunamaker, J. F. (2005). Automatic extraction of deceptive behavioral cues from video. Intelligence and Security Informatics: Proceedings of the Third Symposium on Intelligence and Security Informatics ISI 2005, Atlanta, GA, USA. Berlin: Springer-Verlag. Miller, C., Matusitz, J., O’Hair, D., & Eckstein, J. (In press). The role of communication and the media in terrorism. In D. O’Hair, R. Heath, K. Ayotte, & J. Ledlow (Eds.), Terrorism: Communication and rhetorical perspectives. Cresskill, NJ: Hampton Press. Moriyama, T., Xiao, J., Cohn, J. F., & Kanade, T. (2006). Meticulously detailed eye model and its application to analysis of facial image. IEEE Transactions on Pattern Analysis and Machine Intelligence, 28(5), 738-752. O’Hair, D., Bruce, M., & Shamas, K. (In press). Media, culture, and message strategies: Identifying vulnerable populations. In D. O’Hair & R. Heath (Eds.), Handbook of risk and crisis communication. Mahwah, NJ: Lawrence Erlbaum Associates. O’Hair, D., & Heath, R. (2005). Conceptualizing communication and terrorism. In D. O’Hair, R. Heath, & J. Ledlow (Eds.), Community preparedness, deterrence, and response to terrorism: Communication and terrorism. 1-12. Westport, CT: Praeger. O’Hair, D., Heath, R., & Becker, J. (2005). Toward a paradigm of managing communication and terrorism. In D. O’Hair, R. Heath, & J. Ledlow (Eds.), Community preparedness, deterrence, and response to terrorism: Communication and terrorism (pp. 307-327). Westport, CT: Praeger. O’Hair, D., Yuan, M., & Siriko, T. (In press). Assessing, evaluating, and modeling risk and crisis communication. In D. O’Hair & R. Heath (Eds.), Handbook of risk and crisis communication. Mahwah, NJ: Lawrence Erlbaum Associates. Pape, R. A. (2005). Dying to win: Tthe strategic logic of suicide terrorism. New York, NY: Random House. Pollina, D. A., Dollins, A. B., Senter, S. M., Brown, T. E., Pavlidis, I., Levine, J. A., & Ryan, A. (2006). Facial skin surface temperature changes during a "concealed information" test. Annals of Biomedical Engineering, 34(7), 1182-1189. Ruble, N.M., & Turner, W.L. (2000). A systematic analysis of the dynamics and organization of urban street gangs. The American Journal of Family Therapy, 28(2), 117-132. Sageman, M. (2004). Understanding terror networks. Philadelphia, Pa: University of Pennsylvania Press. Schmidt, K. L., & Cohn, J. F. (2001). Human facial expressions as adaptations: Evolutionary perspectives in facial expression research. Yearbook of Physical Anthropology, 116, 8-24. 36 Shoreditch. http://www.telegraph.co.uk/news/main.jhtml?xml=/news/2006/05/09ncctv09.xml&sSheet=/news /2006/05/09/ixuknewsnew. Spears, R., Lea, M., & Lee, S. (1990). De-individuation and group polarization in computermediated communication. British Journal of Social Psychology, 29, 121-134. Virgil, J.D. (2003). Urban violence and street gangs. Annual Review of Anthropology, 32, 225242. von Hippel, K. (2002). The roots of terrorism: Probing the myths. London Political Quarterly 73 (Supplement 1), 25-39. Vrij, A. (2000). Detecting Lies and Deceit: The psychology of lying and implications for professional practice. Chichester: John Wiley & Sons. Walther, J. B. (1996). Computer-mediated communication: Impersonal, interpersonal, and hyperpersonal interaction. Communication Research, 23(1), 1- 43. Yuan, M. (2001) Representing complex geographic phenomena with both object- and field-like properties. Cartography and Geographic Information Science, 28(2), 83-96. Zuckerman, M., & Driver, R. E. (1985). Telling lies: Verbal and nonverbal correlates of deception. In A. W. Siegman & S. Feldstein (Eds.), Multichannel integrations of nonverbal behavior (129-147). Hillsdale, NJ: Lawrence Erlbaum Associates. 37 Appendix A Participants Maruthi Akella Charles Bond Donald Brown Judee K. Burgoon Kathleen Carley Jeffrey Cohn William Crano Andrew Dollins William Donohue Lawrence W. Hunter Matthew Jensen Robert Kissel Arie Kruglanski Jim Lee Position Assistant Professor of Aerospace Engineering and Engineering Mechanics Professor of Psychology Professor and Chair of the Department of Systems and Information Engineering Professor of Communication, Family Studies and Human Development Professor of Computer Science at the Institute for Software Research International Professor of Psychology Professor of Psychology Researcher Distinguished Professor of Communication Researcher Research Associate, Center for the Management of Information Researcher Distinguished University Professor of Psychology Associate Director of the Center for the Management of Information Institution University of Texas, Austin Texas Christian University University of Virginia University of Arizona Carnegie Mellon University University of Pittsburgh/CMU Claremont Graduate University DOD Polygraph Institute Michigan State University Johns Hopkins University Applied Physics Laboratory University of Arizona Naval EOD Technology Division University of Maryland University of Arizona Applied Research Laboratories, The University of Texas at Austin Lawrence Livermore Natl. Labs/UC Berkley ONR Sandia National Labs University of Arizona University of Oklahoma Cheryl Martin Harry Martz Dave Masters Dale Murray Jay Nunamaker H. Dan O'Hair Stergios Papadakis Marc Sageman Jennifer Samples Gary Strangman Research Scientist Center Director for Nondestructive Characterization Program Manager Researcher/Consultant Regents Professor of MIS and Computer Science Presidential Professor of Communication and Director, Institute for Communication Research President, National Communication Association Researcher Sageman Consulting, Washington, DC Manager Director, Neural Systems Group University Distinguished Professor of Engineering Science and Mechanics and Electrical Engineering Senior Staff, Applied Physics Laboratory Professor of Communication and Telecommunication, Information Studies & Media APL/JHU Private Practice; University of Pennsylvania APL/JHU Harvard University University of Arkansas Vasu Varadan Michael Vlahos Joseph B. Walther Johns Hopkins University Michigan State University 38 May Yuan Professor of Geography; Director, Center for Spatial Analysis University of Oklahoma Attendees Peg (Marguerite Barratt) Edward Clancy Mike Dunaway Bruce Hamilton Harold Hawkins Lee Mastrionni Ernest McDuffie Robert Osiander Mike Pestorius Alex Schwarzkof Billy Short Scott Steward Amber L. Story Lora Weiss George Zarur Position Division Director, Behavioral and Cognitive Sciences Program Manager, Industry/University Cooperative Research Centers Program Director, Biochemical Engineering Institution National Science Foundation National Science Foundation ONR National Science Foundation ONR ONR ONR APL/JHU Applied Research Laboratories, University of Texas at Austin National Science Foundation ONR ONR National Science Foundation Pennsylvania State University Department of Homeland Security Research Scientist Research Engineer Program Director, Industry/University Cooperative Research Centers Program Director Associate Professor of Acoustics and Senior Research Associate Program Manager 39 Appendix B FINAL AGENDA: DETECTING AND COUNTERING IEDS JUNE 5, 2006 8:30 – 9:00 am A. Welcome, introductions, and expectations 1. NSF welcome—Marguerite (Peg) Barratt 2. PIs welcome and expectations—Judee Burgoon and Vasu Varadan 3. Group Introductions 4. Use of GroupSystems—Jim Lee 9:00 – 10:15 am B. General Background 1. on Terrorism Marc Sageman, Future of Terrorism Research Michael Vlahos, The IED, Cultural Adaptation, and New War 2. on Explosive Devices and Detection Devices Harry Martz, Defending the Homeland: The Non-Intrusive Inspection Problem and Technologies Robert Kissel, Explosive Ordinance 3. Synergizing the Social and the Technical Judee Burgoon for Tom McGill, Issues for Social Sciences in IED David Masters, Researching the IED Problem: A View from ONR 10:15 –10:30 am Break 10:30 –11:15 am C. Framing the Problem: 1. What are the key issues? 2. What are possible scenarios? Jay Nunamaker & Jim Lee, electronic brainstorming, categorization, outliner and discussion 11:15-12:15 pm D. The Social Psychology of IED Events: Motivations and Recruitment of Actors William Crano, Minority and Intergroup Relations: Relevance For Terror Prevention Arie Kruglanski, Motivational Factors in IEDs E. The Process of IED Events 1. Organization, planning, communication and network dynamics Jay Nunamaker, Social Framework for Studying Improvised Explosive Devices 2. Communication Joe Walther, New and Old Media: Do Internet Connections Exaggerate Extremes, and Traditional Media Confuse? 40 12:15 – 1:30 pm Lunch on your own 1:30 --2:30 pm F. Theme Development Session I 1. Break-out groups each identify a possible research theme based on preceding talks, enter ideas on GroupSystems 2. Report out to group, discuss 2:30 – 3:30 pm G. Surveillance and Detection of Emplantment/Deployment 1. Multli-modal analyses of behavioral indicators of intent Judee Burgoon, Detecting IED Incidents from Behavioral Indicators Matthew Jensen, Automatically Inferring Meaning from Observable Behavior Jeffrey Cohn, Multi-Modal Analysis of Face and Body Gesture Indicators of Communicative Intent and Identity Charles Bond, International Deception 3:30 --3:45 pm Break 3:45 – 4:30 pm 2. Neurophysiological indicators for close surveillance or interrogation Andrew Dollins, Credibility Assessment using Physiological Indices Gary Strangman, Robust Neuromonitoring for Real-World Applications 3. Spatial analysis May Yuan, Spatial Analysis and Modeling Potential for IED Research 4:30 - 5:30 pm H. Theme Development Session II 1. Break-out groups each identify a possible research theme based on preceding talks, enter ideas on GroupSystems 2. Report out to group, discuss Summing up 6:30 – 8:30 pm Group dinner JUNE 6 8:30 – 9:45 am I. Surveillance and Detection of Emplantment/Deployment, cont. 4. Remote sensing Vasu Varadan, Microwave Scanning for IEDS Dale Murray, Bulk Detection of Explosives J. Countering IEDs: Prevention and Mitigation 1. Modeling, tracking and preventing incidents William Donohue, Automatically Inferring Meaning from Observable Behavior Don Brown, Discrete Choice Models for Incident Prediction 41 2. Risk assessment approaches Dan O’Hair, Effects of Influence Networks, Communication Media, and Message Strategies on Decision Making 9:45 – 10:45 am K. Theme development III 1. Break-out groups each identify a possible research theme based on preceding talks, enter ideas on GroupSystems 2. Report out to group, discuss 10:45 -- 11:00 am Break 11:00 – 12:15 pm L. Methodological Issues 1. Fusion of sensors, images, behavioral data Maruthi Akella, Algorithms for Mobile Heterogeneous Sensor Networks with Applications to Improvised Explosive Device Detection and Surveillance 2. Approaches to real-time or near-real time results Kathleen Carley, Dynamic Network Approach to IED Related Research Gary Strangman, Neural Network Fusion and Pattern Identification for Multi-Modal Physiological Data 3 Discussion of methodological issues and possible themes 12:15 – 1:15 pm Working Lunch (ordered in advance) M. ONR projects Lawrence Hunter, Overview of Counter IED R&D at the Johns Hopkins University Applied Physics Laboratory Stergios Papadakis, THz Scanning Source for Active IED Detection, Imaging, and Characterization Jennifer Samples, Nanoparticle Taggants for Explosive Precursors Cheryl Martin, Support for Predicting Improvised Explosive Devices 1:15 – 1:30 Break 1:30 -- 3:20 pm N. Generation of research project/theme ideas for hard problems and high risk/highpayoff research 1. Electronic balloting to prioritize themes 2. Subgroups formed to flesh out themes 3. Presentations to the entire group, critique and electronic feedback 3:20 – 3:30 pm O. Summing Up and Next steps Depart 42 Appendix C Voting Results Rank Order (Allow bypass) Number of ballot items: 11 Total number of voters (N): 17 Number of Votes 1 2 3 4 5 6 7 8 9 10 9 1 2 2 1 0 0 0 1 1 0 3 5 2 1 2 0 2 1 0 1 0 1 2 3 1 3 3 0 3 1 0 0 2 2 2 2 3 2 1 0 2 1 0 1 3 2 2 0 1 4 2 1 1 0 0 0 3 3 1 6 1 1 1 0 1 0 1 1 5 1 1 1 1 2 4 0 0 3 1 0 2 0 3 3 3 2 0 0 0 0 0 1 1 3 3 3 5 1 1 0 1 1 1 1 2 3 3 0 4 Themes for Funding 1. Database 2. Fusion 3. Model Pred & Validation 4. Deception Detection 5. Culture 6. Disruptive Tech/Strategies 7. Social, Psychological Characteristics of Terrorists 8. Training & Skills 9. Communication/Media 10. Human Computer Interaction 11. Technology Transitions 11 0 0 0 0 2 2 0 0 0 2 11 Rank Sum 163 133 128 116 108 98 92 89 86 81 28 Mean STD 2.41 4.18 4.47 5.18 5.65 6.24 6.59 6.76 6.94 7.24 10.35 2.35 2.38 2.50 2.43 3.33 2.63 2.58 2.77 2.70 2.82 1.27 n 17 17 17 17 17 17 17 17 17 17 17 Group consensus (1.00 = most consensus): 0.38 43