Unconventional Challenges in the Human Domain
The Human Domain: The Unconventional Collection and Analysis Challenge Workshop.
Author | Editor: Canna, S. & Popp, R. (NSI, Inc).
One of the major challenges the US has faced during the opening decades of the 21st Century is the growth in the number and international reach of insurgents, terrorists, WMD proliferators and other sorts of nefarious groups. Among many others, the current situation in Iraq illustrates the difficulty of countering insurgencies once they have become violent and operationally well-established. The apparent transition of contemporary war away from the traditional physical battlefield towards the human element as the battlefield poses new challenges for both intelligence and operations – collections and analysis within the human domain requires new thinking and new methods.
The purpose of this workshop was to identify approaches to recognize and anticipate the conditions in which insurgents, terrorists, WMD proliferators and other sorts of nefarious groups tend to emerge, become operationally established, and choose to employ tactics counter to the interests of the US national security objectives. An underlying presumption of this workshop is that tackling these movements during their emergent phases is less costly in resources and lives, with more non-lethal options available for consideration, than after they have become fully established. Novel approaches to collections and analysis of the interactions within and between elements of the “human dimension” may lead us to greater potential for proactive rather than reactive capabilities, and certainly non-kinetic capabilities, not only within the military context, but within all components of national power.
Clearly, focusing on emergent violent groups and their activities will necessitate a shift in how the nation and its allies collects, analyzes, and acts upon information from and about these volatile environments. The workshop brought together a diverse group of physical and social scientists, academics, analysts, intelligence collection managers, operational planners, and decision makers to discuss novel techniques and methods for a more proactive approach to address the human dimension associated with insurgents, terrorists, WMD proliferators and other sorts of nefarious groups, and the impact of such an approach on training, planning, tasking, collection, analysis, and decision making in the coming decades.
EVOLUTION OF VIOLENT EXTREMISM
There is an important distinction between activism and radicalism. Activism is legal and nonviolent political action, while radicalism is illegal and violent political action. Activism is not necessarily a conveyor belt to radicalism. Individuals can be radicalized from any layer of society including neutral parties, sympathizers, and activists.
Government Sponsored Research on Violent Extremism
The Department of Homeland Security (DHS) is is laying the groundwork to conduct in-depth case studies of individuals engaged in terrorist violence in the US. At the group level, DHS is studying terrorist group rhetoric with a focus on both groups that have engaged in terrorism and groups that would be considered radical but that have not engaged in terrorism. The study will look at groups with similar ideologies and goals and try to understand why some groups choose violence and others do not. DHS will also work to identify characteristics of groups who participate in IED attacks. At the community level, DHS is involved in two primary activities. The first is evaluating whether survey data can shed light on radical activity in the US. Finally, DHS is working with START to study the effectiveness of counter radicalization programs conducted in five countries including Yemen, Indonesia, Columbia, Northern Ireland, and Saudi Arabia.
The Behavioral Analysis Unit is the newest organization within the Federal Bureau of Investigation. It is an operational unit that conducts research. Counterterrorism research within the unit is still in its infancy stages; however, some projects are currently underway dealing primarily with operational threat assessments. The most pertinent research efforts are in collaboration with DoD intelligence units. This work primarily focuses on radicalization by looking at the radicalization process in historic case studies. The FBI is also involved in another research effort in coordination with the WMD directorate, to look at 50 domestic cases in which offenders used WMD in the US.
The Defense Intelligence Socio-Cultural Dynamics Working Group (SCDWG) was created in 2002 to develop an enterprise solution to institutionalize Socio-Cultural Dynamics intelligence analysis. The group was asked to address the full scope of issues, from policy to tactical. The group was asked to define the role of intelligence in this area as well as to identify the “right tool for the right job.”
The Marine Corps Intelligence Activity (MCIA) is working on the next generation of cultural intelligence. They are working on visualizing cultural intelligence by “putting culture on the map.” Their second project is to support influence operations and irregular warfare by building partner capacities focused on the values and themes of sub-national groups. The third is cultural vignettes, which are short and focused products designed to create a specific cognitive map for each mission set. The fourth is building an analytic capacity by supporting uniquely trained cultural analysts, ensuring 24/7 reachback for deployed unites, and creating deployable analytic teams for contingency operations.
UNDERSTANDING SOCIO-CULTURAL DYNAMICS: TRADECRAFT AND OBSERVATIONS FROM A DISTANCE
Studies to understand a country and its people must be planned and carried out with nuance. Ethnic maps, a critical tool for US forces, have long been a political issue. It affects resource allocation, boundaries, and political formulae. In countries experiencing ethnic conflict, building a credible demographic baseline must be done before lasting geopolitical stability can be realized. Maps can help researchers and analysts understand the cultural geography of conflict. Chorological analysis, relating to the description or mapping of a region, can reveal causal linkages and patterns of political and social behavior that researchers were previously unaware of. Mapping the human terrain is needed to understand these complex problems.
Culture at a distance methodology from 1942 mostly comprised of several sources. Researchers studied travelers’ accounts, ethnographies, histories and other second hand sources including social science works. Content analysis finds its roots in these attempts. Post World War II, the DoD funded a great deal of social science research although only a small part of it was “culture at a distance”. They focused instead on models such as game theory. Their main interest was in political instability: its causes and indicators.
Today, the DoD has stated that socio-cultural data collection is everyone’s problem. However, the DoD does not currently have the knowledge and understanding of non-Western cultures and societies necessary to execute these missions. Because of this, there are several new culture at a distance methods. Computational modeling has become the sine qua non of culture at a distance method, but it suffers from considerable handicaps. Its emphasis on computational engineering leads to “everything looks like a nail” thinking. The problem with models is that many do not scale gracefully from explaining small scale issues to larger scale ones. Models are culturally and socially ignorant, which is frequently the case when engineers develop “new cultural theory” with little or no training in the fields of social science.
Socio-Cultural Tradecraft and Open Source Collection Requirements Management
Emerging violent non-state actors use or share a common story about their involvement to motivate and empower their groups, and attract and mobilize their audiences. The degree by which infringements are interpreted, personified, and perceived is the variable to which violence emerges. It is typically a conflict between the haves and the have not’s over a particular issue. Modeling these relationships requires a four prong approach: anthropological research; open source collection; human intelligence collection; and environmental research. The model will synthesize that information to assess behavior.
Cultural geography is the study of spatial variations among cultural groups and the spatial functioning of society. Urban geography is the study of how cities function, their internal systems and structure, and the external influences on them and the study of the variation among cities and their internal and external relationships. The human geography areas of interest include economics, demographics, politics, culture, combined social implications, and technology/infrastructure. Cultural geographers face several challenges. First, experiences in Afghanistan and Iraq have shown that neighborhood level data is necessary for operational and tactical planning (Intelligence Preparation of Battlefield). Second, there is a lack of reliable data sets. For example, there has been no census in Lebanon since 1932. Third, closed countries, such as North Korea, pose a problem. Fourth, no city level data exists or it is difficult to obtain in grey literature. Fifth, there is a lack of current academic studies. Therefore, researchers must take a multidiscipline approach and use sources from a variety of fields to complete their research.
Remote Observing Using Large Data Sets
The nexus for computer forensics is addressing large data sets. Researchers and developers are working with open source information and applying analytic tools to it. They have discovered that it is considered novel in the US to use the same tool to address open source and classified datasets and merge them. One large data set, INSPIRE, was developed to allow an analyst to look at a large number of text documents and look for words and phrases. The program graphically portrays the output. A subsequent project is porting INSPIRE over to high power computing environment.
Content analyzers are another tool for working with very large datasets. The tool makes coding schemes in the process of creating large data sets from large text sources. It has an automatic generation process that is transparent. From this data, the analyst produces dashboard like metrics to describe the data. This has created an impetus to create measures of effectiveness and validation. The team discovered that people and computer coders make different types of errors. Humans make boredom errors. Their accuracy always goes down over time. However, if a computer sees a new word, it will not code it. The operation code measures how you present self and others in text on cooperative or conflictive dimensions. The tool can also extract and compare entities for distrustfulness. They use a combination of tools to look for themes and people. It is a pragmatic way to build structure while reducing the analysts’ reading load.
UNDERSTANDING THE HUMAN TERRAIN: TRADECRAFT AND STREETCRAFT METHODS
Applying Law Enforcement Methods to Gathering Human Terrain
An appropriate balance of skills, knowledge, and abilities need to be incorporated into an analysis team. It needs to be half Army Ranger company and half artist colony. Good analysts are adaptable, intellectually flexible, and have a high degree of comfort with ambiguity. An analyst’s job is simple: they read, write, and present. In the development of a good analyst, no preparation matches analyst’s comfort level with ambiguity. Furthermore, the depth and breadth of an analyst’s rolodex is more important than personal understanding of a region or topic. Analysts must also exhibit finesse – the practice of combining knowledge, thought, and intellectual stamina. Personal conviction of paying attention to a few small details that have enormous impact on outcomes or intelligence is critical. Good analysts are those who care as much about operational environment as producing a final product.
Approximately 90 percent of the food we eat, clothes we wear, and products we use are imported daily via commercial maritime transport. Threats to ships and ports include drug smuggling, stowaways, and terrorists. Drugs are often creatively hidden in cargo, such as cocaine stashed inside individual bananas. Drug couriers can also be hidden away in cargo shipments. These could easily be terrorists or terrorist materials. According to a US Coast Guard report, 25 Islamic extremists entered the US onboard commercial cargo vessels. Narco traffickers are more frequently taking consignments of extremists to the US. There is an increasing level of cooperation between MS-13, narco traffickers, and Islamists. The human collection component is critical to stopping these illegal activities because Customs cannot provide the level of analysis needed.
The Department of Commerce’s task is to prevent the export of US dual-use goods and technology that may be used by rogue states or terrorists to make chemical, biological or nuclear weapons. The Office of Export Enforcement’s priorities are WMD proliferation, terrorism/terrorist support, and unauthorized military/government use. Commerce is unique in that it has civil penalties, in addition to criminal penalties, in which one only needs the preponderance of the evidence to convict. The challenge of identifying and stopping illegal dual use exports is making sure at every step of the process that all of the people involved can validate the end use. It gets murky because end users lie and some components are working with the end users. Globalization is also a national security challenge and many transactions are moving over the internet. The US government does not move at the speed of business.
The Criminal Investigation Task Force (CITF) was created in early 2002 by the DoD to conduct investigations of detainees captured in the Global War on Terrorism. CIFT became the investigative arm of the DoD after OSD was given the mission to prosecute the detainees. OSD transferred the responsibility to the Army, where the CITF now falls under Army Criminal Investigation Command (CID). The need for analysts and intelligence professionals was huge, especially for law enforcement officials not used to working with classified materials. Likewise, law enforcement people need to communicate with the combatant commanders. Terrorists groups often act like organized crime groups. The intelligence community may not have much experience in this, but law enforcement does. It also has experience in conducting interrogations. To conduct interviews effectively the interrogator must put all biases aside and just accept the culture for what it is. The interrogator must get down to the bedrock of cultural foundation.
Applying Social Science Methods to Understanding the Human Terrain
Social science is the application of consistent, rigorous methods of research and analysis to describe or explain social life. Its purpose is inference: using facts we do know (data, observations) to learn about facts we do not know (theories, hypotheses). It is used to make descriptive inferences (semantic program); to place or make observations within a conceptual framework to allow people to understand a phenomenon; and to make causal inferences (syntactic / pragmatic program). Because of the large number of potential causal factors for any social phenomenon, establishing causation is difficult. If we can establish the cause of a social phenomenon, then we can make policy to affect that phenomenon.
Data collection involves various methodologies such as ethnography: participant observation; surveys: interview processes (formal, informal, group, individual, structured, unstructured); record or document review; and history: conducting interviews or reviewing documents about the past. Data analysis involves direct interpretation: analysis by an individual’s reflection and synthesis; quantitative analysis: using standard methods of statistics to ascertain relationships; and formal modeling: analysis by creating a formal system that mimics the world. In evaluating which method is most appropriate, it is important to keep in mind that multidisciplinary methods reinforce one another.
Field research in Iraq is particularly complicated. Researchers must use semantic programs and descriptive inference as well as create a pragmatic program with causal inference. Some of the challenges include sampling, reactive bias, and interpretive bias. Coping with these problems is done by working through local researchers, geographic sampling and purposive sampling, redundancy of methods, and redundancy of research networks. On the ground analysis is informed by area studies, topic specialization, and occasional forays into population.
There is a great need to improve intelligence support to stability operations. To do this, analysts must embrace the importance of understanding the human terrain. There are three key elements to the human terrain. The first is developing a socio-cultural dynamics data network and repository. The second is developing data gathering, visualization, and analytic tools. The third is recruiting, training, and deploying experts to support decision-making.
To understand counterinsurgency, you must first understand and define the elements involved. Collecting information for COIN operations involves identifying key groups in a society and representing where they are located on a map. For each group, researchers seek to identify security (level, sources, threats); income and services (level, sources, gaps); beliefs and communications systems (narratives, symbols, norms and sanctions); and authority structures & figures (identity, structures, levels of authority).
Soldiers as Human Terrain Sensors
The Civil Affairs (CA) information objectives were to provide multi-dimensional situational understanding, provide situational analysis, and to work on the development of possible solution sets. The foci of CA operations are human terrain identification and prioritization (sphere of influence management) and civil reconnaissance. Within the human terrain network, there are sphere of influence engagements. These works identified political/tribal leaders throughout the district and their areas of influence to facilitate reconciliation efforts. They also identified facility managers/public works workers to facilitate transition to Iraqis fixing Iraqi problems. They also prioritized efforts to optimize limited resources for engagement.
There are three critical nodes in the CA analysis process. The first is collection. The information is the critical base element to the process. Without this, there is no system. The second is consolidation. Analysis and understanding the information makes the process function. It is an art and a science. The third is dissemination. Sharing the situational understanding is the key to success. Dissemination must flow up and down in order to coordinate and integrate with partners.
Provincial Reconstruction Teams (PRTs) have become a key tool for the international community to assist Afghanistan in becoming a secure and self-sustaining Republic. They represent, at the local level, the combined will of the international community to help the government and the civil society of Afghanistan become more stable and prosperous. PRTs, due to their provincial focus and civil-military resources, have wide latitude to accomplish their mission of extending the authority of the Government by improving security, supporting good governance and enabling economic development. This engagement of diplomatic, military and economic power by Nations at the provincial level allied with the wide latitude to accomplish their mission has been a strength, as it provides flexibility of approach and resources to support the provincial government structures and improve security.
The Humanitarian Information Unit (HIU) serves as a United States Government (USG) interagency center to identify, collect, analyze and disseminate unclassified information critical to USG decision makers and partners in preparation for and response to humanitarian emergencies worldwide, and to promote best practices for humanitarian information management.
Civilians as Human Terrain Sensors
In order to operate effectively in Afghanistan, warfighters need to understand the cultural context and history of the nation. An anthropology background by itself is not always useful. Human Terrain Teams (HTTs) and others need culture specific knowledge to achieve their mission in foreign countries. An open source portal exists to provide a civilian based source of socio-cultural information on Afghanistan. The portal facilitates a subject matter expert network to encourage interaction among the key players including the Department of State (DoS), non-governmental organizations (NGOs), the US Agency for International Development (USAID), Brigade Combat Teams (BCTs), and soldiers. The site drills down to the provincial level where 21 of 34 provinces have been completed. The site provides information on detailed maps, refugees, education, health, topographic information, etc.
SUPPORTING ANALYTIC DISCIPLINES
Modeling and Visualization Workshop Summary
The objective of the Social Science Modeling and Visualization Workshop, held in January 2008, was to provide a “mixing bowl” for social scientists, modelers, researchers and government stakeholders to discuss the state-of-the-art in methods/models/visualization and their potential application in SMA efforts. Models can clarify complex situations, test assumptions, aid decision making, explore co-evolutionary motivations and bound the expected and outlier behaviors. Modeling can capture dynamical, non-linear processes. They can be multi-level, multi-scale behavior representations. Other benefits include that models can assist in training decision makers on the consequences of their policies/decisions. Visualization is a critical component of modeling. Therefore, it needs to be considered from the beginning of a project. Visualization serves a variety of functions include support of knowledge management, contextual visualization, relational visualization, identification of key features/nodes/information, understanding of decisions/complex processes, and evaluation/inspection of the data. Models must be verified, valid, and credible. The theoretic framework of social science modeling and visualization still needs work. Opportunities and challenges include the need for better, scientifically grounded theories. The federal government is interested in social science, but struggling to determine how to support it. It will take 4-5 years to reach robust funding levels.
Evolutionary Agent-based Modeling and Game Theoretic Simulations
Agent Based Modeling (ABM) is successful in the social science arena because as in the traditional approach to formal modeling, the goal is reducing the social landscape to a set of meaningful variables and specific mathematical equations. ABM brings to social science the ability to begin with a social landscape of entities in the world. It can identify basic relationships between actors. All models are developed or designed to answer a class of questions. ABMs are typically designed to answer questions that are intractable through earlier and more traditional mathematical or statistical methods, such as complex emergent patterns in large systems of agents. The beauty of ABM is that it can conduct experiments about shifting policies. All this can be done in silicon because you cannot manipulate players in the real world. ABM is a specific kind of simulation. It provides an experimental setting for finding patterns that are not very obvious.
Game theory is a branch of applied mathematics that formally models strategic behavior. Game theory is an approach to aid in articulating and understanding important factors of conflicts / disputes / coalitions. Game theory is neither prescriptive nor descriptive; nor is it normative. It is a theoretical tool. Game theory does not provide an answer; it provides a helpful way to frame the question to better understand the situation. If the assumptions are correct, game theory will provide an answer. Some assumptions of the theory include (bounded) rationality, complete information, and preference hierarchy for players (utility). In conclusion, it is important to remember that people are not agents. Game theory requires assumptions about behavior. It provides insight and understanding, but people are more complex than mathematical models. Finally, as in any model, if you put garbage in, you get garbage out.
Data Fusion / Integration and Detecting Patterns in Heterogeneous Data Sets
The defense and intelligence communities are making great progress in the social sciences. Many agencies now realize the importance of social science approaches, in combination with more traditional approaches. Often, people want to know why if we can go to the moon we cannot accurately model social science. It is because people are complex intelligent agents with feedback. They have multiple incentives, multiple allegiances, and multiple groups. Humans also exhibit adversarial behavior. They provide inaccurate and misleading information. Behavior is not invariant. There is a hierarchy of challenges in data fusion. The first is registration and entity resolution. The second is group and network detection. The third is feature construction and recognition. The fourth is complex event detection. The fifth is that concepts in models are inherently fuzzy. There is a prediction versus risk relation. And risk brings additional scrutiny and preparation challenges.
Data fusion technologies encompass a variety of characteristics. It is historically a deductively-based inferencing / estimation process. The resulting observations and estimates produce random variables optimized for minimum uncertainty. The process is adaptive in various ways. It enables observation management, process adaptation, and beneficial actions. Data fusion models can be used to analyze a broad set of areas. The physical domain is the easiest to accomplish fusion. This includes weapon systems and weather reports. The information domain is harder. This includes symbolic representation and interpretations of data and models. The cognitive domain includes beliefs, values, and emotions and is the most difficult to fuse.
Efforts are underway to create an automated multi-source, multi-sensor, multi-data fusion tool. One way they are doing this is by looking at textual information. They take documents of different types, use ABA tools and extract antecedents. The tools can grab information, bring it together, and end up with predictive models. Working toward multi-focused analysis, the tool must overcome the problem of sorting through tons of information.
Large Data Sets and Knowledge Extraction Techniques
Large data and visualization efforts are becoming an ever growing avalanche within the social science community. Loading all the data is hopeless. Therefore, efforts have focused on trying to load the structure of the data. This results in maximum expressiveness with minimum size. The structure of the data for documents includes terms, entities, and concepts. The structure for transaction, travel, and communication records include people, places, organizations, and relationships. However, smart loading still requires human guidance. A sandbox is an environment for “what if” questions and for constructing arguments. It should have quick access to exploratory tools and answer the question, “How are these actors connected?” The system’s job is to track workflow and allow for annotations and assumptions. The analyst can backtrack and replay, branch, and recombine. He may also retract or change assumptions. The output should embody a complete chain of reasoning.
Work is being undertaken to determine whether researchers could construct a scaffold for the data that can say something about the dataset. The geometry in the data set might help pull out information. By comparison if you look at physics, the math needed to model it is very tame math. In biology, the math needed is fundamentally different. Even though researchers have not discovered the laws of social science, there are some regularities, although they have not been captured in math to a sufficient depth.
The scale and speed of the data means that the classic, craftsman approach to machine learning is no longer possible. The answer is a commodity, hands-on model. This is a challenging task. Not only does commodity mean that researchers have to develop algorithms that are robust in the face of noise, skew, and the host of other ills; it also means that they have to work out how to set all algorithm parameters from the data or the context, so that the human analyst need not concern themselves with them. There is no single algorithmic magic bullet. There are lots of pattern recognition algorithms; some are well known, including decision trees, support vector machines, neural nets, nearest neighbors, naive Bayes. Ensembles are the first and the most powerful of the techniques to know about, as they permit you to squeeze all possible accuracy out of your data.
Download Publication
Comments