Click for next page ( 12


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 11
Part I BACKGROUND AND NEED FOR ORGANIZATIONAL MODELS 

OCR for page 11

OCR for page 11
1 Introduction I n 1951, Isaac Asimov published a science fiction novel, Foundation, that imagined a future world in which a maverick genius, Hari Seldon, invented a new science—psychohistory—that was capable of mathe- matically predicting “the reactions of human conglomerates to fixed social and economic stimuli” (Asimov, 1951, p. 19). In Asimov’s novel, Seldon’s psychohistory equations are used to predict the collapse of a galactic empire, allowing a band of scientists (the Foundation) to act to preserve human knowledge and greatly shorten the period of chaos that follows the galactic collapse. Asimov’s vision has inspired generations of scientists. Today scientists find themselves at the edge of what he imagined—working on computa- tional mathematical models of aggregate human behavior that allow them to understand, assess, and, to a very limited extent, predict “the reactions of human conglomerates.” This report assesses how close they have come to that vision and what still remains to be done. The study was requested by the Human Effectiveness Division of the U.S. Air Force Research Laboratory, with additional funding from the Air Force Office of Scientific Research. The Air Force and the other military services are increasingly interested in using models of the behavior of humans, as individuals and in groups of various kinds and sizes to support the develop- ment of doctrine, strategies, and tactics for dealing with state and nonstate adversaries, in support of military planning and operations, acquisition pro- grams, and as training and simulation tools. In this report, we are calling them individual, organizational, and societal (IOS) models. There are many lines of research on such models, in academia, industry, and the military, and 

OCR for page 11
4 BEHAVIORAL MODELING AND SIMULATION it would be difficult for a military program office staff to become thoroughly familiar with all of them or to evaluate the potential of each research pro- gram for use by the military. The modeling efforts span several disciplines, have different goals, and often use different terminologies. The Air Force therefore asked the National Research Council (NRC) to review several relevant IOS modeling research programs, evaluate the strengths and weaknesses of the programs and their methodologies, deter- mine which have the greatest potential for general military use (i.e., not just Air Force specific), and provide the Air Force with guidance for the design of a research program to effectively foster the development of IOS models useful for the military. One of the great strengths of the NRC is its ability to convene committees of experts from a broad range of disciplines and facilitate their cooperative work on the study of a cross-disciplinary topic like this one. STuDy TASK AND OBjECTIvES The formal statement of task from the cooperative agreement between the NRC and the Air Force for this study is as follows: • Review the state of the art of the subset of the social sciences per- ceived as having the greatest payoff in terms of informing future computational model developments. These will include o key conceptual models in the areas of anthropology, sociology, social psychology, political science, organizational theory, and similar social sciences specialties o efforts in developing computational models, “artificial life” sim- ulations, and the like being undertaken by these communities Review the state of the art in societal1 modeling applications serv- • ing the Department of Defense (DoD) and related agencies, with special emphasis given to computational modeling and simulation- based approaches • Review the state of the art in the three computational modeling communities outside DoD, and identify strengths and shortcomings in each: o cognitive science and individual behavioral modeling o network analysis and multiagent organizational modeling o multiresolution modeling and simulation 1 Inthis study, the committee broadened the scope to include individual and organizational models as well, because of the inseparability of all three, given the intended usage. Additional discussion appears in the Concepts and Definitions section below, as well as in chapters following.

OCR for page 11
5 INTRODUCTION • Identify how gaps in societal behavioral modeling applications serving DoD and related agencies might be filled by: o conceptual models in the social sciences o computational modeling approaches now under way in the social science community o closer linkages—via shared research, common development frameworks, interlinked computational models and the like—between the cognitive science community, the network/ organizational modeling community, and the multiresolution modeling and simulation community • Develop a research and development roadmap to fill current appli- cation gaps, for the near, mid-, and far term2 NATIONAL ACADEMIES’ RESPONSE The NRC, an operating arm of the National Academies, responded by appointing a committee of 13 experts, drawn from the social sciences and from several human behavior modeling communities and disciplines, following the procedures mandated for all NRC committee appointments. These procedures are designed to ensure that committee members are chosen for their expertise, independence, and diversity and that the com- mittee’s membership is balanced and without conflicts of interest. The appointments were finalized after the discussion of sources of potential bias and conflict of interest at the committee’s first meeting in April 2005. Brief biographies of the committee members appear in Appendix D. THE COMMITTEE’S APPROACH The committee developed its approach to the task at the first meeting. We discussed each member’s expertise and identified information needs in several domains, including the military’s needs and uses for IOS modeling, research now under way under military contracts (and often not available in the open literature), and the current state of the art of modeling efforts in the social science and computational modeling communities listed in the task statement. We developed plans for obtaining and analyzing the needed information and for organizing the report. The committee also discussed the scope of its task and determined what would and would not be attempted. 2 In our recommendations, we distinguish actions to be taken in the first year, years 2–4, and beyond. These may be interpreted to correspond to near-, mid-, and far-term horizons.

OCR for page 11
 BEHAVIORAL MODELING AND SIMULATION Defining the Project Scope To achieve the objectives of the study, the committee needed to review the state of the art in several modeling disciplines and communities of prac- tice. We decided that it would be neither feasible nor useful, for the purposes of this study, to produce an exhaustive literature review. Rather we decided to summarize relevant knowledge in each of the modeling areas, and to orga- nize our summary review of each area using a template of significant features developed by the committee. The template focused on the applicability of each type of modeling approach to the DoD’s IOS modeling needs. gathering Data The committee used a variety of data-gathering methods, mainly over the period of the first three meetings. We reviewed pertinent literature, scholarly and applied, including publicly available military documents, such as the Quadrennial Defense Review (U.S. Department of Defense, 2006), with each committee member concentrating on his or her area of expertise. We invited the sponsor and other military experts to brief us on the particulars of DoD’s needs for IOS models and on the expectations of potential model users, and we invited managers of DoD modeling research programs to tell us about their programs. We appointed three military operations experts with some knowledge of IOS modeling as consultants to the committee, enlisting their help in developing representative scenarios of situations in which models might be used by DoD, as one way of under- standing the need for IOS models. Data Analysis and Review In our later meetings, the committee discussed the information we had found, developed a framework for presenting our findings and conclusions, and developed recommendations for the study sponsor. The report structure is straightforward: we discuss DoD’s need for modeling and the current knowledge and capabilities (state of the art) in the modeling community. We then highlight the important gaps between the state of the art and the iden- tified needs and discuss ways to bridge the gap in a research program. Concepts and Definitions Because the field of IOS modeling is spread among several disciplines and domains, the same terms are often used with different meanings by dif- ferent authors. We felt it necessary to agree on common definitions for some important terms and then to use the terms in only the defined senses.

OCR for page 11
 INTRODUCTION First, what should we call this area of modeling that spans the range from individual actors who are members of a small group to whole nations, societies, or ethnic groups? We are using the term “individual/ organizational/societal” modeling to convey this broad meaning, to cover the modeling communities that are concerned with the full span of human behavior, including individuals, teams, small groups, large groups (includ- ing different cultures and ethnic/religious groups), societies, nations, and national coalitions. We are of course not viewing this as merely a one- dimensional progression in the size of the “human conglomerate,” but rather as a rich tapestry of many dimensions that are complexly interlinked via relationships that are only now being recognized, let alone understood. Through the course of this report, we hope to point out some of the key relationships, as well as the considerable distance there is to go in terms of understanding the fundamental interdependencies and interactions that exist, in a manner that supports meaningful and useful models. Second, what should we call these different levels of “knowledge repre- sentation” that start with empirically based observations of human activity3 and end with computational instantiations (specifically, computer-based simulations) of human behavior, often referred to simply as “models”? It is certainly beyond the scope of the committee to develop an ontology of human behavior representation, but we think that it is appropriate to attempt to identify at least four levels that proliferate in the modeling com- munity. Going from the general to the specific, they are • Theory: This is an explanation of how something works, in this case how one of the human conglomerates behaves for a given set of traits (or culture) in a given situation or environment. Theories may be global (e.g., at the individual level, a “unified theory of cog- nition”), or they may be local (e.g., the decay of working memory). Theories may be formal or informal, mathematical or verbal, well formed and founded, ill formed and unfounded, and everything in between. • Architecture: This is a more specific statement of a theory, one that places a structure under it, and attempts to either: (a) break down the theory into smaller and perhaps more readily understood com- ponents or subtheories (e.g., at the individual level again, a “cogni- tive architecture”) or (b) link the theory with collateral theories to explain behavior at a larger scale or in more complex environments (e.g., an architecture to link cultural influences on social networks). “Good” architectures attempt to maintain as much generality as 3 Although not always: some might argue that one starts with internalized theoretical con- structs that shape what one observes, rather than the other way around.

OCR for page 11
 BEHAVIORAL MODELING AND SIMULATION possible (i.e., are parameter-free, domain independent, etc.), so as to be able to accommodate the broadest set of behaviors and situations. • Model: This is a yet more specific representation of the human conglomerate, one that can be directly derived from a correspond- ing theory, or, more indirectly, instantiated from an associated architecture, in which the specific instantiation takes into account, for the entity being modeled, that entity’s inherent characteristics (e.g., personality traits, religious beliefs, social connectivity, etc.), the associated domain-specific knowledge base (e.g., knowing the local village streets), and the specific situation and environment of interest (e.g., crowd formation in a village).4 Like theories, models can be global or local, and well founded or not. • Simulation: This is a still yet more specific representation of the human conglomerate, this time instantiated in executable soft- ware. Simulations can be developed directly from theories (e.g., by coding up, say, a mathematical equation embodying a particular theory), from architectures (by developing a simulation within the software/system development environment associated with a par- ticular architecture, if available), or from models instantiated for a specific situation (giving rise to the term “computational models”).5 The power of a simulation is several-fold: simulated “data” can be compared with empirically collected data for model validation purposes; simulations can be used to explore the range of potential outcomes; and simulations can be used to drive new theory devel- opment and empirical data collection efforts, via the generation of new hypotheses based on simulation-based “experiments.”6 Again, we emphasize that this is not intended to be a definitive ontol- ogy of behavior modeling and simulation, but merely an attempt to clarify terms somewhat, terms that are often used interchangeably in the literature (including, occasionally, this report). 4 Although models can be directly instantiated from theories, there is a trend toward increas- ing use of “intermediate” architectures, driven both by the practical benefits gained by the model developers in being able to instantiate well-grounded models quickly for specific situ- ations and by the lessons learned gained by the architecture developers with each new model instantiation. 5 Again, the trend in the well-established modeling and simulation community is to dis- courage “direct coding” from theory to simulation and instead move through the levels out- lined here, because of the advantages gained from established architectures and model-specific databases (which may be reused), although clearly the development overhead is higher. 6 In addition to the other uses identified in Chapter 2.

OCR for page 11
 INTRODUCTION CAuTIONS FOR IOS MODELINg In our discussions with military personnel, and in interactions outside the committee deliberations, the committee became aware that many people may have unrealistic expectations of what a model or simulation of human behavior is able to do. No model is ever likely to be able to predict exactly what an individual or group will do, except in a situation so constrained, with alternatives so well understood, that a model is not needed. Human behavior, individually and in groups, is governed by so many variables, including many that are not likely to be susceptible to capture in a model, that the best any model will do is to narrow the range of plausible behav- ioral outcomes of a defined situation. For example, a model may be able to forecast the most likely range of outcomes of a potential course of action. It may be able to direct attention to situational variables that are known to be important but may have been overlooked in a particular engagement. A well-designed model may draw a decision maker’s attention to possible unintended consequences (“second-order effects”) of a planned course of action. But it will not be able to make point predictions, such as “If we take Action A, the adversary will attack at Point B early tomorrow morning with three simultaneous improvised explosive devices (IEDs).” So we speak of models forecasting a range of outcomes, rather than making precise predic- tions. Certainly models that can produce such forecasts are a worthwhile objective. They can serve many useful purposes, from supporting training, to serving as tactical decision aids, to examining possible outcomes of alter- native strategies or policies. Some of the known difficulties of developing and implementing models are discussed later in the report, but a few may bear mention at this point. The most desirable data to put into a model that would provide the most accurate forecasts often will not be available: the data may not be acces- sible, may not be in a usable form, or may not be verifiably accurate, timely, or complete. In fact, it is common knowledge that adversaries will often attempt to provide false data (disinformation) if they think it will be believed and used. So the development of a model, in itself, is only a small part of the work that must be done to use it, and there is never a guarantee that good information will be available to implement the model when it is needed. These issues must be taken into account in the design of models. The work of developing models of adversary behavior is never com- plete, because any worthy adversary, once it realizes that its modus operandi is known and defenses are being used against it, will make changes in its organization, operations, etc., designed to invalidate the model. So we have the ever-changing methods used by insurgencies to attack friendly

OCR for page 11
0 BEHAVIORAL MODELING AND SIMULATION forces7 or innocent citizens in the Middle East: car bombs, IEDs, suicide bombers, each adapted or supplanted by a different method as soon as effective countermeasures are devised. This means that a modeling effort must include ongoing maintenance and updating functions if it is to remain useful. Another challenge is that some of the research on modeling for military purposes must necessarily be conducted at high security levels, in secure environments. It is likely that much of the fundamental research for the design of modeling methods and tools can be done in open venues by researchers with low or no security clearances, but any work that includes specific and current field information on individuals or groups, specifics on friendly or adversary force capabilities, or detailed operational plans must of necessity be highly classified to prevent the adversary anticipation, adaptation, and/or exploitation discussed above. Finally, it is important to recall that the predecessor report by the NRC in this area (National Research Council, 1998, p. 8) noted that “the mod- eling of cognition and action by individuals and groups is quite possibly the most difficult task humans have yet undertaken. Developments in this area are still in their infancy.” This situation has not changed significantly in the mere 10 years since the publication of that report. But the world has, and, as a result, it has become ever more clear that human behavioral modeling at all levels is critical to DoD specifically and to the nation more generally. ORgANIzATION OF THE REPORT The report is organized into three parts. Part I provides background information and explains the need for organizational models. Chapter 1 gives the background of the study and the committee’s approach to the work. Chapter 2 discusses evolving missions of the military and the appli- cability of IOS modeling to those missions. It includes an introduction to a set of military scenarios that are used throughout the report as exemplars of situations that could benefit from the use of modeling. Part II contains extensive descriptions of the major modeling method- ologies and model types the committee reviewed. Models take many forms, ranging from loose conceptual models to precise mathematical models (Lave and March, 1975). They include agent-based models, cognitive models, expert systems, dynamical systems, and input-output models. The diverse expertise of the committee members contributed greatly to the complete- 7 The term “friendly forces” is used to refer to forces that are either formally or informally allied with the United States and that support its objectives. It may thus refer to the armed forces of allied nations or to forces representing nonstate organizations or factions.

OCR for page 11
 INTRODUCTION ness of this review but also made it challenging to agree on an organizing framework for presenting the review results. Refined through multiple itera- tions, the organizing framework that we developed represents a significant product of the study, as discussed in the introduction to Part II. Chapter 3 presents conceptual and cultural (verbal) models. The sub- sequent model descriptions are then organized according to the level of granularity of the models. We have differentiated “macro” models that describe organizations and their behaviors on a large scale (Chapter 4); “micro” models dealing on a level as detailed as the individual actors within groups or organizations (Chapter 5); and “meso” or intermediate models somewhere between these two, as well as integrated, multilevel models (both in Chapter 6). We discuss games separately in Chapter 7 because although they incorporate formal models, they do not easily fit into any of the other categories. For each methodology, we describe the method and its current state of development, often with some history of the field. We review the appli- cability of the methodology to the modeling requirements identified in Chapter 2, its major limitations, issues of data, verification and validation, and needs for continued research and development. The discussions of models and methodologies are not exhaustive. We have attempted to provide an overview of a broad range of model types and modeling methods, although the committee members, chosen for their range of modeling expertise, naturally discussed in greatest depth the areas with which they are most familiar. In Chapter 8 we discuss some generic issues, such as integration across levels of models, modeling frameworks and tools, model verification and validation, and data sources and quality. Chapter 9 summarizes the state of the art of IOS modeling as presented in Part II and its utility for the applications discussed in Part I. In Part III we identify the gaps between the current modeling capabilities and the military’s modeling needs, and, in Chapter 10, we discuss common problems or pitfalls that may impede the development and application of models or reduce their utility. In Chapter 11 we present recommendations for a research roadmap, a program of use-inspired IOS modeling research and development designed to reduce the gaps and develop the needed capabilities. The report ends with four appendixes. Appendix A provides a list of acronyms and abbreviations used in the report, spelled out, with some infor- mation on their meanings. Appendix B contains detailed military scenarios that served as exemplars for considering how models could be used for military purposes. Appendix C provides detailed material relevant to the discussion in Chapter 8 of DIME/PMESII modeling paradigms. Appendix D provides biographical sketches of committee members and staff.

OCR for page 11
 BEHAVIORAL MODELING AND SIMULATION REFERENCES Asimov, I. (1951). Foundation. New York: Random House. Lave, J., and March, J.G. (1975). An introduction to models in the social sciences. New York: Harper & Row. National Research Council. (1998). Modeling human and organizational behavior: Appli- cation to military simulations. R.W. Pew and A.S. Mavor (Eds.). Panel on Modeling Human Behavior and Command Decision Making: Representations for Military Simula- tions. Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press. U.S. Department of Defense. (2006). Quadrennial defense review report. Washington, DC: Author.