Harrison Complexity in World Politics

Complexity
in World
Politics
Concepts and
Methods of a
New Paradigm
Edited by
Neil E.
Harrison
COMPLEXITY IN WORLD POLITICS
SUNY series in Global Politics
James N. Rosenau, editor

COMPLEXITY
IN
WORLD POLITICS
Concepts and Methods of a New Paradigm
Edited by
Neil E. Harrison
State University of New York Press

Published by
State University of New York Press, Albany
© 2006 State University of New York
All rights reserved
Printed in the United States of America
No part of this book may be used or reproduced in any manner whatsoever

without written permission. No part of this book may be stored in a retrieval system
or transmitted in any form or by any means including electronic, electrostatic,
magnetic tape, mechanical, photocopying, recording, or otherwise
without the prior permission in writing of the publisher.
For information, address State University of New York Press,

194 Washington Avenue, Suite 305, Albany, NY 12210-2384
Production by Judith Block

Marketing by Michael Campochiaro
Library of Congress Cataloging-in-Publication Data
Complexity in world politics : concepts and methods of a new paradigm / edited by

Neil E. Harrison.
p. cm. — (SUNY series in global politics)
Includes bibliographical references and index.
ISBN 0-7914-6807-0 (hardcover : alk. paper)
1. International relations—Philosophy. 2. International relations—Methodology.
3. Complexity (Philosophy) I. Harrison, Neil E., 1949– II. Series.
JZ1305.C657 2006
327.1'01—dc22
2005024118
ISBN-13: 978-0-7914-6807-4 (hardcover : alk. paper)
10 9 8 7 6 5 4 3 2 1
Contents
1 Thinking About the World We Make 1

Neil E. Harrison
2 Complexity Is More Than Systems Theory 25

Neil E. Harrison with J. David Singer
3 Complexity and Conflict Resolution 43

Dennis J. D. Sandole
4 Understanding and Coping with Ethnic Conflict 73

and Development Issues in Post-Soviet Eurasia
Walter C. Clemens, Jr.
5 Beyond Regime Theory: Complex Adaptation 95

and the Ozone Depletion Regime
Matthew J. Hoffmann
6 Agent-Based Models in the Study of Ethnic Norms 121

and Violence
Ravi Bhavnani
7 Alternative Uses of Simulation 137

Robert Axelrod
8 Signifying Nothing? What Complex Systems Theory 143

Can and Cannot Tell Us about Global Politics
David C. Earnest and James N. Rosenau
v
vi CONTENTS
9 When Worlds Collide: Reflections on the 165

Credible Uses of Agent-Based Models in
International and Global Studies
Desmond Saunders-Newton
10 Complex Systems and the Practice of World Politics 183

Neil E. Harrison
Contributors 197
List of Titles, SUNY series in Global Politics 201
Index 205
CHAPTER 1
Thinking About the World We Make
Neil E. Harrison
Despite nearly a hundred years of theorizing, scholars and practitioners alike are
constantly surprised by international and global political events. The abrupt end
of the much-studied Cold War was widely unanticipated, as were the conse-
quences of the collapse of communism in Europe. The defining characteristics
of four decades of international politics were erased in a few short years, but the
globalization of economic and social life has continued. The 1997 Asian finance
crisis rattled the US and European stock markets, civic strife in Venezuela influ-
ences the price of oil, and the needs of AIDS patients in South Africa challenge
international agreements on intellectual property. Out of the blue, terrorists at-
tacked within the United States one sunny September morning. A year earlier,
in the space of a few months the global economy lurched from rapid expansion
to recession and flirted with deflation.
After so much ink has been spilt, we still know so little about international
relations and world politics that events continue to surprise us. There is no agree-
ment on the cause of this failure. Some believe that international theorists think
too small and fail to synthesize relevant insights from a range of disciplines (Buzan
and Little 2001); others criticize the emphasis on positivist methods (Smith,
Booth, and Zalewski 1996); and postmodern scholars reject the ahistorical, ratio-
nalist foundations of most international theory (Der Derian and Shapiro 1989;
George and Campbell 1990). This book takes a different tack. It argues that the
reality of world politics is more complex than dreamt of in current theories.
Current theories of world politics assume that the social world is appro-
priately modeled as a simple system; this book proposes that it should instead be
viewed as a complex system. In this book my colleagues and I describe, and
demonstrate the benefits of, a paradigm of system emergence from complex
1
2 NEIL E. HARRISON
agent interactions that we call “complexity”. The study of complexity in systems

is “complexity science” and descriptive, explanatory, or predictive theories—
formal statements that generate empirically testable hypotheses—based in com-
plexity ideas and concepts are “complex systems theories.”1
Like realism, complexity is a thought pattern, set of beliefs, or ideological
orientation about the essence of political reality that organizes theorizing about
and empirically investigating events in world politics. Realism assumes that es-
sential human characteristics drive political behavior within fixed structures;
complexity views politics as emerging from interactions among interdependent
but individual agents within evolving institutional formations. So world politics
is a more or less self-organizing complex system in which macroproperties
emerge from microinteractions. This and the next chapter outline a taxonomy of
the central ideas and concepts of a complexity paradigm of world politics from
which useful theories or models of complex world politics may be constructed.
This ontological shift from simple to complex systems opens new paths to
knowledge and understanding yet incorporates much current knowledge; it val-
idates novel research methods; and theories founded in this approach will gen-
erate radically different solutions to policy problems. In the next section, I
compare basic concepts of simple and complex systems and thereby frame a com-
plexity paradigm. Following that, I show how complexity concepts can be used in
theories of world politics. In the final section of this chapter I outline the rest of
the book.
FROM SIMPLE TO COMPLEX
A system is a portion of the universe within a defined boundary, outside of

which lies an environment. An atom is a system, as is an animal or a country.
Usually, the definition of the boundary is a convenience used to assist human
analysis, as when scientists define for study an individual ecosystem. A pond is
only arbitrarily separated from its shoreline, the air, and the Sun. Similarly, a de-
finition of “country” may be in terms of its recognized sovereign territory, its ter-
rain and ecosystems, its economy (where the distinction between gross national
and gross domestic product is important), or its state or government.
A system is simple if the units and their relations are relatively fixed, per-
mitting reasonable prediction of future system states. An automobile may be
complicated, but it is a simple system. Each of the parts has a specific role in the
system, and the actions of all the parts are centrally coordinated toward a collec-
tive outcome. The existence of workshop manuals further illustrates the sim-
plicity of the system: they identify all potential problems and explain how to
THINKING ABOUT THE WORLD WE MAKE 3
remedy them. They also illustrate problems and solutions, define the character-
istics of each part, and the range of relations between them in exhaustive detail.
As table 1.1 shows, a living system is complex in many ways that an auto-
mobile is not. The two primary differences between complex and simple systems
are diversity and decentralization. In an automobile there are many diverse parts
constructed for very specialized roles, but there is centralized coordination of
their operations through mechanical or electronic management systems. In liv-
ing systems, not only are the units diverse but each has a range of freedom of
choice denied to parts in a mechanical system. Because units in a complex system
have discretion in their choice of behavior, they are commonly called “agents.”
Decentralized decision-making increases complexity. One measure of com-
plexity is the length of the shortest possible message that fully describes the sys-
tem (Gell-Mann 1994, 30–38). Description of a jaguar in the jungle is longer
than of a quark (a unit within an atom). If all the units of a system are identical,
system description is shorter; only one unit need be described in detail. Thus,
heterogeneity among the units increases description length.2 But if the units also
have behavioral discretion, system description requires description of the units
(perhaps by class), of the range of their available choices, and of the rules of be-
havior that each will follow in making their individual choices.
Centralization of decision-making simplifies complicated systems. Mod-
ern automobiles have sophisticated management systems that use miniature
TABLE 1.1. CHARACTERISTICS OF SIMPLE AND COMPLEX SYSTEMS

Simple Systems Complex Systems
Few agents Many agents

Few interactions Many interactions
Centralized decision-making Decentralized decision-making
Decomposable Irreducible
Closed system Open system
Static Dynamic
Tend to equilibrium Dissipative
Few feedback loops Many feedback loops
Predictable outcomes Surprising outcomes
Examples: Examples:
Pendulum Immune systems
Bicycle Genes
Engine Molecules in air
Boyle’s law Ecosystems
Gravitational system Markets
4 NEIL E. HARRISON
computers to govern feedback cycles and responses to environmental changes.

Although these systems respond almost instantaneously to multiple indicators,
there is only a single programmed response to any change in system condition.
These centralized management systems prohibit freedom of choice in the units.
In living systems, decision-making is decentralized, and units can choose
their actions. Bacteria have fewer choices of behavior than ants, which are, in
turn, more regimented and less “free” than animals. Mammal societies are more
complex than anthills or bacterial infections. As the “degrees of freedom” of
choice for individual members in a system increase, the range of individual
behaviors increases, making the system more complex.
The common assumption, usually implicit, that a system is simple rather than
complex simplifies analysis. If the system is simple, it can be decomposed into its
parts. It is nothing more than its parts and their defined relationships. The auto-
mobile can be disassembled and reconstructed and work just as well as it did before.
Disassembling a living system, or removing any part of it, can destroy the system or,
at least, make it much less than it was previously. Only Victor Frankenstein has yet
been able to deconstruct and reconstruct a human and breathe life into it.
The desire to simplify analysis also leads to the common assumption that
the system under study is closed to other systems, does not exchange energy with
them, and is not affected by them. The desire among social scientists for closed
systems reflects their common admiration for the analytical control of the labo-
ratory sciences. The laboratory is designed to close the system under study.
Boyle’s law that pressure and temperature are inversely related can be demon-
strated to be true only within a closed cylinder within a controlled environment.
Unfortunately, social systems are always open, and wishing them closed often
makes assumption of closure unreasonable.
Simple systems usually are static and tend to equilibrium; complex systems
are always dynamic and they are dissipative. This is most clearly illustrated by the
“arrow of time” (Prigogine 1997). Without an input of energy, a simple system can
remain largely unchanged for long periods. It declines only marginally by interac-
tion with its environment (to that extent, it is an open system). The automobile is
a static system that remains in equilibrium if no energy (for example, gasoline and
human control) is added to the system. In contrast, a living system perpetually
changes. Humans age and die, a dynamic process of constant change in the cells
within our bodies and the relationships between them. And we are “dissipative
structures” because we have to draw energy from our environment in the form of
oxygen, food and water merely to stay alive (Prigogine and Stengers 1984).
Even in simple systems, effects can feedback on their causes. Negative feed-
back slows down processes, and positive feedback speeds them up. The thermo-
stat is the classic example of a simple system with a negative feedback loop. As
the air cools below the set-point temperature, an electrical circuit closes to turn
on the furnace and blow hot air into the room. When the air is returned to its
set point, the circuit opens and the furnace shuts down. The homeostatic be-
havior of animals reflects feedback from activity (hunger, hunt, satiation, sleep).
Environmental selection operates on the individual agent as a form of feedback;
behavior can change from punishment/reward contact with the environment.
Complex systems usually have multiple feedback loops. Positive feedback
loops strengthen the cause and the subsequent effect in an ever increasing cycle
that can lead to nonlinear transitions and system collapse. For example, atmos-
pheric scientists hypothesize that positive feedback loops caused Venus’s swirling
toxic mists and 900-degree surface temperatures (Schneider 1989). Some scien-
tists fear that climate change on Earth could also progress with a nonlinear shift
in the system (Ocean Studies Board et al. 2001).
Complex systems are unpredictable. By its nature, nonlinearity is unpre-
dictable and difficult to represent mathematically, and most complex systems are
potentially nonlinear. In complex systems, prediction as a path-dependent ex-
trapolation of historical processes runs the risk of nonlinear change. Beyond the
very short term, the range of possible system paths for a complex system widens
dramatically. Decentralized decision-making and diversity among agents permits
a wide range of agent actions and openness to changes in environmental condi-
tions (the state of another complex system), and the prevalence of positive feed-
back loops inject further uncertainty into the system under study.
Complex systems may not be predictable, but they may be simulated with
interacting rules for agent behavior. These rules may be few and simple, yet the
outcome of their interaction can simulate complex systems in which agent be-
havior appears random and system order seems accidental. For example, the
flocking behavior of birds looks random and disorganized but can be modeled
with three rules (Waldrop 1992, 241–43). The location of water temples in Bali
can be simulated with a few rules of kinship and farming practice (Lansing, Kre-
mer, and Smuts 1998). The collapse of the Anasazi civilization in the American
Southwest has been explained by the interaction of social rules and environ-
mental changes (Axtell et al. 2002). In comparison to an automobile, the game
of checkers seems uncomplicated. Yet it “provides an almost inexhaustible vari-
ety of settings (board configurations)” (Holland 1998, 76). Because complexity
emerges from the simple rules of checkers, we should expect that “complexity
will be pervasive in the world around us” (76) both natural and social. But it also
“gives hope that we can find simple rule-governed models of that complexity.”
That hope is partially fulfilled by simulations of social systems with agent-based
models in which systems are modeled from the interactive behavior of essential
agents, as described throughout this book.
6 NEIL E. HARRISON
The characteristics summarized in table 1.1 and described in this section

are most commonly associated with each genus of system, simple or complex. No
single descriptor defines either simple or complex systems. For example, simple
systems may have many and diverse parts and complex systems (e.g., of bacteria)
may have homogeneous units; and complex systems can be, at least temporarily,
in equilibrium, while some simple systems appear dynamic. However, the more
descriptors of one system genus that can be attributed to a specific system, the
greater is the probability that that system is of that genus. Thus, complexity is an
accumulation of the characteristics of complex systems.
The next section shows how complexity concepts can be used to construct
a complex systems taxonomy of world politics (that is further elaborated in
chapter 2).
COMPLEX SYSTEMS IN WORLD POLITICS
Intuitively, the social world seems complex in the sense described here, but cur-
rent theories of world politics model it as a simple system. As Ruggie (1993) com-
ments, world politics theories are “reposed in deep Newtonian slumber.”
Newton described a universe formed out of particles that were all made
from the same material and whose movements in absolute space and time were
governed by forces that followed unchanging and universal laws. These laws
could be expressed exactly through mathematics (Capra 1982, 65–67; Ruggie
1993). For example, the properties of gases can be reduced to the mathematically
describable motion of their atoms or molecules. Thus, the image is of a universe
constructed like a perfect mechanical watch. Science, aided by mathematics, was
the method for prizing open the watch case to see the workings inside (Hollis
and Smith 1990, 47).
Locke and other early political and social theorists enthusiastically emu-
lated Newton and attempted “to reduce the patterns observed in society to the
behavior of its individuals” (Capra 1982, 69). A fixed human nature was pre-
sumed to determine human behavior, and “natural laws” governed spontaneous
human society: “As the atoms in a gas would establish a balanced state, so
human individuals would settle down in a society in a ‘state of nature.’” Natural
laws included freedom, equality, and property rights (Locke 1980, 123–27; Kym-
licka 1990, 95–159).
The shadow of Newton’s universe continues to obfuscate knowledge in the
social sciences. For example, while neoclassical economics remains the dominant
explanation of economic phenomena, it is “an economic science after the model
of mechanics—in the words of W. Stanley Jevons—as ‘the mechanics of utility and self-
interest’” (Georgescu-Roegen 1975, emphasis in the original). Economic actors are
assumed to be rational in their pursuit of undefined, subjective self-interest. Their
behavior is assumed to be an objectively rational response to external forces such
as the level of supply and demand of goods and services. If supply exceeds de-
mand and prices fall, economic actors will increase their purchases. In such a
model, agency is limited to only economic interests and programmed responses
to external stimuli.
Recent debates about agency and structure do not hide the similarly mech-
anistic paradigm that still drives orthodox theories of world politics. Essentially
identical units—interests and identities are assumed to be exogenously formed—
are driven by “natural laws” to behave predictably in response to exogenously de-
termined conditions. A rational-choice approach, borrowed from neoclassical
economics, is used in an attempt to generate ahistorical, universal explanations
of relations between states. The result is several significant simplifications of re-
ality. For example, concentrating on the state as the unit of analysis causes an an-
alytically convenient but arbitrary separation of international and domestic
politics, and the theoretical focus on “explaining constancies, not change” privi-
leges structure over agency (Smith 2004).
Constructivist theories—the most recent incarnation of liberalism—posit
that state interests and identities are intersubjectively malleable at the margin
through interaction with other states. While it is now historically located within
international society, as in rational-choice theories, the state remains the unit of
analysis. Thus, I start with the state to better illustrate the primary concepts of a
complexity taxonomy of world politics.
Emergence
A complex system is commonly described as more than the sum of its parts.
That is, properties of the system are emergent, created by the interaction of the
units. The basic unit of any social group is the individual. In biological terms,
the human body is a system; socially, each human is an essential unit within sev-
eral systems, and any social group, including the state, is an emergent system.
Social and political institutions emerge from the interaction of individual hu-
mans and human groups. Groups may be local or national; they may be loose-
knit coalitions or adhesive groups of fervent followers, and may be more or less
centrally organized. Out of the interactions among this mélange of groups and
individuals emerges the set of institutions, people, and practices that scholars
call the “state.”
8 NEIL E. HARRISON
Open Systems
The state is not a closed system: it is open to other natural and social systems.
For example, defined as a political system, it is open to technological, cultural,
and economic systems that influence political choices and processes (Skolnikoff
1993; on political economy, Gilpin 1996 and Strange 1994, 1996 among many
others).3 The state also is open to other states and, as constructivism argues, is in-
fluenced by interactions with them.
Some social systems are both within and outside the state. For example,
unions, major corporations, and nongovernmental organizations (NGOs) cross
boundaries and operate in several jurisdictions simultaneously (Goddard, Passé-
Smith, and Conklin 1996; Korten 1995; Keohane and Nye 1971).
Although the state is evidently an open system, theories of world politics
conventionally assume that all systems are closed to their environment much as
optimal natural science experiments are controlled and isolated from unwanted
external influences. Despite occasional attempts to bring in domestic politics
(Evans, Jacobson, and Putnam 1993; Putnam 1988), the state is usually modeled
as a unit with exogenous identity and objective interests. This greatly reduces the
range of possible causal explanations for any perceived social event, simplifying
causal analysis and hypothesis generation and testing.
The assumption of closure thereby permits historical theorizing and sup-
ports the widespread belief among scholars that general laws can be found. This
would be impossible if social systems were modeled as open, because “constant
conjunctions (empirical regularities) in general only obtain under experimentally
controlled conditions”—that is, under closure (Patomäki and Wight 2000). Open
systems are “susceptible to external influences and internal, qualitative change
and emergence” (232) and “outcomes might be the result of many different causes
and the same cause might lead to different outcomes” (229). Small changes that
can initiate a radical system shift may come from a change in environmental con-
ditions, or from inside, from interactions among its constituent agents. The non-
linearity of open systems prevents the theorist from mapping specific causes to
observed effects. Thus, open dynamic systems are inherently unpredictable
(Doran 1999). But that is no reason to model them as closed systems.
Meta-agents
The state is both an emergent system and a unit within the international system
of states. In Holland’s (1995) terminology, they are “meta-agents” whose “inter-
nal models” (discussed below) emerge from the interaction of domestic agents.
State behavior then results from the interaction of internal model and external
reality, and feedback is available on whether internal system processes and state
behavior “fit” within the environment, not unlike the concepts elaborated in
Putnam’s (1988) two-level game, though in a more fluid and dynamic relation-
ship. The concept of meta-agents can be used in any issue area in which agents
and actions at more than one level of aggregation are involved.
In contrast, orthodox international relations (IR) theory usually takes the
state as the primary unit of interest, while recognizing in passing the potential in-
fluence of substate and nonstate actors. Constructivism and other cognitive the-
ories treat states as subjects, but the state still is assumed to be a unitary actor
whose identity and interests change primarily as a result of interaction with other
states (Wendt 1994). The extent to which states also may self-consciously change
their interests and identity is debated, but the potential for change as a result of
domestic political discourse is usually disregarded (Hasenclever, Mayer, and Ritt-
berger 1997, 186–92).
Internal Models
Each human agent, the essential unit of any social system, has an internal model
of his or her desires and beliefs about how to achieve those desires in the world.4
If their beliefs are out of synch with reality, they will act inappropriately, fail to
achieve their goals, and may be punished. Agents who learn from such an expe-
rience, change their internal models and, thus, their behavior.
Individual agents’ behaviors are responsive and purposeful but not objec-
tively rational. According to Elster (1986, 16), an action is rational if it is the best
way for an actor to satisfy his or her desire based on beliefs that are optimal given
the available evidence and as much information as possible, given the desire. Be-
liefs and desires must be free of internal contradictions. Finally, actions must be
the intended result of beliefs and desires. This is substantially the same description
of rationality used by Green and Shapiro (1994, 6) to explain the foundations of
rational-choice theory. However, by assuming diversity among agents, complexity
does not make the simplifying jump to an assumption of objective rationality. Each
agent can have unique desires and unique beliefs about how to achieve them. The
alignment of behavior with desires and beliefs indicates agent rationality, but there
is no assumption that the outcomes of an agent’s choices will be individually or col-
lectively rational or will match agent intent. This is not Simon’s “substantive con-
ception of rationality” quoted and approved by Keohane (1988): “‘behavior that
can be adjudged objectively to be optimally adapted to the situation.” Because
agents cannot predict the effects of their actions in complex systems, behaviors
of individual agents are “optimally adapted” to their situation only accidentally.
Rationality is subjective—within the agent—rather than objective.
10 NEIL E. HARRISON
Constructivism broadens “the array of ideational factors that affect inter-

national outcomes” and introduces “logically prior constitutive rules alongside
regulative rules” (Ruggie 1998). The concept of internal models potentially ex-
tends the ideational content of world politics theories, while at the same time
making analysis of agent motives more difficult. However, simulation of agent
behavior now is possible.
Internal models drive agent behavior, but those models may change when
tested in a selective environment. Agents that consistently act in ways that are se-
lected by their social environment as suboptimal face eradication. Because states
are themselves systems, the process of matching internal model to external reality
is one of trial and error. As Putnam (1988) has suggested, the state may be not be
able to move its internal model—particularly in terms of its (causal) beliefs about
what is possible—to accord with the reality of the international system. If all states
are adaptive complex systems, then the international system emerges from coevo-
lution. International norms influence behavior through the internal process of in-
ternal model formation, one component of which is the desire to participate in a
society of nation-states. In chapter 5, Hoffmann investigates how states changed
their beliefs during negotiations over regulation of ozone depleting substances
and how the internal model of the United States adapted to these changes.
Dynamic Systems
Superficially, complexity appears to have some affinity with other world politics
systems theories, like neorealism and world systems. As Waltz (1979, 91) de-
scribes the international system, it is “formed by the coaction of self-regarding
units,” and its structure is “formed by the coaction of their units” and “emerge[s]
from the coexistence of states. . . . International-political systems, like economic
markets, are individualist in origin, spontaneously generated, and unintended.”
However, the similarity is more perceived that real, as shown in more detail in
chapter 2.
The distinction is in the details: in conventional systems theories, struc-
ture is a fixed or only slowly changing determinant of agent behavior. In complex
systems, structure is dynamic but “organization” is fixed. The “organization of a
living system is the set of relations among its components that characterize the
system as belonging to a particular class (such as bacterium, a sunflower, a cat, or
a human brain)” (Maturana and Varela 1980, 18). To describe the organization,
it is only necessary to describe the relationships and not the components. For ex-
ample, self-organization is “a general pattern of organization, common to all liv-
ing systems, whichever the nature of their components.” The structure of a
complex system is the actual relations among actual physical components: “[I]n
other words, the system’s structure is the physical embodiments of its organiza-
tion.” While organization is static—a cat cannot become a dog—structure is dy-
namic. Thus, structure is not fixed but a fleeting embodiment—in social systems
manifested by institutions—of the deep organization within apparent chaos. As
the momentary embodiment of prior agent interactions, complex system struc-
ture changes dynamically.
In complex systems, structure has a social role but no purpose. In func-
tional social theories like constructivism and neoliberalism, “history is path-
dependent in the sense that the character of current institutions depends not
only on current conditions but also on the historical path of institutional devel-
opment” (March and Olsen 1998, 959). Because “rules, norms, identities, orga-
nizational forms, and institutions that exist are the inexorable products of an
efficient history . . . surviving institutions are seen as uniquely fit to the environ-
ment, thus, predictable from that environment” (958). Complexity science
makes no such assertions: it does not assume or judge the fitness or efficiency of
emergent institutional arrangements. Institutions and rules are the consequence
of history but may not fit agents’ purposes.
The common (usually implicit) assumption that the international system is
homeostatic is a stronger version of the orthodox presumption that events in dif-
ferent spatiotemporal locations may be compared. It is equally untenable. Simple
dynamic systems find a point of equilibrium that is “sustained by micro-mecha-
nisms operating in finely attuned and compensating ways” (Elster 1983, 31–32).
Despite its “balance of power” bromide, classical realism is really about the
processes of systemic change from dynamic forces. Realism presumes that just as
the neoclassical market continually returns to an equilibrium between demand
and supply, the international system returns to a balance between many forces.
Complex social systems are never homeostatic: in both markets and world
politics the frequent and temporary equilibrium points are always distinct phe-
nomena. Each state of balance, like a human standing still through tensions be-
tween opposing muscles, is a fleeting event within a specific set of conditions, a
point on a path of change. The dynamic European system has found several mo-
mentary points of balance between myriad forces. Tudor England understood
the need to change alliances to continually balance power in Europe. Though
power was balanced in Europe before World War I and in the Cold War, the
conditions were unique to each period.
Causation
The uncertainty of complex social systems calls into question conventional

world politics assumptions about causation. Conventional world politics
12 NEIL E. HARRISON
theories presume that causation is proximate and proportionate. Like most of

social science, they have adopted Hume’s rules for causal explanations (Hume
1975).5 These rules require that the cause can be shown to precede the effect,
that cause and effect are contiguous (there was no intermediate event), and that
there is a “necessary connection” between events such that this cause can be
shown to always precede this effect under consistent conditions. For at least
four reasons, these rules are not appropriate causal explanations in complex so-
cial systems. First, they only apply in closed systems in which conditions can be
controlled. But if social systems are open, it is unlikely that conditions will re-
main constant or be comparable between different states of affairs. In an open
system, a cause may have different effects at different times due to changed con-
ditions. Therefore, it is not surprising that no general laws of world politics
have been found. Second, social systems are so complex that parsimonious the-
ories that attempt to isolate single (or few) causes for observed effects may dan-
gerously oversimplify models. In complex social systems, the events noted at the
start of this chapter (among others) are surprising only when we expect to find
a singular cause. Understood as the emergent consequence of multiple inter-
acting prior events, such events are less astounding. The events of September
11, 2001, may be the result of all of the explanations commonly offered: failures
of collection, coordination, and distribution of intelligence; a clash of cultures;
hatred by fanatics; and so on. But each of these “causes” were themselves caused
by multiple prior events. Osama Bin Laden is the product of his family, Islam,
the Saudi culture, and personal experience defending Afghanistan against the
Soviets. The clash of cultures (or civilizations: Huntington 1993) is as much a
consequence of U.S. actions as of Muslim choices. Intelligence failures resulted,
in part, from decisions that restricted human intelligence gathering, decisions
made by successive US governments after several high-profile misadventures in
the 1970s. Thus, September 11 could have emerged from a plethora of choices
and events across the globe over decades, not as an inevitable consequence of
any of them but as a path-dependent phenomenon. And if it was not path de-
pendent, it was a symptom of a nonlinear system shift that cannot be predicted
or explained. In neither case is conventional thinking about causation useful.
Third, the immediate cause of an effect may, as part of a higher-order Markov
chain, itself be the effect of an earlier, and possibly more important, cause. If an
earlier cause is more important than later ones in the chain, this implies action
at a distance in space/time that both Newton and Hume reject (Elster 1983,
26–30). Fourth, causation may be simultaneous as in open, emergent systems
where the interaction of parts of the system constitute the system.
In addition to these limits to the normal rules of causal explanation, as-
sumptions of the proportionality of cause and effect are often erroneous. As dis-
cussed above, in open, emergent systems, small perturbations in the system may
have very large effects, making identification of the connection between cause
and effect nearly impossible and explanation problematic. Was the fact of Kaiser
Wilhelm’s withered arm or his relationship with his English nanny a sufficient
cause of World War I (Röhl 1998)?
In a complex system, many factors symbiotically cause an effect. Theorists
should look to the evolution of the system, not to individual events, for causes of
observed effects. Patomäki and Wight (2000, 230) argue that “ontologically, the
social world can only be understood as a processual flow that is intrinsically open
and subject to multiple and at times contradictory causal processes.” Uninten-
tionally, this is a fair exposition of complex systems. Social phenomena only
occur because agents act within an existing and real context that is “not re-
ducible to the discourses and/or experiences of the agents,” as constructivists
argue. As Maturana and Varela (1980, 98) wrote: “[O]ur problem is the living or-
ganism and therefore our interest will not be in properties of components, but
in processes and relations between processes realized through components.” In
social systems, processes are not as automatic as they are in insects and bacteria.
Humans and social groups are conscious and self-aware entities (that is, their in-
ternal models are more elaborate and complex) who, therefore, may act strategi-
cally toward some goal within their perception of their environment.
PLAN OF THE BOOK
Most social sciences have begun to embrace complex systems concepts. Ideas
from thermodynamics coupled with a concern for economic systems’ environ-
mental effects (Georgescu-Roegen 1975, 1971) led to the development of eco-
logical economics that specifically models the economy as an open system
(Barbier, Burgess, and Folke, 1994; Krishnan, Harris, and Goodwin 1995;
Costanza 1991; Daly 1991). Brian Arthur and others have identified the pres-
ence and effect of feedback loops in economic systems (van Staveren, 1999;
Arthur, Durlauf, and Lane 1997; Arthur 1990; Arthur 1989; Anderson, Arrow,
and Pines 1988; Romer 1986). Complex systems approaches have attracted soci-
ological interest (Luhmann 1998, 1990; Eve, Horsfall, and Lee 1997; Knapp
1999; Hanneman 1988; Collins, Hanneman and Mordt 1995) and touched pub-
lic administration and organization studies (Griffin 2002; Stacey 2001; Marion
1999; Elliott and Kiel 1999). Even political science is not immune (Richards
2000; Axelrod 1997; Jervis 1997; Cilliers 1998; Cederman 1997; Cederman and
Gleditsch 2004), though efforts are disparate and inchoate. This book is de-
signed to drive forward the complexity research agenda as a viable alternative to
orthodox theories of world politics by establishing the central concepts and ideas
14 NEIL E. HARRISON
needed for the development and empirical assessment of complex systems theo-
ries of issue-areas in world politics.
The next nine chapters further develop the concepts outlined in this chap-
ter and illustrate their application to several world politics issue areas. Chapter 2
begins to sketch out a taxonomy of complexity by comparing complex systems
concepts to those developed more than three decades ago for a general systems
taxonomy. Systems theories that were relatively popular in the early days of the
Cold War have, in recent years, have fallen into disrepute as overly “grand” in
purpose. Harrison, with Singer’s aid, compares and contrasts conceptual de-
scriptions between general systems and complex systems taxonomies. Several
concepts are common to the two approaches, but this chapter also identifies the
important differences between the two taxonomies. Complexity is not a warmed-
over version of general systems theory but builds on its ideas to generate theories
that better explain issue-areas in world politics.
As this is a new approach to understanding world politics/IR, this book
does not attempt to illustrate its application to the whole range of possible is-
sues. The next four chapters show how complexity can generate new insights
and hypotheses when applied to selected issue areas. They are arranged from
the least to most technical in their use and application of complexity con-
cepts. Because this book is an introduction to complexity in IR that is in-
tended to initiate research rather than to develop applications adapted to
all issue-areas of international relations, these chapters are only exemplars
of the application of the complexity paradigm. None formally models their
case but they all describe how their hypotheses might be further elaborated or
empirically tested.
In chapter 3, Dennis Sandole argues that complexity creates opportunities
to integrate and synthesize apparently opposing worldviews. He reconsiders the-
ories of identity-based conflict in the post-9/11 world and proposes a theoreti-
cal framework to demonstrate that traditionally competing Realpolitik and
Idealpolitik (conflict resolution) approaches can coexist. Not only can they co
exist, but more robust guides to identify conflict and formulate policy responses
can be constructed by integrating both approaches into a single framework.
In chapter 4, Walt Clemens attacks a knotty puzzle that has emerged from
the collapse of the Soviet empire: why have some ex-Soviet states fared far better
than others? Natural resources, education, and ethnic homogeneity do not ex-
plain why the Baltic states and Slovenia are joining the European Union, while
oil-rich and more-homogeneous states are embroiled in factional fighting or war,
or have stagnated in neo-Stalinism. Using complexity concepts, Clemens pro-
poses an innovative explanation of why some newly freed states appear to have
failed while others are joining the EU.
Drawing on complex adaptive systems theories (a version of complexity

that uses more life science concepts), Clemens notes that some states were “fit-
ter” than others and so better able to exploit opportunities that opened for them
after the collapse of the Soviet empire. Seeking the sources of that fitness, he
finds that long-standing, religiously inspired institutions in the Protestant coun-
tries developed internal models in the population that reduced ethnic tensions
and increased acceptance of democratic virtues. He also shows that his marker
for fitness correlates with measures of development and describes how to empir-
ically test his hypothesis.
Matt Hoffmann looks at the coevolution of states’ internal models
in chapter 5. He considers two puzzles in the formation of the interna-
tional regime designed to protect the stratospheric ozone layer: why did the
norm governing participation change and why did the United States accept
this new norm?
Hoffmann shows that rational explanations are deficient and that complex
systems concepts can help us to unravel both puzzles. From a complexity perspec-
tive, evolution of the universality norm is a simple story of complex adaptation.
As some Southern Hemisphere states’ internal models changed to universal par-
ticipation, the flux in the system eventually led other states to adapt to the new in-
ternational norm. Hoffmann shows that when the United States reconsidered its
internal model (with some pressure from domestic groups), it recognized it would
have to accept the universality norm and negotiate in good faith with the South
to achieve its goal of an effective treaty. He concludes with suggestions for theo-
retical, empirical, and methodological development of these ideas.
In recent years, genocide within a country has become an international
issue. The stimulus to this international interest in domestic interracial relations
was the terrifying genocidal violence in Rwanda in 1994 that killed possibly as
many as eight hundred thousand people. In addition to the moral implications,
since Rwanda it is now clear that genocide in one country has serious conse-
quences for its neighbors, making it a legitimate concern for the international
community (“The Road Out of Hell,” Economist, March 27, 2004, 25–27). In
chapter 6, Ravi Bhavnani shows how complexity concepts can help us to under-
stand why the speed and magnitude of the killing was so much greater than in all
previous ethnic attacks in that country.
Conventional explanations of the scale of the Rwandan violence are inad-
equate. Bhavnani shows how bottom-up simulations can generate new hypothe-
ses about the spread of ethnic violence. Building on evidence from the field and
reasonable assumptions about relationships between extreme and moderate
Hutus, he describes a simulation of how the killing rampage took hold so quickly
and led to murder even within families.
16 NEIL E. HARRISON
The next three chapters explain the empirical validity of simulations, discuss
potential problems with constructing complex systems theory, and show how
multiple ABMs may be used to improve forecasting and decision-making. Chapter
7 is a reprint—used with permission—of part of Robert Axelrod’s chapter entitled
“Advancing the Art of Simulation in the Social Sciences,” in Simulating Social Phe-
nomena, edited by Rosario Conte, Rainer Hegselmann, and Pietro Terna (Berlin:
Springer-Verlag, 1997), 21–40. Axelrod argues that simulation is best thought of as
a new way of doing social science. Inductive methods are needed to find patterns
in, for example, opinion surveys and macroeconomic data, and sometimes in in-
ternational interactions. If social agents are assumed, as in conventional theories,
to be objectively rational actors, deductive methodology suffices. Simulation is the
third way—the only way, if agents are assumed to be adaptive. In the social sciences,
the most common form of simulation is agent-based modeling (ABM), which
builds systems from the bottom up rather than, as with deductive methods, from
the top down. Like deduction, simulation starts with explicit assumptions, but it
cannot prove theorems. Like induction, it looks for patterns, but it uses data gen-
erated from defined rules rather than the real world. Axelrod argues that, in social
science simulations, simple is better: like thought experiments simulations can
deepen understanding of fundamental processes.
David Earnest and Jim Rosenau in chapter 8 question whether political
systems are complex systems, as commonly understood, and argue that simula-
tion of political systems begs the questions it attempts to answer: who are the ac-
tors and who has authority? They reject complexity as a theory, because it fails
the standard of theory in positivist epistemology and offers no alternate episte-
mology; and implicitly they reject more limited applications of complex systems
theory. While Axelrod describes simulation as a third way, Earnest and Rosenau
argue that it is no way: it lacks both the empirical appeal of induction and the
disconfirmative value of deduction. For them, thought experiments are “much
ado about nothing.” They acknowledge that complexity is an attractive paradigm
but argue that more development is required before it may generate viable theo-
ries of world politics.
Chapter 9 is an indirect response to Earnest and Rosenau’s critiques. In
Desmond Saunders-Newton’s opinion, while there are epistemological problems
with ABMs, these are neither insurmountable nor critical problems. As scholars
debate the fine points of ontology and epistemology, complex systems thinking
and ABMs already are being put to use in the service of policymakers to generate
and assess multiple policy options.
Saunders-Newton argues that complex systems thinking and computa-
tional methods that emphasize agent-level phenomena are part of a new trans-
discipline that allows analysts to rigorously consider increasingly complex
phenomena in an interdisciplinary way. In addressing the epistemological issues

surrounding computational methodologies, he argues that efficacy or usefulness
is more important than the quality of model isomorphism and method. He then
describes how several computational social science models (including ABMs), in-
tegrated with computer-assisted reasoning methods, are being used in the Pre-
Conflict Management Tools Program being tested at the National Defense
University for its ability to assess social vulnerability.
The concluding chapter draws some general lessons from the four cases and
shows how they illustrate important complexity concepts. It then assesses the va-
lidity of simulations and computational models and the epistemological ques-
tions they raise. It also shows that, from complexity concepts and ideas, complex
systems theories for issue-areas can be specified and models for specific problems
generated. Yet, because political systems are complex—and becoming more com-
plex—a new epistemology and new methods are needed to understand them. Fi-
nally, it shows that recognition of complexity in politics suggests radically new
policies for addressing international problems and pursuing national interests.
NOTES
1. The terms “world politics” and “international relations” are used in-
terchangeably throughout this book. I use “world politics” to better reflect the
multilevel structure of the political world to which complex systems thinking is
so well adapted. Patomäki and Wight (2000, 232–33) opine that the “key error”
of much international theorizing is “to treat levels of the state and the interna-
tional system as related as agents to structures” instead of as “layers” within
world politics. The terms “complex system” and “complex adaptive system” are
often used interchangeably; the concepts described here principally derive from
the latter. I use the term “paradigm” in the sense of a set of assumptions, con-
cepts, values, and practices that comprise a view of reality, and in that sense it is
quite comparable to “worldview” (Hughes 2000).
2. Common knowledge also shortens description. “Bicycle” conveys to
most people a clear image of the system. Imagine how much more complex
would be a description to a Martian who is completely unfamiliar with a bicycle
or any of the common parts used in its assembly.
3. Smith (2004) comments that world politics/IR theorists err in thinking
of the state as solely political. Whether the state is modeled as political interact-
ing with other subsystems of society or as a political unit of a social system
among economic and cultural agents depends on the question being addressed.
4. In this book, the terms “internal model,” “mental model,” and “schema”
are used interchangeably.
18 NEIL E. HARRISON
5. As a skeptic, Hume also argued that causation is a human construct.

All we ever see is the conjunction of events, and we impute a causal relationship.
But without an explanatory force linking cause and effect, causation cannot be
“real” (Patomäki and Wight 2000).
REFERENCES
Anderson, Philip W., Kenneth J. Arrow, and David Pines, eds. 1988. The Econ-
omy as an Evolving Complex System. Santa Fe Institute Studies in the Sci-
ences of Complexity, 5. Redwood City, CA: Addison-Wesley.
Arthur, W. Brian. 1989. “Competing Technologies, Increasing Returns, and
Lock-in by Historical Events.” Economic Journal 99 (March 1): 116–31.
———. 1990. “Positive Feedbacks in the Economy.” Scientific American 262, no. 2
(February): 92–99.
Arthur, W. Brian, Steven N. Durlauf, and David A. Lane, eds. 1997. The Econ-
omy as an Evolving Complex System II. Reading, MA: Addison-Wesley.
Axelrod, Robert M. 1997. The Complexity of Cooperation: Agent-Based Models of
Competition and Cooperation. Princeton, NJ: Princeton University Press.
Axtell, Robert L., Joshua M. Epstein, Jeffrey S. Dean, George J. Gumerman,
Alan C. Swedlund, Jason Harburger, Shubha Chakravarty, Ross Ham-
mond, Jon Parker, and Miles Parker. 2002. “Population Growth and
Collapse in a Multiagent Model of the Kayenta Anasazi in Long House
Valley.” Proceedings of the National Academy of Science of the United States
of America, supp. 3, vol. 99, no. 10:7275–79.
Barbier, Edward B., Joanne C. Burgess, and Carl Folke. 1994. Paradise Lost: The Eco-
logical Economics of Biodiversity. London: Earthscan for the Beijer Institute.
Buzan, Barry, and Richard Little. 2001. “Why International Relations Has
Failed.” Millennium: Journal of International Studies 30, no. 1:19–39.
Capra, Fritjof. 1982. The Turning Point. New York: Simon & Schuster.
Cederman, Lars-Erik. 1997. Emergent Actors in World Politics: How States and Na-
tions Develop and Dissolve. Princeton, NJ: Princeton University Press.
Cederman, Lars-Erik, and Kristian Skrede Gleditsch. 2004. “Conquest and
Regime Change: An Evolutionary Model of the Spread of Democracy
and Peace.” International Studies Quarterly 48, no. 3 (September): 603–29.
Cilliers, Paul. 1998. Complexity and Postmodernism: Understanding Complex Systems.
New York: Routledge.
Collins, R., R. Hanneman, and G. Mordt. 1995. “Discovering Theory Dynamics

by Simulation.” Sociological Methodology 25: 1–46.
Costanza, Robert, ed. 1991. Ecological Economics: The Science and Management of
Sustainability. New York: Columbia University Press.
Daly, Herman E. 1991. Steady-State Economics. 2nd ed. Washington, DC: Island
Press.
Der Derian, James, and Michael J. Shapiro, eds. 1989. International/Intertextual
Relations: Postmodern Readings of World Politics. Issues in World Politics,
ed. James N. Rosenau and William C. Potter. Lexington, MA: Lexing-
ton Books.
Doran, Charles F. 1999. “Why Forecasts Fail: The Limits and Potential of Fore-
casting in International Relations and Economics.” International Studies
Review 1, no. 2 (Summer): 11–41.
Elliott, Euel, and L. Douglas Kiel, eds. 1999. Nonlinear Dynamics, Complexity and
Public Policy. Commack, NY: Nova Science Publishers.
Elster, Jon. 1983. Explaining Technical Change. Cambridge: Cambridge University
Press.
———. 1986. Introduction to Rational Choice, ed. Jon Elster, 1–33. New York:
New York University Press.
Evans, Peter B., Harold K. Jacobson, and Robert D. Putnam, eds. 1993. Double-
Edged Diplomacy: Institutional Bargaining and Domestic Politics. Studies in
International Political Economy, ed. Stephen D. Krasner and Miles
Kahler, 25. Berkeley: University of California Press.
Eve, Raymond A., Sara Horsfall, and Mary E. Lee, eds. 1997. Chaos, Complexity
and Sociology: Myths, Models, and Theories. Thousand Oaks, CA: Sage.
Gell-Mann, Murray. 1994. The Quark and the Jaguar: Adventures in the Simple and
the Complex. New York: W. H. Freeman and Company.
George, Jim, and David Campbell. 1990. “Patterns of Dissent and the Celebra-
tion of Difference: Critical Social Theory and International Relations.”
International Studies Quarterly 34, no. 3 (September): 269–93.
Georgescu-Roegen, N. 1971. The Entropy Law and the Economic Process. Cam-
bridge, MA: Harvard University Press.
———. 1975. “Energy and Economic Myths.” Southern Economic Journal 41, no. 3
(January): 347–81.
Gilpin, Robert G. 1996. “The Nature of Political Economy.” In International
Political Economy: State-Market Relations in the Changing Global Order, ed.
20 NEIL E. HARRISON
C. Roe Goddard, John T. Passé-Smith, and John G. Conklin, 9–24.

Boulder, CO: Lynne Rienner.
Goddard, C. Roe, John T. Passé-Smith, and John G. Conklin, eds. 1996. Inter-
national Political Economy: State-Market Relations in the Changing Global
Order. Boulder, CO: Lynne Rienner.
Green, Donald P., and Ian Shapiro. 1994. Pathologies of Rational Choice Theory:
A Critique of Applications in Political Science. New Haven, CT: Yale Uni-
versity Press.
Griffin, Douglas. 2002. The Emergence of Leadership: Linking Self-Organization and
Ethics. New York: Routledge.
Hanneman, R., 1988. Computer-Assisted Theory Building. Newbury Park, CA: Sage.
Hasenclever, Andreas, Peter Mayer, and Volker Rittberger. 1997. Theories of
International Regimes. Cambridge: Cambridge University Press.
Holland, John H. 1995. Hidden Order: How Adaptation Builds Complexity. Read-
ing, MA: Perseus Books.
———. 1998. Emergence from Chaos to Order. Cambridge, MA: Perseus Books.
Hollis, Martin, and Steve Smith. 1990. Explaining and Understanding International
Relations. Oxford: Clarendon Press.
Hughes, Barry B. 2000. Continuity and Change in World Politics: The Clash of Per-
spectives. 3rd ed. Englewood Cliffs, NJ: Prentice-Hall.
Hume, David. 1975. Enquiries Concerning Human Understanding and Concerning the
Principles of Morals. 3rd. ed., with revisions and notes by P. H. Nidditch.
Oxford: Clarendon Press.
Huntington, Samuel P. 1993. “The Clash of Civilizations?” Foreign Affairs 72, no.
3:22–49.
Jervis, Robert. 1976. Perception and Misperception in International Politics. Prince-
ton, NJ: Princeton University Press.
———. 1997. System Effects: Complexity in Political and Social Life. Princeton, NJ:
Princeton University Press.
Kaplan, Morton A. 1969. “Variants on Six Models of the International System.”
In International Politics and Foreign Policy, ed. James N. Rosenau.
Toronto: Free Press.
Keohane, Robert O. 1988. “International Institutions: Two Approaches.” Inter-
national Studies Quarterly 32, no. 4 (December): 379–96.
Keohane, Robert O., and Joseph S. Nye, eds. 1971. Transnational Relations and
World Politics. Cambridge, MA: Harvard University Press.
Kiel, L. Douglas. 1994. Managing Chaos and Complexity in Government. San Fran-
cisco: Jossey-Bass.
Knapp, Peter. 1999. “Evolution, Complex Systems and the Dialectic.” Journal of
World-Systems Research 5:74–103.
Korten, David C. 1995. When Corporations Rule the World. West Hartford, CT:
Kumarian Press and Berrett Koehler.
Krishnan, Rajaram, Jonathan M. Harris, and Neva R. Goodwin, eds. 1995. A
Survey of Ecological Economics. Washington, DC: Island Press.
Kymlicka, Will. 1990. Contemporary Political Philosophy: An Introduction. Oxford:
Clarendon Press.
Lansing, J. Stephen, James N. Kremer, and Barbara B. Smuts. 1998. “System-
Dependent Selection, Ecological Feedback and the Emergence of Func-
tional Structure in Ecosystems.” Journal of Theoretical Biology 192:377–91.
Locke, John. 1980. Second Treatise of Government. Ed. C. B. Macpherson. Indi-
anapolis: Hackett.
Luhmann, Niklas. 1990. Essays on Self-Reference. New York: Columbia University
Press.
———. 1998. Observations on Modernity. Stanford, CA: Stanford University Press.
March, J. G., and J. P. Olsen. 1998. “The Institutional Dynamics of International
Political Orders.” International Organization 52, no. 4 (Autumn): 943–69.
Marion, Russ. 1999. The Edge of Organization: Chaos and Complexity Theories of For-
mal Social Systems. Thousand Oaks, CA: Sage.
Maturana, Humberto, and Francisco Varela. 1980. Autopoiesis and Cognition.
Dordrecht, Holland: D. Reidel.
Ocean Studies Board, Polar Research Board, and Board on Atmospheric Sci-
ences and Climate. 2001. Abrupt Climate Change: Inevitable Surprise.
Washington, DC: National Academy of Sciences.
Patomäki, Heikki, and Colin Wight. 2000. “After Postpositivism? The Promises
of Critical Realism.” International Studies Quarterly 44, no. 2 (June):
213–37.
Prigogine, Ilya. 1997. The End of Certainty: Time, Chaos, and the New Laws of Na-
ture. In collaboration with Isabelle Stengers. New York: Free Press.
22 NEIL E. HARRISON
Prigogine, Ilya, and Isabelle Stengers. 1984. Order Out of Chaos: Man’s New Dia-
logue with Nature. New York: New American Library.
Putnam, Robert D. 1988. “Diplomacy and Domestic Politics: The Logic of Two-
Level Games.” International Organization 42, no. 3 (Summer): 427–60.
Richards, Diana. 2000. Political Complexity: Nonlinear Models of Complexity. Ann
Arbor: University of Michigan Press.
Röhl, John C. G. 1998. Young Wilhelm: The Kaiser’s Early Life, 1859–1888. Cam-
bridge: Cambridge University Press.
Romer, Paul. 1986. “Increasing Returns and Long Run Growth.” Journal of
Political Economy 94, no. 5: 1002–35.
Ruggie, John Gerard. 1993. “Territoriality and Beyond: Problematizing Moder-
nity in International Relations.” International Organization 47, no. 1
(Winter): 139–74.
———. 1998. Constructing World Polity: Essays on International Institutionalization.
London: Routledge.
Schneider, Stephen H. 1989. Global Warming: Are We Entering the Greenhouse Cen-
tury? New York: Vintage Books.
Skolnikoff, Eugene B. 1993. The Elusive Transformation: Science, Technology, and
the Evolution of International Politics. Princeton, NJ: Princeton University
Press for the Council on Foreign Relations.
Smith, Steve. 2004. “Singing Our World into Existence: International Relations
Theory and September 11.” International Studies Quarterly 48, no. 3
(September): 499–515.
Smith, Steve, Ken Booth, and Marysia Zalewski, eds. 1996. International Theory:
Positivism & Beyond. Cambridge: Cambridge University Press.
Sprout, Harold, and Margaret Sprout. 1971. Toward a Politics of the Planet Earth.
New York: Van Nostrand Reinhold.
———. 1978. The Context of Environmental Politics: Unfinished Business for America’s
Third Century. Lexington: University Press of Kentucky.
Stacey, Ralph D. 2001. Complex Responsive Processes in Organizations: Learning and
Knowledge Creation. London: Routledge.
Strange, Susan. 1994. States and Markets. 2nd ed. London: Frances Pinter.
———. 1996. The Retreat of the State: The Diffusion of Power in the World Economy.
Cambridge: Cambridge University Press.
van Staveren, Irene. 1999. “Chaos Theory and Institutional Economics: Meta-
phor or Model?” Journal of Economic Issues 33, no. 1 (March): 141–67.
Waldrop, M. Mitchell. 1992. Complexity: The Emerging Science at the Edge of Order
and Chaos. New York: Simon & Schuster.
Waltz, Kenneth. 1979. Theory of International Politics. Reading, MA: Addison-
Wesley.
Wendt, Alexander. 1994. “Collective Identity Formation and the International
State.” American Political Science Review 88:384–96.
This page intentionally left blank.
CHAPTER 2
Complexity Is More Than Systems Theory
Neil E. Harrison with J. David Singer
The previous chapter emphasized the differences between conventional theories

of world politics and international relations and complexity. In 1971 Singer pro-
posed a general systems taxonomy that potentially supported both systemic ex-
planation and more limited theorizing. This chapter shows how the important
concepts of general systems taxonomy compare with concepts in a complex sys-
tems taxonomy. In addition, we argue that complexity’s modifications to general
systems taxonomy create a more flexible taxonomy that incorporates current
knowledge and integrates theories that hitherto have been considered incom-
mensurate views of world politics. We anticipate that the complex systems tax-
onomy can generate radically new hypotheses about world politics and develop
innovative ways of testing them and thereby increase knowledge and improve
policy-making.
CONCEPTS IN COMMON
A taxonomy “has no descriptive, predictive or explanatory power, since it contains

only definition propositions. But to serve as the basis for building models and the-
ories, it must specify two things quite clearly: the basic constructs by which the rel-
evant domain is to be described and the definitional relationships between and
among these constructs” (Singer 1970, 3). As a framework for theorizing about
world politics and international relations, complexity modifies and expands on
Singer’s general systems taxonomy, which itself built upon current knowledge
from orthodox theories (such as it was) and corrected the epistemological and
methodological defects found in most general systems theories. Despite correcting
25
26 NEIL E. HARRISON WITH J. DAVID SINGER
faults of general systems theories and a well-argued explanation of the benefits of

a general systems approach, in the last two decades Singer’s taxonomy has attracted
little theoretical interest.
This section shows that some of the concepts from the general systems ap-
proach appear only slightly modified in the complex systems taxonomy. In both
taxonomies the international system is modeled as comprising multiple, hierar-
chical systems of interacting agents, and the systems are open to their environ-
ments but can be distinguished for theoretical purposes. Each system can
contain multiple feedback loops and may be susceptible to path dependencies,
and none is likely to display rational behavior. Yet, as described in the next sec-
tion, complexity also changes, develops, and adds to general systems concepts in
ways that increase its potential utility for building effective theories of world pol-
itics and international relations.
Nested Systems
Textbooks and scholars agree that “there are different ontological layers in the
world” (Patomäki and Wight 2000, 232). Most textbooks recognize several levels
of analysis from individuals to the international system in which to seek expla-
nations of global or international events. Commonly, at least five are identified:
system, state, society, government, and individual. Yet, in pursuit of a false sim-
plicity, international theory has been largely confined to competing, singular lev-
els of analysis. Singer (1970) comments, for example, that the “perhaps fatal . . .
flaw lies in the general tendency to focus on only one level of analysis, rather
than treat the interactions that occur across the several relevant levels. The com-
mon focus on a single level of analysis blinds theorists to influential processes op-
erating at other levels of analysis.”
Both the general systems and complex systems taxonomies explicitly
model the ontological layers in world politics as interrelated systems. In the gen-
eral systems taxonomy the world political system is modeled as “a hierarchy of
nested sets of subsystems, each embraced by those at the next higher level of
analysis and embracing those at all lower levels. It follows from this that any sys-
tem or set of systems at one level of analysis constitutes the environment of all
the entities existing at any lower level” (Singer 1971, 12). For each state, the in-
ternational system is only “real” as an environment within which it operates.
Singer argued (1971, 17) that the nation-state remains a “useful object of analy-
sis, but that at the same time the many entities comprising those social coalitions
known as nations may also serve as useful objects of analysis.” But the state is not
a solid body; it is a “coalition of all social entities at the individual, primary, and
secondary levels” (1971, 17), and government agencies are only components in
COMPLEXITY IS MORE THAN SYSTEMS THEORY 27
the aggregation known as the state. Similarly, complexity views the state as a
“meta-agent” (Holland 1995) that forms its variable internal model out of the on-
going interactions of social aggregations within its domestic political processes.
However, as described below, complexity and general systems approaches diverge
in their conception of cause-and-effect relations across levels of analysis.
Agents, Not Actions
Singer (1971, 7–11) identifies two schools of general systems theories: system-of-
action and system-of-entities. In the former school, systems are identified around
“actions, behavior, interaction, relationship, or role,” largely ignoring the entities
that “participate and experience them.” The latter school models systems
“around individuals, groups, associations, or aggregations of people”—that is, so-
cial entities. In contrast to the “actions” school, scholars in this tradition usually
explicitly posit that systems “will show rather similar patterns and processes as
well as a fair degree of structural isomorphism.” He then argues persuasively that
systems must be conceived of in terms of the characteristics of their constituent
entities rather than in terms of agent actions. He shows that this approach is
methodologically more tractable; it permits more effective separation of subsys-
tems (e.g., political from economic) where this would be theoretically more use-
ful; it more clearly distinguishes the system from its environment (is the social
system part of the political or vice versa?); and it clarifies levels of analysis (at
what level are individual actions of decision-makers?). Yet, in international rela-
tions and world politics theories, scholars often focus on the actions and abstract
away from agents. For example, Kaplan’s (1957) model describes a system of ac-
tions and interactions between states, and Waltz (1979) elaborates a model of
the international system that he compares to a market and in which, beyond
crude power measures, the characteristics of the actors are of no interest. His-
torical-materialist theories focus on the structural forces that dictate state behav-
ior, regardless of the characteristics of the states.
As in Singer’s general systems taxonomy, complexity assumes that the
characteristics of social entities generate agent actions and participate in con-
structing system structure. Complexity posits that internal models cause agent ac-
tions and the pattern of agent behavior reflects the interaction of the agent’s
internal model with environmental constraints. Thus, agent-based models as-
sume that agents choose actions that are consistent with their individual desires
and their beliefs about how to satisfy those desires. In social systems, institutions
are the environmental selection rules that govern punishment and reward for
agent actions. Institutions are the dynamic, path-dependent consequences of
prior agent interactions through earlier patterns of institutions.1
Open Systems
Political systems are conventionally assumed to be closed and homeostatic: dis-

turbances are temporary and the system tends to return to equilibrium. Yet, in
his 1971 monograph Singer simply states that all social systems are open—that is,
their boundaries are “permeable to information and energy from [their] envi-
ronment”—and that no social system can realistically be treated as closed. There
is, he writes, “clearly no such thing as a completely closed social system” (13).2 If
so, assuming closure to simplify analysis and reduce empirical effort is not a rea-
sonable parsimony but a gross distortion of reality. While every model is neces-
sarily an “ersatz” reality, it must retain a recognizable link to the portion of reality
it purports to represent or it will generate inaccurate “knowledge.”
Complexity similarly treats all social systems as open, but, as Singer has
commented, it is an empirical question as to how open each is (1971, 13). Some
scholars approvingly describe rural communities’ openness to their “natural” en-
vironment, especially in poorer countries. However, such open communities
often are both somewhat closed to their social environment and much simpler
than more complex, modern societies that are more open to other social systems
but less open to ecological systems (Harrison n.d.). Openness is not a measure of
complexity, but complex systems usually are open.
Feedback
Conflict recurs in the international system because the conflicting incentives and
temptations within nations and the lack of effective constraints between nations
support positive feedbacks to conflict: “[I]ntra-national and inter-national events
all impinge on one another in a cyclical and ongoing process within which the
self-aggravating propensities frequently exceed the self-correcting ones by an un-
acceptably large amount” (Singer 1970, 165). National elites use rhetoric for do-
mestic political consumption that can incite potential enemies, the public and
military desire the psychological comfort of discernible superiority, media am-
plify internation conflicts, and the benefits of participation in the ideological
mainstream preserve the distribution of power and inhibit changes in the his-
torical patterns that transform inevitable conflicts into costly rivalries. Self-
restraint within political elites and the media has diminished with the increase
in the number of competing media companies, their geographical coverage, and
the diminishing time lag between event and report, and corrective mechanisms
within the international system have atrophied. Technologically induced imme-
diacy reduces opportunities for editorial restraint. Similarly, technology has
reduced the time available for consideration of alternatives by decision-makers,

and raw data crowd information and reduce the quality of its assessment.
But positive feedback also can be beneficial. Constructivists would point
to positive feedbacks operating in the formation of norms that underpin inter-
national environmental treaties (as Hoffmann describes in chapter 5). The Mon-
treal Protocol is often cited as a precedent for the 1992 signing of a global treaty
to mitigate emission of gases believed to fuel climate change. Earlier examples
come from the literature on the formation of the European Community, in
which functional links or communications links were thought to increase trust
and lead to eventual integration between countries (Mitrany 1966; Haas 1964;
Deutsch 1953).
Drawing on models of ecological systems, complexity posits that the in-
teraction of multiple independent, volitional agents allows positive feedback
loops to develop that can drive the system to “flip” to a new state (Levin 1999).
Revolution might be an example of a flip in a social system. Like ecosystems, so-
cieties have often collapsed from runaway internal feedback loops (Mäler 2000;
Tainter 1988).
Path Dependence and Randomness
Path dependence is the idea that system development from time t to t⫹1 is not
wholly random and can only fall within limits created by the prior state of the
system. Living systems are considered path dependent: the current system state
is related to and is, in part, determined by the prior state of the system, and that
to its prior state all the way back to its nascence. Similarly, social systems evolve
continuously, and the international system may change its structure without be-
coming another system. The Cold War was a period in the evolution of the in-
ternational system that was in part caused by all of history that preceded it. It was
not a discrete system and cannot be separated from its history. But to state that
the Cold War evolved out of prior history is not to claim that it was an inevitable
effect of historical causes. The choices of multiple discretionary agents (from in-
dividuals to states) inject randomness into outcomes. Thus, in complex systems
the arrow of time is not reversible (Prigogine 1997).
Studies show that in conflict prior experience matters in at least two
ways. First, conflict stimulates innovation in search of increased military capa-
bility. Armed conflict is an incentive to modernize both equipment and tactics
(Smith 1985). If generals are always fighting the last war, it is better for them if
the last war was more recent and a success. A major defeat may eliminate the
state or so emasculate it that future aggression is militarily impractical (e.g., the
Austro-Hungarian and Ottoman Empires after World War I). Second, there
is institutional memory: how decision-makers and the public perceive the ben-
efits and costs of conflict. Success in conflict tends to bolster militarily adven-
turous groups, and recent failure may cause caution. Failure, however, may
encourage a desire to regain position and respect. For example, defeat in
World War I supported German aggression in the 1930s.
The Correlates of War (COW) project is a major effort to overcome a
common cause of the evident failure to understand and better control interna-
tion conflict. By collecting data for more than a century of interstate and civil
wars, the project seeks patterns and commonalities among conflicts and avoids
the historical fallacy that defines each conflict as a discrete and separable event.
If the experience of conflict influences later conflict choices, feedback mecha-
nisms within the state are the likely link. States have memories that influence fu-
ture perceptions and choices. It is conventional wisdom, for example, that a
“Vietnam syndrome” influenced decades of decision-makers, causing an appar-
ent US aversion to armed conflict.
Research has failed to find statistically significant linkages between war ex-
perience and later conflict choices (Singer and Cusack 1981). Analysis of the
COW data showed that none of the usual hypotheses about learning from war
experience is supported: “[T]he probability of the major powers getting into war
is statistically independent of when and with what effects they experienced their
prior wars.” This does not mean that each conflict must be treated as a discrete
event, but it does show that the feedback mechanisms within states are signifi-
cantly more complex than is commonly believed. There is a randomness to the
influences of memory and history that is not captured by simple theories.
Through the concept of emergence (discussed further below), a complex
systems theory of national security potentially allows for both path dependency
through experience (state memory and capability) and randomness. Because
state behavior emerges from domestic interactions, current conditions and in-
stitutions, and the variable distribution of power between politically influential
groups, influence state internal models. But current social conditions and power
relations are themselves historical artifacts. Thus, historical experience is per-
ceived through the ever-changing lens of the present, which itself emerges from
the past.
Rationality
Rationality assumptions are used as a convenient simplification in both orthodox

theories and general systems approaches. But in the latter rationality is assumed
only at the lowest level of aggregation: the individual human. At this level, rational-
ity is consonance of behavior with desires and beliefs about how to achieve those de-
sires within the perceived environment (primarily the social world). In orthodox the-
ories, rationality is more often assigned to higher levels of aggregation. Realism,
liberalism, and constructivism theorize at the level of the international system and
assume that states and other important agents in that system are rational actors.3
This greatly simplifies analysis but is inherently misleading.
For social aggregations rationality has several meanings. It can mean that
the outcome “was (or might turn out to be) desirable, successful, or functional
for the perpetrators” (Singer 1990, 6). But positive feedbacks can magnify the
preferences, and states rarely learn from past wars: “[K]nowledge may be neces-
sary for rational human intervention, but the bloody pages of international his-
tory remind us that it is hardly sufficient” (Singer and Cusack 1981, 417). While
it is claimed to be rational for both parties to defect in the prisoner’s dilemma,
both suffer individually suboptimal outcomes. Rationality also can mean that
the decision process was rational. But successful or functional outcomes may em-
anate from thoroughly irrational processes. Conversely, “the most careful, thor-
ough, and rational process can, with some frequency, culminate in disaster, even
though there tends . . . to be a positive relationship between high rationality in
the process and the desirability of the outcome” (6–7).
Singer draws three conclusions. First, rationality in social aggregations
can only describe the processes they follow, not the outcomes of those
processes. Second, reality is too complex to call behavior rational if agents pur-
sue outcomes that coincide with their individual preferences. President Bush
may have wanted to install democracy in the Middle East through Iraq, but out-
comes are always unpredictable and small causes, like pictures of prisoners tor-
tured by US military police, may derail the most laudable policies. Third, the
rationality of processes must be judged in relation to specific social aggrega-
tions. Even if individuals and groups only minimally respond to their private in-
terests, which may be rational for them, the effect is “extrarational” for the
aggregation. Establishment and celebration of military organization, positive
feedback in elites and media, social rewards to conformity, sunk costs (“they
shall not have died in vain”), an inability to consider all options, and protecting
the individual credibility of leaders are among the factors that make social de-
cision processes extrarational. In the language of complexity, only the individ-
ual can be rational (again, in process only) and all behaviors of social
aggregations emerge from social interactions: “the ‘invisible hand’ of parochial
sub-system interests is ubiquitous, virtually assuring that deviation from rational
choice and the implied prudent pursuit of collective interests will remain the
norm” (Singer 1990, 18). Social system behaviors are neither rational nor irra-
tional. Singer’s terminology is exact: they are extrarational.
Related to the orthodox presumption that social aggregations are rational

is the belief, sometimes implicit, that systems have a purpose or function in a
teleological sense. If systems are emergent, they have no purpose beyond the in-
tentions and preferences of the subsystems. As the subsystems (for example, do-
mestic interest groups) interact in pursuit of their preferences, the system
emerges (legislation is crafted and enacted).4 As Singer comments (1971, 13), so-
cial systems are not “inherently supposed to perform and survive, or seek to do
so.” Social systems collapse when they no longer serve the needs of their con-
stituent agents and the costs of belonging exceed the benefits though authority
permits formal institutions (such as government agencies) to preserve themselves
beyond the limits of social acceptance (Tainter 1988).
While aggregations may not be rational, complexity diverges from gen-
eral systems theory by accepting that a social aggregation like a nation-state may
have preferences and interests. Singer’s general systems taxonomy specifically
“denies that any social entity other than a human being can think, hope, pre-
fer, expect, perceive, and . . . also insists that any social entity can behave
(Singer 1971, 19, emphasis in original). But if the nation-state is conceived as
a meta-agent, it may, like an individual human agent, construct an internal
model of its environment that guides its behavior. Unlike realism, the internal
model is not assumed to be objectively rational and, therefore, relatively un-
changing, but the process by which it is constructed may be assessed for its ra-
tionality. In chapter 5 below, Hoffmann explains how the US internal model
of the ozone issue changed over time, responding to changes in international
negotiations and domestic political interactions.5
DIFFERENCES BETWEEN THE GENERAL AND THE COMPLEX
While the previous section discussed the similarities between several concepts
used in the general systems taxonomy and concepts in complexity, the latter is
more than a warmed-over version of the general systems approach. Several com-
plexity concepts are additional to general-systems concepts and others are modi-
fications of concepts used in the earlier approach. This section highlights the
important differences between general systems theory and complexity.
Emergence
Although the “interaction of individual properties (both within and among sin-
gle humans) may produce emergent effects” (Singer 1968), emergence was re-
jected by most general systems scholars as “unnecessary and scientifically
misleading” (Singer 1971, 18). The effects are “neither structural or behavioral,”
and if they are cultural they can be observed as individual psychological proper-
ties. He rejected the “bromide” that “a social system is more than the sum of
its parts” because the cultural properties of large social systems can be better
described in a “strictly aggregative fashion, by observing the distribution and con-
figuration of individual psychological properties” (Singer 1968, 144). Complex-
ity also relates system culture to individual psychological properties but models
culture as emerging from the interplay of diverse agent internal models within
institutional strictures. Singer also argued that “a system is nothing more than
the sum of its parts and the relationships and interactions among them” (Singer
1971, 19) and that a system is “not composed of [external] systems, or of any other
phenomena beyond its own component units and the relationships and interac-
tions among them.” In complexity a social system is more than merely the aggre-
gation of its parts: the system is modeled as emerging from the relationships and
interactions between member agents. In contrast to Singer (1968), the emergent
effects of agent interactions are both dynamic and important and cannot be cap-
tured by observation of structure, behavior, or individual psychology.
Although the emergent properties of a system cannot be captured by study-
ing the system’s parts, emergence is real. It would be unscientific to reject a the-
oretically useful concept merely because accepted scientific methodologies
cannot record the phenomenon. Fortunately, as elsewhere in the science, devel-
opments of new methods permit new thinking and empirical testing of novel hy-
potheses. With George Zweig, Gell-Mann had developed a theory of quarks. He
was always skeptical that such partially charged subatomic particles would ever be
found, and by the end of the 1960s it seemed that he was right; no evidence had
been found for their existence (Riordan 2004). But by 1973, with new research
techniques, “everything seemed to be coming up quarks.” It took most of the
decade for the theory of quarks to become generally accepted, and it was later
recognized with a Noble Prize. Similarly, one of the predictions of Einstein’s gen-
eral theory of relativity was tested by astronomical observation for the first time
in 2004, nearly ninety years after the theory was first proposed.
Developments in computer modeling now permit simulation of emergent
systems and avoid “metaphysical pursuits” that attempt to isolate and measure
emergence as a definable property of a complex system. When Singer wrote his
critique of emergence, it was not practical to model complex systems. The mod-
eling for the Club of Rome project Limits to Growth was a massive effort on
mainframe computers by specialist programmers at the Massachusetts Institute of
Technology (Meadows et al. 1972). Not only are the predictions of that project
now largely discredited, but also the design of the model, limited as it was by in-
complete data sets and computing power, is considered crude.6 Social scientists
who are not computer specialists can now program ABMs that model “mystical”
emergent properties as an integral part of the whole system. Indeed, the concept
of emergence is central to ABM construction, and social scientists can now
experiment as never before.
Experimentation
The “natural” sciences—most notably physics, mechanics, and chemistry—are still

the measure of “scientific” for most social scientists. Their laboratory experimen-
tation allows repeatable, controlled manipulation of isolated potential causal fac-
tors, a technique rarely available to the social scientist. Field experiments
sometimes are possible, but they permit less control than in the laboratory. Singer
(1977, 3) proposed that “the historical experiment is a perfectly legitimate mode
of research” that may offer advantages over laboratory or field experiments.
Increased control over principal factors in laboratory experiments allows
more accurate observation and measurement, and an increased ability to ascer-
tain covariation, permitting causal inference. In the social sciences, control over
factors may be increased with comparative case studies, statistical manipulation
(as in COW), and simulations. In the “all-machine simulation . . . the magnitude
and variation of every input is fully controlled by the researcher” (Singer 1977).
Inputs may range from the “purely speculative to the thoroughly grounded.”
Computer simulations can test out myriad ideas against history until a good fit
is found. In this way social research replicates the level of control of the classic
laboratory experiment. Atmospheric scientists use computer simulations of cli-
matic history to explain historical and to project future climate change. The de-
velopment of desktop computers as powerful as supercomputers of twenty years
ago, object programming, and concepts such as neural networks now permits (at
least in principle) such historical experiments, as well as more speculative simu-
lations designed to generate as much as to test hypotheses, as discussed further
in chapters 7 through 9.
Cause and Effects
Complex-systems concepts encourage theory that covers multiple levels of analy-

sis, but can it support construction of theories of cause-effect interactions across
levels of analysis? We believe that complexity provides a conceptual framework
for theories that can accommodate both causes working from below and from
above the system under study.
Within a system, emergence connects causes at lower level of analysis with
effects at higher levels of aggregation. As discussed earlier, a state as meta-agent
may form its internal model from the interaction of domestic constituencies and
the interplay among participants in decision-making groups within the executive.

For example, the ability of an individual to shape state policy depends on the in-
stitutional arrangements that regulate the individual’s influence on state behav-
ior and the acceptance of his or her internal model by others in the decision
processes. The chief executive is usually accorded more influence on state policy
than other participants, and his or her internal model will tend to dominate. But
in the process of negotiating policy, internal models may change, especially in
terms of causal beliefs of what is possible. The “butterfly effect” of less institu-
tionally gifted individuals also may emerge up through levels of aggregation to
influence the state internal model.
Beyond the emergence of behavior from internal interactions is the greater
problem of theorizing links between causes at higher levels with effects at lower
levels. The question is: how does environment affect system behaviors? For ex-
ample, how does the state of the international system influence state policies, or
how do national policies determine individual behavior?
Singer (1961) argued that explanation of behavior at each level of analysis was
problematic. For example, explanation from the state level requires several, often
implicit assumptions, the most important of which is that state decisions are not
influenced, even in part, by the perceptions held by individual decision-makers of
the conditions in the system. Yet Jervis (1976), among others, has detailed the many
ways in which individual decision-makers’ perceptions of the state’s environment
are formed and how they influence state interests and behaviors.
In complexity, environment affects system behavior in two ways. First, it con-
strains what is possible and “selects” behaviors that are most appropriate within
current institutional arrangements. Second, perceptions of environment influence
agents’ internal models. And there may be interaction among both processes.
Institutional arrangements in the environment create selection processes
that act on system behavior. If misperception leads to maladapted behavior, the
agent will be “punished,” at a minimum by being prevented from moving toward
satisfaction of its desires. Selection means that agents adapt or are eliminated;
coadaptation implies dynamic recursive adaptive responses between multiple
agents. Agent learning is the cognitive adjustment that increases behavioral sur-
vivability in a selective environment.
Agents “learn” when they change their desires or their beliefs about how
to achieve their desires. The post-9/11 War on Terror is a realignment of pref-
erences in decision-makers’ internal models. Its prosecution through armed at-
tacks on Afghanistan and Iraq and formation of the Department of Homeland
Security was the result of negotiation among decision-makers’ causal beliefs.
Small groups within the state (especially those closely associated to decision-
makers, perhaps energy companies or conservative Christian groups) or outside
(al Qaida or Abu Ghraib prison guards) also may influence the behavior of states
through their impact on the internal models of decision-makers. International
organizations also may influence state choices in several ways, as through infor-
mation dissemination, rules of due process (in the UN Security Council, for ex-
ample), peacekeeping missions, state-building activities, or as the locus of norm
construction and treaty negotiations.
A BETTER TAXONOMY
In his A General Systems Taxonomy for Political Science (1971), Singer sets out six
criteria for a good taxonomy that are equally applicable here. First, it should have
theoretical relevance to the phenomena we hope to account for. And if it can be
relevant to theory for a broad range of phenomena, so much the better. In com-
plexity the dependent variable or outcome is the behavior of social systems; in
world politics it is the emergence of political events (for example, policies and
agent actions) and institutions. But it may be hoped that regularities will be
found among social systems at all levels of aggregation and in all issue-areas of
world politics. The anticipated predictor variables are rules, both internal and
institutional, respectively within and between agents.
Second, knowledge should be transferable between empirical domains.
To achieve this, constructs must be sufficiently abstract to embrace “concepts
that are substantially identical” (1971, 5), allowing for idiosyncrasies of differ-
ent fields. But, where possible, they also should include current knowledge
within the field. By linking concepts from conventional theories to complexity,
this and the previous chapter show that this paradigm can include much that
is known about world politics and, as discussed below, it may integrate views
that are usually considered incompatible.
Third, a good taxonomy indicates what is not known and needs to be
learned. Recognizing the many weaknesses and gaps in our knowledge, a taxon-
omy of world politics should be liberal and allow a range of eclectic approaches
to many different phenomena. Historical experimentation with agent-based
models is a flexible method adaptable across issue areas and levels of analysis.
Fourth, a taxonomy need not be parsimonious, especially if parsimony would
prevent many plausible models: “Parsimony is a virtue only when well advanced to-
ward the verification stage of the discipline and may often be a liability when we are
still in the discovery stage” (6). Fishing expeditions are permissible. One way to
achieve sufficient permissiveness of testable hypotheses “is to develop a minimum
number of classes of variables so that, while many options remain open, the taxon-
omy also remains conceptually clean and manageable” (6). Complexity entails only
a few broad concepts that may be adapted to theory goals. So, complex systems the-
ories may be more parsimonious than competing orthodox theories.
Fifth, coverage of all levels of aggregation is needed: the taxonomy must
“be able to deal with several levels of analysis and . . . the interface between these
levels [must] not be a source of slippage and confusion” (6). We have argued
that, with concepts such as emergence, environment, and internal model, com-
plexity is eminently and uniquely able to satisfy this demand.
Sixth, constructs must be operational. Chapters 3 through 6 of this vol-
ume demonstrate how complex systems concepts can be operationalized in vari-
ous issues and at different levels of analysis. Finally, semantic clarity is essential:
it is “preferable to select words that do convey generally accepted meaning and
then, if necessary[,] specify the restricted or expanded definition intended” (6).
This volume is intended to begin this work.
Encompassing and Improving Orthodoxy
Certainly the first, and perhaps most important, test of a new taxonomy is its
ability to open new research agendas by better integrating existing theories and
knowledge and thereby explaining some of what was previously inexplicable. Or-
thodox theories may be classified in several ways; one useful approach is to dis-
tinguish them by their “view” of reality, which may be external or internal. The
complexity paradigm should support theories that fully or partially integrate
these two apparently incommensurate views.
The external or “scientific” approach assumes that the social world, and
the natural world in which it exists, is an environment, independent of human
agents and potentially predictable (S. Smith 1994; Hollis and Smith 1990). Be-
havior is then assumed to be explicable using methods borrowed from the nat-
ural sciences. The expectation is that there are regularities in behavior that may
be explained by universal causal “laws.” “Behavior is generated by a system of
forces or a structure” (Hollis and Smith 1990, 3) and decision-makers are re-
placeable and only represent their position in the system with little personal vo-
lition. In Singer’s general-systems taxonomy, whole systems are only the sum of
their parts and could, therefore, be disaggregated and comprehended by analyz-
ing the parts and their relations.
Orthodox theories like constructivism pursue an alternative, “inside”
approach that views the social world as constructed of rules and meaning
through human interaction. Each agent tries to pick an intelligent course
through multiple social engagements in which other agents bring their individ-
ual characteristics to their social roles (Hollis and Smith 1990, 6). Here the goal
is to understand behavior, and the means are often hermeneutical, examining

“human action from within, seeing it as intentional and meaningful behaviour”
(S. Smith 1994, 400).
The two approaches usually are assumed to be incompatible. Explanation
from the outside, however scientific, is incomplete without consideration of the
units: “The anarchical character of the international system . . . strongly suggests
that the units affect the shape of relations, however firm the shove [from struc-
tures]” (Hollis and Smith 1990, 198). It also is difficult to see “how the system
changes its structure in a closed system without a change in the units and purely
functional explanations are bound to be suspect, unless they include a causal
contribution from the units.” The inside approach is rejected by some as inter-
pretive and, therefore, inherently unscientific. And the two levels cannot not be
combined to achieve an “overarching theory which explained how system-level
and unit-level factors interacted to produce state behaviour” (100).
Complex systems theories potentially integrate outside and inside ortho-
dox views. The central concept of emergence marks complexity as favoring the
inside view, yet its experimental methods are potentially as scientific as those of
the revered “hard” sciences. The agent/structure problem is a manifestation of
levels of analysis that turns on “the ‘reality’ of systems or on the need to feature
them in explanations” (Hollis and Smith 1990, 197–98). If the system is real and
must be analyzed as a whole, it must be shown that “wholes are more than their
parts and that science is capable of establishing such a proposition” (198). ABMs
can simulate behaviors of whole systems from the inside without consciously in-
terpreting behavior and with no presumption of motives or meaning at any level
by using randomized internal models and rules of interaction. As discussed,
ABMs also may be constructed to simulate historical reality. Only theoretical de-
velopment and empirical and experimental application will demonstrate if world
politics theories based in complexity can overcome the incompatibility of the
two views of orthodoxy and open useful new research agendas in issue-areas. The
following chapters begin that task.
NOTES
This chapter was written by Neil Harrison based on David Singer’s selection from among
his published works of those that anticipate complexity, and his comments and advice on
earlier drafts.
1. This formulation echoes the recursive interaction of biological entity
with its environment that is well accepted in biology and discussed in detail in
Levins and Lewontin 1985.
2. Except for “lost” tribes in places like Amazonia or New Guinea, this
would seem a fair assessment. Autarky, once valued by “developmentistas” (see,
for example, Palma 1978 and Gunder Frank 1969), is probably impossible in the
modern world.
3. In critical theories derived from Marxian analysis, social aggregations of
classes and states are assumed to rationally pursue their interests.
4. The equating of policy-making with sausage production reflects the
inherently extrarational outcome of the policy process.
5. Also see Harrison 2000, which shows how domestic politics influenced
U.S. policies on climate change.
6. Even programming of linear-world models is much more sophisticated,
and the models are more accessible to social scientists not trained in the arcane
tricks of effective computer programming. For example, see Hughes 1996, which
comes with a computer model and relevant data on a CD.
REFERENCES
Deutsch, Karl W. 1953. Nationalism and Social Community. Cambridge, MA:

MIT.
Gunder Frank, André. 1969. Latin America: Underdevelopment or Revolution. New
York: Modern Reader.
Haas, Ernst B. 1964. Beyond the Nation-State. Stanford, CA: Stanford University
Press.
Holland, John H. 1995. Hidden Order: How Adaptation Builds Complexity. Read-
ing, MA: Perseus Books.
Harrison, Neil E. 2000. “From the Inside Out: Domestic Influences on U.S.
Global Environment Policy.” In U.S. Climate Change Policy, ed. Paul G.
Harris, 89–109. New York: St. Martin’s.
———. N.d. “Complex Systems, Social Complexity, and the Environment.” Draft
manuscript.
Hollis, Martin, and Steve Smith. 1990. Explaining and Understanding International
Relations. Oxford: Clarendon Press.
Hughes, Barry B. 1996. International Futures: Choices in the Creation of a New
World Order. 2nd ed. Boulder: Westview.
Jervis, Robert. 1976. Perception and Misperception in International Politics. Prince-
Kaplan, Morton A., 1957. System and Process in International Politics. New York:
Wiley.
Levin, Simon A. 1999. Fragile Dominion: Complexity and the Commons. Reading,
MA: Perseus Books.
Levins, Richard, and Richard Lewontin. 1985. The Dialectical Biologist. Cam-
bridge, MA: Harvard University Press.
Mäler, Karl-Göran. 2000. “Development, Ecological Resources and Their Man-
agement: A Study of Complex Dynamic Systems.” European Economic
Review 44 (2000): 645–65.
Mitrany, David. 1966. A Working Peace System. Chicago: Quadrangle.
Palma, Gabriel. 1978. “Dependency: A Formal Theory of Underdevelopment
for the Analysis of Concrete Situations of Underdevelopment.” World
Development 6:881–924.
Patomäki, Heikki, and Colin Wight. 2000. “After Postpositivism? The Promises
of Critical Realism.” International Studies Quarterly 44, no. 2 (June):
213–37.
Prigogine, Ilya. 1997. The End of Certainty: Time, Chaos, and the New Laws of Na-
ture. In collaboration with Isabelle Stengers. New York: Free Press.
Riordan, Michael. 2004. “Science Fashions and Science Facts.” Viewed at
http://www.physicstoday.org/vol-56/iss-8/p50.html on March 22.
Singer, J. David. 1961. “The Level of Analysis Problem in International Rela-
tions.” In The International System: Theoretical Essays, ed. K. Knorr and
S. Verba, 77–92. Princeton, NJ: Princeton University Press.
———. 1968. “Man and World Politics: The Psycho-Cultural Interface.” Journal of
Social Issues 24, no. 3:127–56.
———. 1970. “Escalation and Control in International Conflict: A Simple Feed-
back Model.” General Systems 15:163–73.
———. 1971. A General Systems Taxonomy for Political Science. New York: General
Learning.
———. 1977. “The Historical Experiment as a Research Strategy in the Study of
Politics.” Social Science History 2, no. 1 (Fall): 1–22.
———. 1990. “The Invisible Hand, Extra-Rational Considerations and Decisional
Failure.” Paper presented at the Annual Convention of the American
Political Science Association, San Francisco, September 2.
Singer, J. David, and Thomas Cusack. 1981. “Periodicity, Inexorability, and
Steermanship in International War.” In From National Development
to Global Community: Essays in Honor of Karl W. Deutsch, ed. Richard

L. Merritt and Bruce M. Russett, 404–22. Boston: George Allen
& Unwin.
Smith, Merritt Roe. 1985. Military Enterprises and Technological Change. Cam-
bridge, MA: MIT.
Smith, Steve. 1994. “Rearranging the Deckchairs on the Ship Called Modernity:
Rosenberg, Epistemology and Emancipation.” Millennium: Journal of
International Studies 23, no. 2:395–405.
Sterling-Folker, Jennifer. 2002. “Realism and the Constructivist Challenge:
Rejecting, Reconstructing, or Rereading.” International Studies Review 4,
no. 1 (Spring): 73–97.
Tainter, Joseph A. 1988. The Collapse of Complex Societies. Cambridge: Cambridge
University Press.
Wallerstein, Immanuel. The Capitalist World Economy. Cambridge: Cambridge
University Press, 1979.
Waltz, Kenneth. 1979. Theory of International Politics. Reading, MA: Addison-
Wesley.
CHAPTER 3
Complexity and Conflict Resolution
Dennis J. D. Sandole
The events of September 11, 2001, undermined much conventional analysis

in world politics and international relations (IR). Much as the fall of the Berlin
Wall was not anticipated by IR scholars, the terrorist attacks in New York and
Washington did not neatly fit within conventional explanations of international
conflict. In this paper, therefore, I attempt to (1) respond theoretically and prag-
matically to the events and aftermath of September 11, 2001; (2) deal with
Realpolitik (and one of its concomitants, ethnocentrism) and conflict resolution as tra-
ditionally contending, but potentially complementary, approaches to dealing with
threats to order and security at the domestic and international levels; and (3) pro-
vide a theoretical and pragmatic basis for further research, theory building, and
practice in domestic and world affairs, with a view to dealing effectively with both
the deep-rooted causes and the very clear symptoms of the “new” terrorism and re-
lated identity-based conflicts fueled by the ending of the Cold War. To make sense
of the attacks and to illustrate the potential complementarity of Realpolitik and
conflict resolution approaches, I draw on several complexity concepts, including
emergence; nonlinear, “catastrophic” responses to initial conditions; and syner-
gistic coexistence of traditionally competing frameworks and ideas.
REALPOLITIK
Realpolitik is the traditional power paradigm governing efforts to manage the

uncertainty and disorder inherent in “Hobbesian space.” At its most virulent
extreme, it is expressed as dictatorship domestically and as imperialism inter-
nationally, with all the attendant manifestations of structural, cultural, and
43
44 DENNIS J. D. SANDOLE
physical violence—including genocide—implied by the defense and perpetuation

of a preferred status quo at the expense of those who do not benefit from it (see
Galtung 1969, 1996).
Realpolitik has a long lineage, going back in recorded history to at least 416
BC, the midpoint of the Peloponnesian War, when Athens attempted to nego-
tiate control over the neutral island state of Melos, a situation chronicled elo-
quently by Thucydides:
[S]ince you know as well as we do that right, as the world goes, is only in
question between equals in power, while the strong do what they can and
the weak suffer what they must . . . the contest not being an equal one . . .
but a question of self-preservation and of not resisting those who are far
stronger than you are . . . of men we know, that by a necessary law of their
nature they rule wherever they can. And it is not as if we were the first to
make this law, or to act upon it when made: we found it existing before us
. . . all we do is make use of it, knowing that you and everybody else having
the same power as we have, would do the same as we do . . . it is certain
that those who do not yield to their equals, who keep terms with their su-
periors, and are moderate towards their inferiors, on the whole succeed
best. (1951, 331–36)
This is clearly an old story, which has been repeated thousands of times up
to the present day, with Hans Morgenthau (1973, 4) being one of the more “re-
cent” successors to Thucydides and reminding us all about the “laws” that gov-
ern human behavior to Realpolitik effect. He says,
• Political realism believes that politics, like society in general,

is governed by objective laws that have their roots in human
nature.
• Human nature, in which the laws of politics have their roots,
has not changed since the classical philosophies of China,
India, and Greece endeavored to discover these laws.
In other words, for Morgenthau and other realists, human nature—which

makes “statesmen think and act in terms of interest defined as power” (1973, 5)—
has not changed since Thucydides made his observations in 416 BC. Hence, the
“key concept of interest defined as power is a objective category which is univer-
sally valid” (8). In the modern Westphalian world, power as interest is usually re-
served for the protection of the nation-state, but it has also been used in defense
of the tribe and the ethnic group.
COMPLEXITY AND CONFLICT RESOLUTION 45
Ethnocentrism
Ethnocentrism is a natural corollary of Realpolitik: power is used by the privileged

to maintain themselves and their groups at the expense of others. According to
William Graham Sumner (1906), who coined the term, ethnocentrism is
the technical name for this view of things in which one’s own group is the
center of everything, and all others are scaled and rated with reference to
it. . . . Each group nourishes its own pride and vanity, boasts itself supe-
rior, exalts its own divinities, and looks with contempt on outsiders. . . .
the most important fact is that ethnocentrism leads a people to exaggerate
and intensify everything in their own folkways which is peculiar and which
differentiates them from others. It therefore strengthens the folkways.
(quoted in LeVine and Campbell 1972, 8).
Sumner also generalized “that all groups show this syndrome.” In other
words, according to him and extensive research carried out by Henri Tajfel
(1978, 1981) and others, ethnocentrism—following Thucydides’ and Morgen-
thau’s characterizations of Realpolitik—is the universal tendency for humans to
divide humankind into two groups: “them” and “us.” The criteria for doing
so are not fixed and can be based on, among other things, nationality, eth-
nicity, religion, race, class, region, or gender—criteria for which Realpolitik can
mobilize defenses.
Accordingly, ethnocentrism enhances intragroup community, especially
under threat from out-groups (see Simmel 1955; Coser 1956), and in-group eth-
nocentrism works against intergroup community. Indeed, it is safe to say that, es-
pecially within a Realpolitik frame, ethnocentrism makes for a zero-sum
relationship between peace at the intragroup level and war at the intergroup level.
Again, according to Sumner (1906):
The insiders in a we-group are in a relation of peace, order, law, govern-

ment, and industry, to each other. Their relation to all outsiders, or others-
groups, is one of war and plunder, except so far as agreements have
modified it. . . .
The relation of comradeship and peace in the we-group and that of hos-
tility and war towards others-groups are correlative to each other. The exi-
gencies of war with outsiders are what make peace inside, lest internal
discord should weaken the we-group for war. . . .Thus war and peace have
reacted on each other and developed each other, one within the group,
the other in the group relation. (quoted in LeVine and Campbell 1972,
7–8, emphasis added)
So pervasive is the tendency to subdivide humanity in this way, even under

minimal intergroup differences, that fascinating experiments have been conducted
with children, with the resulting “them-us” hostility between the contrived groups
bordering on the remarkable. In their famous “Robbers Cave” experiments, for
example, Sherif and Sherif (1953) were able to stimulate the development of hos-
tile relationships between two groups of boys who had originally been friendly
members of one and the same group. And in the famous (or to some, “infamous”)
Blue Eyes/Brown Eyes exercises conducted by Jane Elliott, originally with her
fourth-grade pupils in Riceville, Iowa, shortly after Rev. Dr. Martin Luther King
had been assassinated:
Elliott told her children that brown-eyed people were superior to blue-
eyed, due to the amount of the color-causing chemical, melanin, in their
blood. She said that blue-eyed people were stupid and lazy and not to be
trusted. To ensure that the eye color differentiation could be made
quickly, Elliott passed out strips of cloth that fastened at the neck as col-
lars. The brown eyes gleefully affixed the cloth-made shackles on their
blue-eyed counterparts.
Elliott withdrew her blue-eyed students’ basic classroom rights, such as
drinking directly from the water fountain or taking a second helping at
lunch. Brown-eyed kids, on the other hand, received preferential treat-
ment. In addition to being permitted to boss around the blues, the browns
were given an extended recess.
Elliott recalls, “It was just horrifying how quickly they became what I told
them they were.” Within 30 minutes, a blue-eyed girl named Carol had re-
gressed from a “brilliant, self-confident carefree, excited little girl to a
frightened, timid, uncertain little almost-person.”
On the flip side, the brown-eyed children excelled under their new-
found superiority. Elliott had seven students with dyslexia in her class that
year [1968] and four of them had brown eyes. On the day that the browns
were “on top,” those four brown-eyed boys with dyslexia read words that
Elliott “knew they couldn’t read” and spelled words that she “knew they
couldn’t spell.” (Kral 2000, 2, emphasis added)1
Elliott conducted the original exercise “to demonstrate to her fourth-grade

students how harmful the myth of White superiority is and what, as a result of
this myth, it meant to be Black in America” (1). Since then, she has appeared on
television, e.g., The Tonight Show, Oprah Winfrey Show, and PBS’s documentary
Frontline (see PBS 1985) in the United States. She has also gone on to conduct
these exercises for adults, including police officers, around the world, letting
them “experience” for themselves that
We learn to be racist, therefore, we can learn not to be racist. Racism is not

genetical. It has everything to do with power. (quoted in Coronel 1996, 2,
emphasis added)
In the language of complexity, Elliott has demonstrated with a simple experi-

ment how mental models are socially constructed and can adapt to interactions
with others, especially those with authority.
Related to the Realpolitik-ethnocentrism nexus is the seductive totality and
simplicity of the clash of civilizations idea of Samuel Huntington (1993, 1996),
and earlier of Benjamin Barber’s (1992) jihad. Since September 11, 2001, some
commentators have been using the “clash” or “jihad” to characterize the polariz-
ing global relationship between Judaic-Christian and Islamic “civilizations”:
probably the ultimate global expression of “us-them” hostility in the history of
humankind. This idea may become, self-fulfillingly, more fact than fiction: a de-
velopment that, especially if accompanied by nuclear weapons, would certainly
not be in the interests of the United States or anyone else.
Sources of Ethnocentrism
Members of the conflict/conflict resolution community tend to be humanists,

liberals associated with flexible, optimistic views of human nature. They tend
to agree with Albert Bandura (1973) and others that whatever humans do in
conflict situations is a function of learning: change what they learn and change
their behavior! This view, which is associated with the Idealpolitik paradigm
(see Sandole 1999a, 110–13), tends to ignore—on ideological, political, emo-
tional and practical grounds—the role of biology in conflict, especially violent
conflict behavior.
Elsewhere (1990), I have argued that biological factors play a role in
human behavior—as part of a complex constellation of social, political, economic,
and other factors—and adherents to that view can now “come out of the closet”
without fear of being ostracized as purveyors of Nazi eugenic philosophies and
programs. I have also argued (1999a, 180–85) that it is not simply a question
of “nature” or “nurture,” or indeed, in some simple additive sense, of “na-
ture” and “nurture,” but of both interacting in complex ways, such that each
may be affected by the other.2 Given the observations of, among others, Ed-
ward O. Wilson (1979), John Pfeiffer (1984), and Joseph Montville (1988)
that our brains seem to be preprogrammed to bifurcate everything, including
fellow human beings, into membership in in-groups and out-groups, it seems
reasonable to conclude that “nature” has invested Homo sapiens with this
particular kind of “hard wiring” to protect us from one another in Hobbes’s

infamous “state of nature,” where
men live without a common Power to keep them all in awe, . . . in that
condition which is called Warre; and such a warre as is of every man
against every man . . . where the life of man [is] solitary, poore, nasty,
brutish, and short. (Hobbes 1950, 103, 104)
Clearly, learning, culture, and other aspects of “nurture” can impact sig-
nificantly whom an agent defines as “threatening,” and how he or she responds
to them; but the biological predisposition to bifurcate fellow members of the species
into “them” and “us” nevertheless seems to be there, ready to interact with cul-
ture to create certain “histories,” certain “facts on the ground,” that then become
the bases of violent conflict spirals, including the genocidal ethnic cleansing that
has returned to Europe in the wake of the ending of the Cold War.
In this regard, R. Paul Shaw and Yuwa Wong (1989) argue that the moti-
vations that predispose human beings toward defense of their in-groups are part
of “human nature”; that is, the “seeds of warfare” lie in ultimate (in contrast to
proximate) causes—inclusive fitness and kin selection.
Inclusive fitness has two parts: (1) “increased personal survival and in-
creased personal reproduction (classical Darwinian fitness)”; and (2) “the en-
hanced reproduction and survival of close relatives who share the same genes by
common descent (a kinship component)” (Shaw and Wong 1989, 26).3
Kin selection “implies that assistance, favors or altruism would be directed
at individuals who were genetically related enough to give the common gene
pool greater survival advantages. Genetic relatedness would be greatest with
members of one’s lineage and one’s own kin or nucleus ethnic group” (Shaw and
Wong 1989, 27).
Here we have the crux of the matter concerning ethnocentrism for evolu-
tionary psychologists: “[P]roviding an ultimate, evolutionary rationale for coopera-
tion and civility among genetically related individuals also provides an ultimate
rationale for anticipating origins of reduced cooperation among less related indi-
viduals” (41, emphasis added).
This amounts to a “sociobiology of ethnocentrism” (44–45, emphasis added)
underpinning “we-them” distinctions, including those as framed in the “clash of
civilizations.” Perhaps the ultimate example of complexity in human affairs is that
Humans have outfoxed themselves. They have learned to maximize inclu-

sive fitness—through ethnocentrism, out-group enmity, nationalism and
patriotism—to the extent that they have created the means to destroy the
very inclusive fitness they seek to foster and protect. . . . unless some kind
of action is forthcoming . . . there is no reason to believe that Homo sapiens

will escape nuclear devastation, if not extinction. (197)
CAN “NURTURE” INFLUENCE

THE “NATURE” OF ETHNOCENTRISM?
Concerned members of the international community could join with Jane El-
liott and start to teach children in the schools, not that racism, anti-Semitism,
and other isms are “normal,” but that they are learned, oftentimes dysfunctional ex-
pressions of our biological predisposition to bifurcate people into friend and foe.
Given that the predisposition is part of our “wiring,” that is, originally meant to
have survival value, we are sort of stuck with it. We are not, however, stuck with
the culturally/experientially determined referents of that predisposition. Indeed,
as implied, some of those definitions may be counter to our survival, either as
members of in-groups or as an entire species. Hence, it would be in our best in-
terests to work on changing those definitions, and on changing the mental mod-
els through which individuals comprehend the world around them and in terms
of which they choose their behaviors.
Imagine classrooms at all levels, up to university level, where pupils and
students are actively encouraged, by conflict resolution–trained facilitators, to
brainstorm the kinds of emotions they experience when they think about, dis-
cuss, or interact with members of certain groups. (They would thereby make it
exceedingly difficult to do what Roger Fisher and William Ury (1983, chap. 2)
counsel: to “separate the people from the problem.”) They would brainstorm
where those feelings come from, the consequences of those feelings, examples
throughout the country and the world where those kinds of feelings have trans-
lated into violent conflict situations, how to work on changing those feelings, dif-
ficulties in doing so, and so on.
This is a complex tall order: the feelings that we experience have a “natural”
base; they are, therefore, part of our “human nature.” However, the culturally de-
fined targets of those feelings are not part of our nature: they may be wrong, un-
fair, self-fulfillingly counterproductive and dangerous and, therefore, should
be—and can be—changed!
This would be quite a challenge to bring into any level of classroom, but
it is a necessary one if we are to make a dent on the levels of violence that have,
for example, brought genocide back to Europe, motivated nineteen young Arab
Muslim males with box cutters to turn passenger-filled aircraft into cruise mis-
siles against the World Trade Center and the Pentagon, or turned the United
States into the most violent country in the industrialized world (see Sandole
1999a, 4). Recent examples of its violence include the Washington, DC–area
sniper incident (October 2–24, 2002) and the murder of three professors at the
University of Arizona by a failing student (October 28, 2002); and as Stepp
(2002, A3) points out, “The homicide rate for U.S. infants . . . [is] now virtually
equal to the murder rate for teenagers, according to a new analysis of govern-
ment data that revealed a surprising demographic milestone.”
But what about the adults, some of whom may be killing their kids (A17)?
To what extent can nurture close the nature-nurture gap of ethnocentrism for
them, especially in the post-9/11 world?
The War on Terror is currently being waged within a Realpolitik frame-
work, again, elevating the level of analysis to a more global version of us-them
hostilities. President George W. Bush’s strident declaration that “you are either
with us or the terrorists” has radicalized Muslims all over the world. It has also
made many Americans feel a closer sense of community, but at the expense of
the security of many American Arabs and Muslims, who feel threatened and vic-
timized by governmental security services as well as by purveyors of hate crimes
(see Pierre 2002a, 2002b). In other words, we are returning to the dangerous
simplicity of a bipolar world, where, given the Bush administration’s continu-
ing war rhetoric to keep the patriotic fervor flowing beyond the ebbing impact of
military successes in Afghanistan and Iraq, and the worsening insurgency in Iraq,
the Realpolitik-ethnocentrism nexus and the intragroup peace versus intergroup
war dynamic are taking on a more global, civilizational, “jihadic” character.
Hence, although Realpolitik has been conceived as a rational approach to
the defense of individual and national interests, it has, in practice, tended to be-
come more a part of the problem than of the solution: more and more it has
been revealed to be a significant source of self-stimulating and self-perpetuating con-
flict systems (see Vasquez 1993; Sandole 1999a).
Complex Systems
Complexity offers insights in this regard. One of its major assumptions is that,
among other things, everything is connected to everything else (see Waldrop
1992). Accordingly, any attempt at problem solving must be at least multi- if not
interdisciplinary. But few of us have been educated that way: we receive our de-
grees usually in only one discipline, and therefore we as analysts may also be
more a part of the problem than we are of the solution.
Conflict Resolution
As an interdisciplinary field, conflict resolution is intuitively similar to complexity

and provides a conceptual basis for capturing the complexity of complex con-
flicts. Part of its appeal is that it does not replace Realpolitik as such, because on
occasion we need the military to prevent or stop atrocious acts of violence such
as the genocidal conflicts of recent times in Rwanda and Bosnia. Also, we could
have used our police or military as armed marshals on board the four hijacked
aircraft on Tuesday, September 11, 2001, to prevent the planes crashing into
their intended targets. However, to be effective in the long run, Realpolitik must
always be included in a larger frame, a metaparadigm, where it coexists and co-
evolves with, for example, Idealpolitik, Marxism, and something I call “non-
Marxist radical thought,” which focuses on basic human needs (see Sandole
1993; Sadole 1999a, 110–13, 117–20, 137–40).
COMPLEX SYSTEMS AND CONFLICT STUDIES
Given small differences in the start-up conditions of biological, economic, physi-

cal, and other systems, the consequences may be catastrophically or otherwise rad-
ically different.4 To paraphrase Heisenberg’s uncertainty principle from quantum
mechanics (see Nagel 1961, 293–305), in such cases neither analysts nor policy-
makers would be able to predict a system’s behavior with unlimited precision.
Nevertheless, there would also be discernible patterns underlying chaos, thereby
keeping alive the possibility of prediction. The distributed decision-making in
complex systems and their consequent dynamism and tendency toward nonlin-
earity make them unpredictable: though patterns may hold for a period of time,
their sensitivity makes them liable to change out of all proportion to any stimulus.
As with many, if not all, innovations in thought, complexity had been
around awhile before it was conceptualized as such (see Saperstein 1995). For
instance, Kenneth Boulding remarked that in conflict analysis and resolution,
Human beings are moved not only by immediate pressures but by distant
goals that are contemplated in the imagination. These goals are susceptible
of change, often of dramatic change, as a result of apparently slight changes in
current information. On the other hand, they also have a good deal of sta-
bility, and this gives a stability to the system in the large that it may not
have in the small. (1962, 24, emphasis added)
Lewis F. Richardson’s (1939, 1960a) work on the dynamics of an arms

race is another source of ideas in orthodox conflict studies that show similari-
ties with those of complexity: in a dyadic relationship, depending upon each
actor’s sensitivities to the other’s arms levels (mutual fears), plus the constraints
of each on further arms spending (limiting factors), and underlying grievances,
there could be a stable balance of power with regard to “rate[s] of rearmament or
disarmament”; or there could be an unstable equilibrium in which either complete

disarmament or a runaway arms race is possible. There could also be radical shifts
between stable and unstable systems (in either direction), “for relatively modest
variations in . . . assumptions” regarding mutual fears, limiting factors, and griev-
ances (Nicholson 1989, 152). Or, within the unstable condition, there could be
radical shifts from complete disarmament to a runaway arms race (or vice versa),
resulting from “small shift[s] in the position of the initial point” of armament ex-
penditures at “time zero” (152; also see Boulding 1962, chap. 2; Rapoport 1960,
chap. 1; and Saperstein 1995).5
COMPLEX SYSTEMS AND CONFLICT RESOLUTION:

MANAGING ENTROPIC CONFLICT SYSTEMS
IN THE POST–COLD WAR WORLD
It is clear that conflict researchers and policy-makers cannot predict with cer-
tainty what kinds of conflicts-as-process will emerge from various kinds of conflicts-
as-startup conditions (see Sandole 1999a, 129–31), or predict the course of any
particular conflict-as-process. The danger in this, of course, is that conflict re-
searchers may be paralyzed into recommending nothing and policy-makers para-
lyzed into doing nothing, or at least nothing of major significance: witness
Bosnia and Herzegovina, at least up to the Dayton Peace Accords of October–
December 1995.
But we should be fair: the danger of paralysis derives from the possibil-
ity that conflicts-as-process could, unpredictably, and because of very small
shifts in existing conditions, escalate out of control (a continuing risk in Iraq).
In other words, beyond some threshold, conflicts-as-process could escalate
into self-stimulating/self-perpetuating spirals, where attempts to deal with them
could backfire, leading to destruction of the conflict systems themselves. In
such cases, we can talk of entropic conflicts: conflicts that approach entropy, or
progressive disorder.
The danger that, unpredictably, conflicts can assume an entropic charac-
ter (as Iraq may already have)—what Gregory Bateson (1973, 98) refers to as a
schismogenic “regenerative causal circuit or vicious circle”—is implicit in Realpoli-
tik: the use of a “measured” amount of force, even as part of an Idealpolitik strat-
egy to achieve negative peace as a necessary (but not sufficient) condition of
positive peace, could backfire, making matters worse. This may explain why, with
the exception of the NATO bombing campaign that, in part, led to the Dayton
Peace Agreement, Robert Axelrod’s (1984) fascinating theory of cooperation has
not been applied to Bosnia.
Axelrod has argued that in all situations involving the prisoner’s

dilemma—a classic game of Realpolitik analysis and prescription that applies to
the interpersonal as well as international levels—the best way to act is in terms of
the TIT FOR TAT strategy:
TIT FOR TAT’s robust success [in prisoner’s-dilemma situations] is due to

being nice, provocable, forgiving, and clear. Its niceness means that it is
never the first to defect, and this property prevents it from getting into un-
necessary trouble. Its retaliation discourages the other side from persisting
whenever defection is tried. Its forgiveness helps restore mutual coopera-
tion. And its clarity makes its behavioral pattern easy to recognize; and
once recognized, it is easy to perceive that the best way of dealing with TIT
FOR TAT is to cooperate with it. (176, emphasis added, also see 54)
For TIT FOR TAT to work, however, “the future must have a sufficiently large
shadow”; that is, it “requires that the players have a large enough chance of meet-
ing again and that they do not discount the significance of their next meeting
too greatly” (174).
Extending Axelrod’s theory to the wars in former Yugoslavia during
1991–95 leads to the following scenario:
1. Slovenian and especially Croatian declarations of indepen-

dence from the Yugoslav Federation in June 1991 resurrected
Serbian fears (especially among Serbs living in Croatia) of Croa-
tian defection from the stable TIT FOR TAT equilibrium that had
existed up to that point.
2. Serbian military successes, plus the “nonprovocability” of the
international community, stimulated the development and ex-
acerbation of a violent, asymmetrical conflict-as-process, that is,
“ethnic cleansing,” which was prosecuted by the Serbs against
the major victims of the wars in former Yugoslavia, Bosnian
Slavic Muslims.
3. In the absence of the “provocability” of the Bosnian Slavic Mus-
lims, the international community was effectively shamed into
becoming “provocable” and retaliating against the Serbian
“defection” from the previously stable TIT FOR TAT equilibrium,
although in a very restrained way (as in the live-and-let-live sys-
tem of trench warfare during World War I; see Axelrod 1984,
chap. 4). Subsequently, the international community was “for-
giving” toward the Serbs to avoid stimulating new or exacerbat-
ing ongoing violent conflict spirals.
4. The international community embarked on a “train-and-equip”

program for a joint Bosnian Muslim–Croat army (see Pomfret
1996a) so that Bosnian Muslims in particular could, in the fu-
ture, become appropriately “provocable.”
In terms of this analysis, the “provocability” of the international commu-

nity (more so than of the Bosnian Muslims) was the issue. Until Dayton, the in-
ternational community had not been sufficiently “provocable” in Bosnia and
Herzegovina, perhaps because of paralysis associated with the unpredictability of
the consequences of even minor adjustments in complex systems capable of gen-
erating entropic conflict processes. As Michael Lund (1996, 111) puts it: “From
1990 into 1992, it may be remembered, a major obstacle to European and U.S.
involvement in the Yugoslavian imbroglio was considerable uncertainty as to the
wider ramifications of the gathering storm” (emphasis added).
But even with Dayton, the “provocability” of the international community
remained an issue; for instance, during the summer of 1996 demands by the in-
ternational community and threats of sanctions were followed by vague promises
by the Bosnian Serbs to comply, and their failure to do so was followed by a
breakdown on sanctions. This only emboldened the Serbs. “It was, according to
many western diplomats, a humiliating retreat and one that was greeted with ju-
bilation in the self-styled Republic of Srpska” (Hedges 1996b).
While Serbs celebrated the first “anniversary” of the fall of the UN “pro-
tected safe area” of Srebrenica, war crimes investigators were sorting through the
remains of men and boys captured and shot after the Muslim enclave fell. Serbs
marked their victory . . . and reiterated their goal of keeping the territory “ethni-
cally pure.”6
Taken together with the observation that the conditions specified by Day-
ton for “free and fair elections” in Bosnia—freedom of movement, freedom of ex-
pression, freedom of press, and freedom of association—had not been met, even
though the Organization for Security and Cooperation in Europe (OSCE) de-
clared that national elections could nevertheless take place on September 14,
1996 (see Hedges 1996a), and municipal elections a year later, then it is clear
that, nearly one year following the cessation of hostilities in Bosnia, with the ex-
ception of the NATO bombing campaign leading up to Dayton Axelrod’s theory
remained basically untried and untested in former Yugoslavia.
Paul Stern and Daniel Druckman (1994/1995, 114) view Axelrod’s theory
as an example of the strongest evidence of the “hegemonic position of realism”
in US international relations thinking and practice, because it effectively legiti-
mates cooperation within the Realpolitik paradigm. Axelrod’s theory is certainly
appropriate for Realpolitik-defined realities, such as the wars in former
Yugoslavia. In Bosnia and Herzegovina, for instance, Anthony Lewis (1996, 11)
had concluded that “The only thing that ever moved the Bosnian Serbs to more
than empty promises during the war [there] was force.” Also, a Bosnian Serb of-
ficial in the city of Brcko characterized the Dayton Peace Agreement’s call for the
return of refugees as “a clear attempt to change the biological structure of the
city.” He went on to assert, with Muslim refugees in mind, that “We will defend
our frontiers biologically” (Dobbs 1996a, emphasis added), thereby implying a
continuation of the doctrine and practice of “ethnic cleansing.”
But of the four elements of TIT FOR TAT, only two—provocability and clar-
ity—reflect Realpolitik as such. TIT FOR TAT’s other two elements—niceness and for-
giveness—locate it in a more “complex” constellation of options, very much like
that suggested by Stern and Druckman’s own “contours of a new paradigm”
(1994/1995, 115–17) and by my own “4 ⫹ 2 framework,” which combines
Realpolitik, Idealpolitik, Marxist, and non-Marxist radical definitions of reality,
plus cooperative and competitive means for dealing with conflict (see Sandole
1999a, 110–13). This “complex” orientation shares with Fisher and Keashly’s
(1991) “contingency model” the prescription of using what is necessary under one
set of conditions, but of using other tools as well when those conditions have
changed (also see Fisher 1993; Fisher 1997, chap. 8). Indeed, TIT FOR TAT is a re-
sponse to “complexity”: it can encourage, through learning, the development of
cooperation out of the “coevolutionary dance of competition and cooperation”
(see Waldrop 1992, 259–60, 262–65, 292–94). TIT FOR TAT is an example of a
process of agent interactions in which each agent learns to coevolve with others.
But for TIT FOR TAT to be ultimately successful, there must be, in addition
to a “sufficiently large shadow” of the future (which, admittedly, ethnic cleansing
had eroded), stability in the sense of Richardson’s (1939, 1960a) “balance of
power”—another Realpolitik aspect!—between the “coevolving” parties in their re-
spective capabilities to inflict pain on each other. Unless a stable balance exists,
the parties may engage in what Lewis Coser (1956, 136) refers to as a “trial by or-
deal,” in which “conflict may be an important balancing mechanism” designed
to achieve the very equilibrium that may be absent to begin with:
Conflict consists in a test of power between antagonistic parties. Accom-

modation between them is possible only if each is aware of the relative
strength of both parties. However, paradoxical as it may seem, such knowl-
edge can most frequently be attained only through conflict, since other
mechanisms for testing the respective strengths of antagonists seem to be
unavailable.
Consequently, struggle may be an important way to avoid conditions
of disequilibrium by modifying the basis for power relations. (137,
emphasis added)
Apropos less lethal forms of conflict handling (e.g., mediation or arbitra-

tion), Coser tells us that such “Efforts . . . encounter the difficulty that the assess-
ment of the actual power relations between the contenders can hardly be made
before their relative power has been established through struggle” (135–36).
The US-led effort to arm the Bosnian Muslims was designed to “make pos-
sible a reassessment of relative power and thus serve as a balancing mechanism
which helps to maintain and consolidate societies” (137) and to provide a mate-
rial basis for increased Muslim “provocability,” especially in the relative absence
of such on the part of the international community, thereby establishing a stable
balance of power and ensuring that TIT FOR TAT succeeds in Bosnia without fur-
ther international intervention. Indeed, according to James Pardew, the official
originally in charge of the US program, the weapons “would be used for Bosnia’s
defense and would contribute to stability in the region. The purpose of the train-
and-equip program [therefore] is to prevent war by creating a military balance in
Bosnia” (Pomfret 1996b, emphasis added; also see USIP 1997).
There is, however, a problem with “balance of power,” as there is with
Realpolitik in general. As Shaw and Wong (1989, 47) imply, “trials by ordeal” to
determine “relative strength,” as manifested in former Yugoslavia, are associated
with “groups as forces of selection [that] represent an emergent, proximate,
environmental cause [of war]”:
Since failure to maintain a balance of power could have resulted in ex-

tinction, groups and their expansion figure as forces of selection in our the-
ory. Motivated by resource competition, conflict, and warfare, struggles to
maintain balances of power [have given] rise to more complex societal
units which [have] continued the legacy of intergroup warfare. . . . It is by
this process that out-group enmity and ethnocentrism have been rein-
forced and carried over from nucleus ethnic group to band, to tribe, to
chiefdom, to nation-state. (45)
Applying complexity concepts, therefore, involves more than stable bal-

ances associated with negative peace of an enforced temporary respite from violent
conflict; it also involves building upon and transcending these and, in positive
peace fashion, exploring the dynamic of deep-rooted processes and conditions
that make, in the shorter run, the balances a “natural” consequence of social
processes (see Galtung 1969, 1996). Seven years after the U.S. intervention,
Bosnia was unable to function as a sovereign state. It depended on a North At-
lantic Treaty Organization (NATO) force of twelve thousand foreign troops, and
the functions of government relied on Western representatives with sweeping
powers. And the same nationalist parties that incited the conflict had been
reelected (WP 2002). In December 2004, the NATO force was replaced by a
British-led EU peace keeping force of some 7,000 personnel.
The problem with the Dayton Peace Agreement for Bosnia is not only that
the physical and emotional reconstruction of the country (positive peace) has
lagged behind the enforced prevention of violence (negative peace), but also—
with provocability still an issue—that the negative peace is not a stable one. TIT
FOR TAT, therefore—and with it, complexity in general—still remain to be fully ap-
plied to Bosnia.
THE “COMPLEXITY” OF COMPLEX SYSTEMS
As implied thus far, Realpolitik philosophers, theorists, and practitioners tend to re-
spond to the disorder, unpredictability, and insecurity inherent in “Hobbesian
space” by advocating and/or pursuing the enhancement of predictability, regularity,
and stability (the “PRS needs,” see Sandole 1984)—and, therefore, of order and secu-
rity—in their domestic and international environments through the simplistic bifur-
cation of the species into “them” and “us” and by the threatened or actual use of
force against “them” whenever circumstances within the Realpolitik/ethnocentric
frame call for such. Hence, Kenneth Waltz’s (1964) earlier defense of a “bipolar” in-
ternational system as inherently more conducive to stability than a multipolar sys-
tem; John Mearsheimer’s (1990a, 1990b) lamenting of the end of the Cold War
and its simplicity; and following the terrorist attacks of September 11, 2001, Presi-
dent George W. Bush’s strident declaration to the entire world that “[y]ou are either
with us or the terrorists!” This desire for simplicity in the face of real complexity is
the foundation of most conventional perspectives on conflict and its resolution.
At the extreme “right-wing” end of the Idealpolitik-Realpolitik continuum,
we find authoritarians who have a low threshold for uncertainty and insecurity
and who, therefore, tend to find democracy too chaotic. But the irony here is
that, as “extreme” Realpolitik practitioners implement more and more antidemo-
cratic and threat- or force-based measures in pursuit of order and security, their ef-
forts tend to become more and more counterproductive and self-defeating (see
Burton 1972, chap. 6), generating “security dilemmas” (Herz 1950) and the even-
tual collapse of their own systems.
If such policymakers are alive and well at the end of the day and respond
to the “cognitive dissonance” (Festinger 1962) generated by their failed Realpoli-
tik-based policies and expectations with a “paradigm shift” (Kuhn 1970) to Ide-
alpolitik-based norms and polices, then we might have a situation as we did
following the termination of World War II in Europe, when the erstwhile mor-
tal enemies Germany and France established the basis for what has become the
European Union: one of the most illustrative, successful experiments in peace-

building (still ongoing!) in pursuit of “positive peace.”
But there is irony here as well: even Idealpolitik may contain the seeds of its
own destruction, because “too much stability” may lead to boredom and atrophy.
As Paul Sites (1973) and Kenneth Boulding (1962), among others, have argued,
whatever else we as humans may “need” to develop effectively—physically, emotion-
ally, psychologically, and socially—we also have a need for stimulation, for “drama.”
This is the crux of complexity: the “need” to nudge systems at the “edge of chaos”
so that neither chaos nor order prevails at the zero-sum expense of the other. Quite a
challenge, especially when people are stressed by threats at home and abroad.
One of the major lessons of complexity, therefore, is to never take anything
for granted for too long on either end of the Idealpolitik/voluntary order–
Realpolitik/force-based order continuum. Hence, Viktor Frankel (1985) was able
to survive the horrors and brutalities of a Nazi concentration camp (an incredi-
bly negative setting) by discovering “meaning” in his adversity (a remarkably posi-
tive occurrence). Similarly, former Yugoslavia was “able” to implode genocidally
in the 1990s (an incredibly negative event) after years of intergroup stability and
relative prosperity (a positive setting). Frankel’s feat is an example of the potential
for diversity in human mental models, Yugoslavia’s flip to violent conflict may re-
flect complex system sensitivity to initial conditions and small events.
Given that complex systems can shift “catastrophically” from one end of
the continuum to the other with apparently little effort, one implication here for
the architects of globalization is the need to creatively influence the balance be-
tween order (“McWorld”) and disorder (“Jihad”), so that, for example, those in
the developing world who have traditionally borne the brunt of colonialism and
imperialism have a chance to close the gap between the “haves” and the “have-
nots,” lest frustration/aggression-based cycles of violence degenerate further into
the “new” terrorism (see Sandole 2002)!
One of the interesting aspects of the “new” terrorism—where terrorists are
quite prepared to die in the execution of their acts to inflict catastrophic damage
and destruction on their symbolic and human targets—is that the terrorists, moti-
vated by fundamental ideology more than by a political agenda, are not deterred by
traditional Realpolitik threats or the actual use of force. Consequently, their ulti-
mate intention and effect are to generate maximum unpredictability, instability,
and, therefore, disorder and insecurity in the West and those supported by the West.
Once the “bite-and-counterbite” dynamic of terrorism versus counterter-
rorism reaches some critical threshold, another consequence is each side’s over-
perception of and overreaction to the actions of “the Other” (see Zinnes, North, and
Koch 1961; Holsti, North, and Brody 1968). Hence, the self-fulfilling confirma-
tion of Huntington’s (1993, 1996) otherwise contentious “clash of civilizations”
or of Barber’s (1992) “jihad,” with weapons of mass destruction in the bargain, all
further enhanced by the U.S. rush to war against, and apparently long-term occu-
pation of, another developing country in the Arab/Muslim world, Iraq.
Three-Pillar Framework
While complexity may have generated paralysis over Bosnia, complex-systems con-
cepts have given new meaning to a possible antidote: frameworks that can poten-
tially integrate most if not all disciplines in an effort to explain and to facilitate
dealing with the foci of any one of them. I have developed the “Three-Pillar” com-
prehensive mapping of conflict and conflict resolution (see Sandole 1998a; 1999a,
chap. 6; 2003) as one such framework for identifying and integrating factors associ-
ated with traditionally competing frameworks. I have used this framework as a basis
for developing a “new European peace and security system” (NEPSS) potentially rel-
evant to preventing “future Yugoslavias” (see Sandole 1999a, chap. 7 and below).
Basically, the three-pillar framework comprises pillar 1, conflict—latent con-
flict (pre-MCP); manifest conflict processes (MCPs); or aggressive manifest con-
flict processes (AMCPs)—while pillar 2 deals with conflict causes and conditions,
and pillar 3, conflict (third-party) intervention (see table 3.1).
TABLE 3.1. THREE-PILLAR COMPREHENSIVE MAPPING OF

CONFLICT AND CONFLICT RESOLUTION
Pillar 2 Pillar 1 Pillar 3
Conflict
Conflict Causes (Latent [Pre-MCP]) Conflict
and Conditions MCP/AMCP Intervention
Individual Parties Third-Party Objectives
Societal Issues Conflict Prevention
International Objectives Conflict Management
Global/Ecological Means Conflict Settlement
Conflict-handling Conflict Resolution
Orientations Conflict Transformation
Conflict Environments
Third-Party Approaches
Competitive and/or
Cooperative Processes
Negative and/or Positive
Peace Orientations
Track 1 and/or Multitrack
Actors and Processes
Under pillar 1 (the “Middle Kingdom”), we have the parties, the issues
about which they are in conflict, the long-term objectives they hope to achieve by
waging conflict over certain issues, the means they are employing; their preferred
conflict-handling orientations, and the conflict “spaces” within which their con-
flict is occurring.
Pillar 2 comprises four levels of explanation—individual, societal, interna-
tional, and global/ecological—that capture potential causes and conditions of the
conflict occurring in the “conflict spaces” of pillar 1.
Finally, pillar 3 deals with third-party objectives such as violent conflict
prevention, management, settlement, resolution, and/or transformation; plus
the means for achieving any of these objectives: competitive (confrontational)
and/or cooperative (collaborative) processes; “negative peace” and/or “positive
peace” orientations; and “track 1” (official, governmental) and/or “multitrack”
(nongovernmental, unofficial, and other) actors and processes.
Reflecting the complexity perspective, the three-pillar framework maps the
conflict, its causes, and potential interventions at multiple levels. Complex prob-
lems are characteristics of whole systems. So, they do not have simple solutions,
and it is not possible to anticipate the effects of interventions. The working hy-
pothesis of the three-pillar framework is that to design and implement an effec-
tive intervention into any particular conflict “space” under pillar 1, a potential
third party under pillar 3 will have to “capture the complexity” of the conflict as
represented by all four levels of potential “drivers” under pillar 2.
Complexity is especially relevant here, as it provides the essential concep-
tual basis for combining into a coherent whole traditionally competing frame-
works and ideas. For example, if we were to ask an anthropologist, an economist,
a historian, an international relations specialist, a political scientist, a psycholo-
gist, and a sociologist for their views on why former Yugoslavia imploded into a
genocidal frenzy during the 1990s, we would likely get radically different re-
sponses. Similarly, if we were to ask a businessperson, a citizen activist, a diplo-
mat, a humanitarian aid worker, a journalist, a military officer, and a religious
leader about how to deal with Yugoslav-type conflicts, we would also get differ-
ent responses. All these, however, can be accommodated within the three-pillar
framework. This is precisely what I have attempted to do with “NEPSS.”
The New European Peace and Security System (NEPSS)
I have used the three-pillar framework as a basis for designing the NEPSS: an in-
tervention into post–Cold War Europe that just might be relevant to preventing
“future Yugoslavias” (see Sandole 1998b; 1999a, chap. 7; 1999b) and that, appro-
priately adapted, also may be relevant to conflict interventions outside Europe.
NEPSS comprises descriptive and prescriptive elements—that is, develop-

ments that are actually occurring as well as those that could or should occur. De-
scriptively, NEPSS makes use of existing international organizations in Europe
such as OSCE, the European Union (EU), the Council of Europe (CoE), and
the North Atlantic Treaty Organization (NATO).
NEPSS also employs the basic structure of the OSCE as a conceptual and
operational framework for enhancing the complementarity and synergy of all
mechanisms working together on common problems.
Within this framework, NATO represents an example of political and
military aspects of a reframed, more comprehensive sense of security, the
European Union (EU) an example of economic and environmental aspects, and
the Council of Europe (CoE) an example of humanitarian and human rights as-
pects of comprehensive security. More importantly, each of these heretofore
Cold War institutions has been reaching out to its former enemies, inviting
them to become members or join together in constituting new, post–Cold
War institutions.
For example, at its November 2002 summit in Prague, NATO, which had
already taken in the Czech Republic, Hungary, and Poland as members, issued
invitations to seven other former members of the communist world—Bulgaria,
Estonia, Latvia, Lithuania, Romania, Slovakia, and Slovenia—all of which be-
came members by March 2004. Subsequently, at its December 2002 summit in
Copenhagen, the EU issued invitations to Cyprus, the Czech Republic, Estonia,
Hungary, Latvia, Lithuania, Malta, Poland, Slovakia, and Slovenia, all of which
became members by May 2004. Quite simply, these developments are nothing
short of revolutionary, facilitating a genuine paradigm shift from Realpolitik, zero-
sum national security to Idealpolitik, positive-sum common security.
But revolutionary though these developments are, all these organizations
are basically interstate in nature, while the problems posed by conflicts in former
Yugoslavia and elsewhere are essentially intrastate in nature. Hence, there has
been a need for something else to deal with the conflicts of the post–Cold War
world. This is where the prescriptive element enters the picture.
Prescriptively, NEPSS is characterized by integrated systems of conflict resolution
networks, with vertical and horizontal components. Under the vertical, we would
have a mapping of Europe in terms of the local, societal, subregional, regional, and
global levels of analysis, with track 1–9 actors and processes—governmental/official;
nongovernmental/professional; business; private citizen; research, training, and
education; activist; religious; funding; and media—corresponding to each level (see
Diamond and McDonald 1996). The idea here is that “all conflicts are local.” And,
assuming an early warning system to activate the preventive diplomacy envisaged by
Michael Lund (1996) and others (e.g., Peter Wallensteen [1998] and Walter Kemp
[2001]), conflicts developing at any local level could be responded to by a synergis-

tic combination of track 1–9 resources at that level—plus, to the extent necessary
and possible, societal, subregional, regional, and global levels as well. Should the
vertical dimension fail to prevent “the house from catching on fire,” then there
may be a need for the horizontal dimension to become operational. This would in-
volve the judicious use of Realpolitik force, but basically within an Idealpolitik frame-
work, to achieve negative peace (suppression of the fire) but only as a necessary (not
sufficient) condition for achieving positive peace: the elimination of the (pillar 2) un-
derlying causes and conditions.
While some recent developments in Europe are suggestive of progressive re-
inforcement of NEPSS’s descriptive character and the “vertical” dimension of its pre-
scriptive character—such as the emergence from the November 1999 OSCE
Summit in Istanbul of the Charter for European Security, inclusive of the Platform
for Co-operative Security (see OSCE Istanbul 1999a, 1999b)—other developments
are suggestive of the sole narrow use of Realpolitik force (e.g., the destruction of
Grozny and killings of tens of thousands of Chechen civilians in the Russian Fed-
eration). Even the 1999 NATO air war against Serbia over Kosovo—albeit clearly
for the humanitarian purpose of preventing further genocidal ethnic cleansing of
Kosovar Albanians—falls more into the category of the narrow use of Realpolitik
force basically within a Realpolitik (instead of an Idealpolitik) framework.
These and other developments—the relentless Israeli-Palestinian carnage;
the “new” terrorism; the U.S.–Iraq war; the possibility of a nuclear war between
India and Pakistan over Kashmir, with one or both sides trying to “preempt” the
other; and the escalating development of a “clash of civilizations” or “jihad” be-
tween the Judaic-Christian and Islamic worlds against the background of easily
available weapons of mass destruction—are not only tragic but, via the law of
unintended consequences, potentially very destabilizing.
The primary message of a complexity approach to conflict analysis and res-
olution is that there may be a need for a policeman to pull the attacker off a vic-
tim—for example, for NATO to stop genocide. But this is not the same as
declaring or insinuating, for example, that all Afghans, all Chechens, all Pales-
tinians, all Saudis, all Wahhabis, all Arabs, or all Muslims are “terrorists,” and
then proceeding to eliminate (or be perceived to be eliminating) the entire pop-
ulation and its culture as a way to deal with the “terrorist” problem. Paradoxi-
cally, this problem is being, in part, created self-fulfillingly by this perspective and
corresponding behavior!
Within the terms of the argument posed here, therefore, Realpolitik force
must always take place within a bigger picture, a framework that also allows for
and encourages conflict resolution (dealing with the underlying causes of the fire
at hand) and conflict transformation (dealing with the long-term relationships

among the survivors of the fire), as well as (violent) conflict prevention (preventing
the house from catching on fire in the first place), conflict management (if initial
conflict prevention fails, preventing the spread of the fire), and conflict settlement
(if management fails, forcefully putting out the fire).
If peace is not positive as well as negative—if it does not ultimately deal with
the underlying “conflicts-as-startup conditions”—then “conflict-as-process” will
never be far from the surface, always available to come back to haunt us time and
time again! This is the ultimate message and categorical imperative of a com-
plexity approach to conflict analysis and resolution: not only to think and act
outside the Realpolitik-only box, but to combine it synergistically with other, usu-
ally competing, ways of knowing and acting.
CONCLUSION
It would take extremely enlightened leadership—in the United States, Europe, Rus-
sia, China, and Japan, among others (e.g., Israel, Iraq, North Korea, Palestine, Saudi
Arabia)—to pursue “positive” as well as “negative peace” in coordinated response to as-
saults to the “global commons”: superordinate goals that no one state can achieve on
its own, but only in collaboration with others (see Sherif 1967). But for some inex-
plicable (perhaps, in part, “biological”) reasons, ecological degradation, exponen-
tial population growth, and a growing gap between haves and have-nots, among
other compelling elements of the global problematique (e.g., AIDS), have failed to rise
to the status of William James’s (1989) “moral equivalent of war.”
Perhaps, “if we have time,” we can leave it to the children: the next gener-
ation of decision-makers. In the meantime, however, especially after the Bali
bombings (October 12, 2002), Moscow Chechen hostage crisis (October 23–26,
2002), Madrid train bombings (March 11, 2004), and London transit bombings
(July 7, 2005), global terrorism itself just might provide the motivation for the in-
ternational community to come together, and not just to “root out” terrorism
(Realpolitik), but to deal with its root causes as well (Idealpolitik).
This would be a truly superordinate undertaking that could galvanize the
international community into developing a culture of global problem-solving that
transcends traditional ethnocentrism and a reliance on Realpolitik-only perspec-
tives and measures, paving the way for a new definition of “the enemy” as any
and all assaults to the global commons: a truly complex approach to a set of com-
plex problems at the “edge of chaos.”
Among the conceptual tools that could facilitate movement in this
constructive (albeit ambitious) direction is the Three-Pillar Framework (3PF) or
3PF-generated new European peace and security system (NEPSS) discussed briefly
in this chapter. Analysts working together with policy-makers could use either or
both to capture the complexity of complex conflict situations. In this way, they
could deal with relationships that have gone wrong and the underlying causes and
conditions driving negative developments in those relationships as well as the symp-
toms (indicators) of those negative relationships.
To a very large extent, the US-led invasions and occupations of
Afghanistan and Iraq (especially the latter) seem to be addressing only the symp-
toms of the conflicts that have torn these Muslim countries apart; furthermore,
those interventions may actually be exacerbating the causes of 9/11-type terror-
ism. Such counterproductivity is the price that policy-makers—and the rest of us—
might continue to pay for rejecting or otherwise avoiding conceptual tools that
transcend symptoms and capture the complexity of complex conflicts.7
NOTES
An earlier version of this paper was presented at the 44th Annual Convention of
the International Studies Association (ISA), Panel on “Global Complexity:
Agent-Based Models in Global and International Studies,” Portland, Oregon,
February 25–March 1, 2003. The author gratefully acknowledges comments and
suggestions made by Neil Harrison and Patrick James.
1. Also see Rosenthal and Jacobson’s (1968) classic study “Pygmalion in
the Classroom.”
2. For other discussions in this regard, see Cowley (2003) and Oldham
(2003).
3. “Inclusive fitness thus equals an individual’s Darwinian (egoistic) fit-
ness augmented by an allowance for the effect that the individual can have on the
reproductive success of those who share identical genes by common descent”
(Shaw and Wong 1989, 26–27, emphasis in the original).
4. This section reflects and builds upon parts of chapter 8 (especially
pp. 193–201) of Sandole (1999a).
5. Sensitivity to initial conditions is a characteristic of complex systems
that can also be found in some simple systems. In complex systems, however,
small changes in initial conditions may lead to nonlinear system changes, a flip
in the system to something quite different—as when a forest becomes a desert.
But it does not follow that a system in which small changes in initial conditions
“cause” large system changes is necessarily complex.
6. “There is no place for Turks [the derogatory term the Serbs use for
Bosnian Muslims, whose ancestors adopted the Islamic faith of Turkish in-
vaders] in Republika Srpska,” said General Milenko Zivanovic, the regional com-
mander, who led the final assault on Srebrenica” (AP 1996, emphasis added).
(Also see Honig and Both 1996 and Rohde 1997.)
7. Thus far, I have applied the 3PF to an analysis of the causes of 9/11-
type terrorism (see Sandole 2002) and the 3PF-generated NEPSS to a design for
an EU intervention into post-NATO Bosnia (see Sandole 2004).
REFERENCES
AP. 1996. “Serbs Mark a ‘Victory,’ Victims Are Exhumed.” Associated Press.
International Herald Tribune, 12 July: 10.
Axelrod, Robert. 1984. The Evolution of Cooperation. New York: Basic Books.
———. 1997. The Complexity of Cooperation: Agent-Based Models of Competition and
Collaboration. Princeton, NJ: Princeton University Press.
Bandura, Albert. 1973. Aggression: A Social Learning Analysis. Englewood Cliffs,
NJ: Prentice-Hall.
Barabasi, Albert-Laszlo, and Eric Bonabeau. 2003. “Scale-free Networks.” Scien-
tific American, 288, no. 5 (May): 60.
Barber, Benjamin R. 1992. “Jihad Vs. McWorld.” Atlantic Monthly 269, no. 3
(March): 53–65.
Bateson, Gregory. 1973. Steps to An Ecology of Mind: Collected Essays in Anthro-
pology, Psychiatry, Evolution, and Epistemology. Frogmore, England:
Paladin.
Bertalanffy, Ludwig von. 1973. General System Theory. Harmondsworth, England:
Penguin.
Boulding, Kenneth E. 1962. Conflict and Defense: A General Theory. New York:
Harper and Row.
Burton, John W. 1972. World Society. Cambridge: Cambridge University Press.
———. 1997. Violence Explained. Manchester, England: Manchester University
Press.
Coronel, Mercita. 1996. “Interview with Jane Elliott.” Contrast, 28 March, 1–2.
http://www.magenta.nl/EyetoEye/contraste.html.
Coser, Lewis A. 1956. The Functions of Social Conflict. New York: Free Press; Lon-
don: Collier-Macmillan.
Cowley, Geoffrey. 2003. “Our Bodies, Our Fears (Anxiety and Your Brain: How
Living With Fear Affects the Mind and Body).” Newsweek, 24 February,
42–52.
Diamond, Louise, and John W. McDonald, Jr. 1996. Multi-Track Diplomacy: A
Systems Approach to Peace. 3rd ed. West Hartford, CT: Kumarian Press.
Dobbs, Michael. 1996a. “The Story of Brcko Is the Story of Bosnia.” International
Herald Tribune, 28 June, 2.
———. 1996b. “U.S. Ready to Participate in New Force in Bosnia: Gore Signals
Willingness, NATO Officials Say.” Washington Post, 14 November,
A25.
Festinger, Leon. 1962. A Theory of Cognitive Dissonance. Stanford, CA: Stanford
University Press.
Fisher, Roger, and William Ury. 1983. Getting to Yes: Negotiating Agreement With-
out Giving In. London: Penguin Books.
Fisher, Ronald J. 1993. “The Potential for Peacebuilding: Forging a Bridge from
Peacekeeping to Peacemaking.” Peace & Change 18, no. 3 (July):
247–66.
———. 1997. Interactive Conflict Resolution. Syracuse, NY: Syracuse University
Press.
Fisher, Ronald J., and Loraleigh Keashly. 1991. “The Potential Complementar-
ity of Mediation and Consultation within a Contingency Model of
Third Party Intervention.” Journal of Peace Research 28, no. 1 (February):
29–42.
Frankl, Viktor E. 1985. Man’s Search for Meaning. Revised and updated. New
York: Washington Square Press.
Galtung, Johan. 1969. “Violence, Peace and Peace Research.” Journal of Peace
Research 6, no. 3:167–91.
———. 1996. Peace By Peaceful Means: Peace and Conflict, Development and Civiliza-
tion. London: Sage Publications, for the International Peace Research
Institute, Oslo (PRIO).
Gleick, James. 1987. Chaos: Making a New Science. New York: Penguin Books.
Hedges, Chris. 1996a. “Bosnian Vote in the Fall Wins a Wary Approval.” Inter-
national Herald Tribune, 26 June, 6.
———. 1996b. “West’s About-Face on Karadzic: Familiar and Sad.” International

Herald Tribune, 2 July, 1, 10.
Herz, John H. 1950. “Idealist Internationalism and the Security Dilemma.”
World Politics 5, no. 2 (January): 157–80.
Hobbes, Thomas. 1950. Leviathan. New York: E. P. Dutton; London: J. M. Dent
and Sons. (Orig. publ. in 1651.)
Holsti, O. R., R. C. North, and R. A. Brody. 1968. “Perception and Action in the
1914 Crises.” In Quantitative International Politics: Insights and Evidence,
ed. J. David Singer. New York: Free Press; London: Collier-Macmillan.
Honig, Jan Willem, and Norbert Both. 1996. Srebrenica: Record of a War Crime.
New York: Penguin Books.
Huntington, Samuel P. 1993. “The Clash of Civilizations?” Foreign Affairs 72, no.
3 (Summer): 22–49.
———. 1996. The Clash of Civilizations and the Remaking of World Order. New York:
Simon & Schuster.
James, William. 1989. “The Moral Equivalent of War.” In International War: An
Anthology and Study Guide, ed. Melvin Small and J. David Singer. 2nd
ed. Homewood, IL: Dorsey Press. (Orig. publ. in McClure’s Magazine,
August 1910, 463–68.)
Jervis, Robert. 1997. System Effects:Complexity in Political and Social Life. Princeton,
NJ: Princeton University Press.
Kemp, Walter A., ed. 2001. Quiet Diplomacy in Action: The OSCE High Commis-
sioner on National Minorities. The Hague: Kluwer Law International.
Kral, Brigitta. 2000. “The Eyes of Jane Elliott—Blue Eyes Brown Eyes Exercise.”
Horizon, 1–5. http://www.horizonmag.com/4/jane-elliott.asp.
Kuhn, Thomas S. 1970. The Structure of Scientific Revolutions. 2nd ed. Chicago:
University of Chicago Press.
LeVine, Robert A., and Donald T. Campbell. 1972. Ethnocentrism: Theories of
Conflict, Ethnic Attitudes and Group Behavior. New York: John Wiley.
Lewis, Anthony. 1996. “Karadzic and Co. Have to Be Tried.” International Herald
Tribune, 9 July, 11.
Lund, Michael S. 1996. Preventing Violent Conflicts: A Strategy for Preventive Diplo-
macy. Washington, DC: U.S. Institute of Peace Press.
Mearsheimer, John J. 1990a. “Back to the Future: Instability in Europe After the
Cold War.” International Security 15, no. 1 (Summer): 5–56.
———. 1990b. “Why We Will Soon Miss the Cold War.” Atlantic Monthly,
August, 35–50.
Montville, Joseph V. 1988. Foreword to Vamik Volkan’s The Need to Have Ene-
mies and Allies: From Clinical Practice to International Relationships, by
Vamik Volkan. Northvale, NJ: Jason Aronson.
Morgenthau, Hans J. 1973. Politics Among Nations: The Struggle for Power and
Peace. 5th ed. New York: Alfred Knopf.
Nagel, Ernest. 1961. The Structure of Science: Problems in the Logic of Scientific
Explanation. London: Routledge and Kegan Paul.
Nicholson, Michael. 1989. Formal Theories in International Relations. Cambridge Stud-
ies in International Relations, 3. Cambridge: Cambridge University Press.
Oldham, Pamela. 2003. “A Flight of Mind: The Act of Dissociation Can Protect
Children Emotionally from Trauma, but Repeated Use May Cause
Lasting Harm.” Washington Post, 18 February, F1, F5.
OSCE Istanbul. 1999a. OSCE Istanbul Summit Declaration. 18–19 November ’99.
Istanbul: Organization for Security and Cooperation in Europe.
———. 1999b. OSCE Istanbul Summit. 18–19 November ’99: Charter for European Secu-
rity. Istanbul: Organization for Security and Cooperation in Europe.
Pagels, Heinz R. 1988. The Dreams of Reason: The Computer and the Rise of the Sci-
ences and Complexity. New York: Simon & Schuster.
PBS. 1985. “A Class Divided.” Frontline (PBS documentary produced by W. Peters).
http://www.pbs.org/wgbh/pages/frontline/programs/categories/info/
309.html.
Pfeiffer, John. 1984. “Human Nature: The Universe Inside Your Skull.” Science
Digest 92:92.
Pierre, Robert E. 2002a. “Fear and Anxiety Permeate Arab Enclave Near Detroit:
Muslim Americans Feel They Are Targets in War on Terror.” Washing-
ton Post, 4 August, A3.
———. 2002b. “Victims of Hate, Now Feeling Forgotten: Family of Man Killed
After 9/11 Finds Little Charity, But Much Hardship.” Washington Post,
14 September, A1, A10.
Pomfret, John. 1996a. “Croats and Muslims Edge Toward a Joint Army.” Inter-
national Herald Tribune, 8 July, 5.
———. 1996b. “U.S. Starts Delivery of Heavy Weapons to Bosnia’s Muslim-Croat
Forces.” Washington Post, 22 November, A44.
Rapoport, Anatol. 1960. Fights, Games, and Debates. Ann Arbor: University of
Michigan Press.
Richardson, Lewis F. 1939. “Generalized Foreign Politics: A Study in Group Psy-
chology.” British Journal of Psychology, Monograph Supplements 23, no. 7
(June): 1–91.
———. 1960a. Arms and Insecurity. Chicago: Quadrangle Books; Pittsburgh: Box-
wood Press.
———. 1960b. Statistics of Deadly Quarrels. Chicago: Quadrangle Books; Pitts-
burgh: Boxwood Press.
Rohde, David. 1997. Endgame: The Betrayal and Fall of Srebrenica, Europe’s Worst
Massacre Since World War II. New York: Farrar, Straus and Giroux.
Rosenau, Pauline Marie. 1992. Post-Modernism and the Social Sciences: Insights,
Inroads, and Intrusions. Princeton, NJ: Princeton University Press.
Rosenthal, Robert, and Lenore Jacobson. 1968. Pygmalion in the Classroom. New
York: Holt, Rinehart and Winston.
Sanders, T. Irene. 2002. “To Fight Terror, We Can’t Think Straight.” Washing-
ton Post, 5 May, B2.
Sandole, Dennis J. D. 1984. “The Subjectivity of Theories and Actions in World
Society.” In Conflict in World Society: A New Perspective on International
Relations, ed. Michael Banks. New York: St. Martin’s; Brighton,
England: Wheatsheaf.
———. 1987. “Conflict Management: Elements of Generic Theory and Practice.”
In Conflict Management and Problem Solving: Interpersonal to International
Applications, ed. Dennis J. D. Sandole and Ingrid Sandole-Staroste. Lon-
don: Frances Pinter; New York: New York University Press.
———. 1990. “The Biological Basis of Needs in World Society: The Ultimate
Micro-Macro Nexus.” In Conflict: Human Needs Theory, ed. John W.
Burton. New York: St. Martin’s; London: Macmillan.
———. 1993. “Paradigms, Theories, and Metaphors in Conflict and Conflict Res-
olution: Coherence or Confusion?” In Conflict Resolution Theory and
Practice: Integration and Application, ed. Dennis J. D. Sandole and Hugo
van der Merwe. Manchester, England: Manchester University Press;
New York: St. Martin’s Press.
———. 1998a. “A Comprehensive Mapping of Conflict and Conflict Resolution:
A Three Pillar Approach.” Peace and Conflict Studies 5, no. 2 (Decem-
ber): 1–30.
———. 1998b. “A Peace and Security System for Post-Cold War Europe: Prevent-
ing Future Yugoslavias.” In Encyclopedia of the European Union, ed.
Desmond Dinan. Boulder, CO: Lynne Rienner.
———. 1999a. Capturing the Complexity of Conflict: Dealing with Violent Ethnic Con-
flicts of the Post-Cold War Era. London: Pinter/Cassell (Continuum
International).
———. 1999b. “A Design for Peace and Security in Post-Cold War Europe.” Eth-
nic Conflict Research Digest (University of Ulster) 2, no. 1 (February):
41–42.
———. 2002. “The Causes of Terrorism.” In Terrorism: Concepts, Causes, and Con-
flict Resolution, ed. R. Scott Moore. Fort Belvoir, VA: U.S. Defense
Threat Reduction Agency.
———. 2003. “A Typology.” In Conflict: From Analysis to Action, ed. Sandra Cheldelin,
Daniel Druckman, and Larissa Fast. London: Continuum International.
———. 2004. “Building Peace in Post-NATO Bosnia: A Recommended Action Plan.”
In From Peace Making to Self Sustaining Peace—International Presence in South
East Europe, ed. Predrag Jurekovic, Frederic Labarre, and Ernst M. Felber-
bauer. Vienna: National Defence Academy. www.bmlv.gv.at/wissen-
forschung/publikationen/publikation.php?id=160.
Saperstein, Alvin M. 1995. “War and Chaos.” American Scientist 83, no. 6
(November–December): 548–57.
Shaw, R. Paul, and Yuwa Wong. 1989. Genetic Seeds of Warfare: Evolution,
Nationalism, and Patriotism. Boston: Unwin Hyman.
Sherif, Muzafer. 1967. Group Conflict and Cooperation: Their Social Psychology. Lon-
don: Routledge and Kegan Paul.
Sherif, Muzafer, and Carolyn W. Sherif. 1953. Groups in Harmony and Tension.
New York: Harper and Row.
Simmel, Georg. 1955. Conflict and the Web of Group-Affiliations. Trans. Kurt H.
Wolff and Reinhard Bendix. New York: Free Press; London: Collier-
Macmillan.
Sites, Paul. 1973. Control: The Basis of Social Order. New York: Dunellen.
Stepp, Laura Sessions. 2002. “Infants Now Murdered As Often As Teens: Actual
Rate May Be Higher, Experts Say.” Washington Post, 10 December, A3,
A17.
Stern, Paul C., and Daniel Druckman. 1994/1995. “Has the ‘Earthquake’ of
1989 Toppled International Relations Theory?” Peace Psychology Review
1, no. 2 (Autumn/Winter): 109–22.
Sumner, William Graham. 1906. Folkways. New York: Ginn.

Suplee, Curt. 1992. “The Order of Things.” Review of Complexity: Life at the Edge
of Chaos, by Roger Lewin, and Complexity: The Emerging Science at the
Edge of Order and Chaos, by M. Mitchell Waldrop. Washington Post Book
World 22, no. 51 (20 December): 1, 10.
Tajfel, Henri, ed. 1978. Differentiation between Social Groups: Studies in the Social
Psychology of Intergroup Relations. New York: Academic Press.
———. 1981. Human Groups and Social Categories: Studies in Social Psychology. Lon-
don: Cambridge University Press.
Thucydides. 1951. The Peloponnesian War. The unabridged Crawley translation
with an introduction by John H. Finley, Jr. New York: Modern Library.
USIP. 1997. “Dayton Implementation: The Train and Equip Program.” Special
Report of the Dayton Implementation Working Group, September. Wash-
ington, DC: U.S Institute of Peace.
Vasquez, John A. 1993. The War Puzzle. Cambridge Studies in International
Relations, 27. Cambridge: Cambridge University Press.
Wallensteen, Peter, ed. 1998. Preventing Violent Conflicts: Past Record and Future
Challenges. Report no. 48. Uppsala, Sweden: Department of Peace and
Conflict Research, Uppsala University.
Waltz, Kenneth N. 1964. “The Stability of a Bipolar World.” Daedalus 93 (Sum-
mer): 881–909.
Wilson, Edward O. 1979. On Human Nature. New York: Bantam Books.
Woodcock, Alexander, and Monte Davis. 1978. Catastrophe Theory. New York:
E. P. Dutton.
WP. 2002. “After the War.” Editorial. Washington Post, 12 November, A24.
Zeeman, E. C. 1977. Catastrophe Theory: Selected Papers, 1972–1977. Reading,
MA: Addison-Wesley.
Zinnes, D. A., R. C. North and H. E. Koch, Jr. 1961. “Capability, Threat, and
the Outbreak of War.” In International Politics and Foreign Policy: A
Reader in Research and Theory, ed. James N. Rosenau. New York: Free
Press; London: Collier-Macmillan.
CHAPTER 4
Understanding and Coping with Ethnic

Conflict and Development Issues
in Post-Soviet Eurasia
Walter C. Clemens, Jr.
Generated by scholars from various disciplines, complexity science integrates

concepts from many fields to produce a new slant on evolution.1 Its exponents
seek a general theory able to explain many different types of phenomena—social
as well as biological and physical. If complexity fulfills this goal, it should also
help us to understand ethnic and other problems in post-Soviet Eurasia and
other troubled regions. The contributions of complexity to this understanding
are evaluated in this paper.
This chapter contends that basic complexity concepts do much to explain
the movement toward or away from resolution of ethnic problems in newly in-
dependent states. These concepts do not contradict explanations centered on
the success or failure of movement toward democratization (Snyder 2000), but
rather enrich them and offer linkages to other fields of knowledge. Complexity
starts with a wider lens than democratization but includes it. The concept of so-
cietal fitness, a major concern of complex systems theories, subsumes political,
economic, and cultural strengths. The precise role played by each strength in
shaping societal fitness becomes an important but secondary question.
The analysis here suggests that ideas and concepts from complexity can en-
hance our ability to describe and explain the past and present. But for several
reasons discussed in chapter 10 and elsewhere in this volume, complex systems
theories have much less utility for projecting alternative long-term futures or pre-
scribing international strategy. Still, ideas and concepts from complexity can
enlarge our vision and complement other approaches to social science.
73
74 WALTER C. CLEMENS, JR.
All the hypotheses discussed here are pitched at the macrolevel: they focus
on emergent properties of state and society, or on the international system as an
emergent phenomenon of the interactions of states. As discussed in the intro-
duction to this volume, complexity produces bottom-up theories and models.
However, this chapter does not specifically address the ultimate actor—individu-
als, often the decisive factors in tipping the balance of forces one way or the
other. A full assessment of the past, present, and future of any social system
would have to analyze the key individuals and groups who shape it.
Having registered these caveats, let me summarize the essence of complex-
ity and then apply it to explain divergent policy outcomes in the former Com-
munist states of Eastern Europe and the USSR.
ESSENTIALS OF COMPLEX ADAPTIVE SYSTEMS THEORY
Nonlinearity and complexity are hallmarks of human social networks. Com-

plexity theorists endeavor to explain the process of complex adaptation within
complex systems—whether they be ecosystems, the Internet, or political systems.2
The version of complexity used here—derived from the interpretation of complex
adaptive systems (or CAS) developed by Stuart Kauffman and others at the Santa
Fe Institute—is anchored in eight basic concepts. Three of these—emergence,
agent-based systems, and self-organization—were described in chapter 1. This sec-
tion considers in more depth the ideas related to coevolution, fitness, criticality,
and punctuated equilibrium that are particularly relevant to the discussion in
this chapter.
Coevolution.3 No organism evolves alone. Every individual, species, and so-
ciety coevolves with others and with their shared environment. A change in any
one actor or environment can alter the environment of multiple actors and chal-
lenge their fitness. The more variables shape a system, the harder it is to antici-
pate how change in one element will affect others (the “butterfly effect”).
Fitness. CAS defines fitness as the ability to cope with complexity. To sur-
vive challenges and make the most of opportunity, a fit organism can process in-
formation about and deal with many variables. The theory posits that all life
forms exist on a spectrum ranging from instability (chaos) to ultrastability (or-
dered hierarchy). Fitness is found in the middle ranges of this spectrum between
rigid order and chaos—not in a crystal, where every atom resides in an ordered hi-
erarchy, nor in gases whose molecules move at random. Move too far toward ei-
ther pole, and you lose fitness. Creative and constructive responses to complex
challenges, however, are more likely to be found close to the edge of chaos than
toward the other end of the spectrum.
ETHNIC CONFLICT AND DEVELOPMENT 75
The behavior of social systems emerges from the interactions of their

members. This means that their fitness is a function of the interaction of indi-
viduals within the social and political parameters of the society. Because the fit-
ness of countries is an emergent property, it is not possible to predict with
precision how countries will react to changes in the environment. Up to a point,
countries that are more decentralized are expected to be more adaptable and,
therefore, fitter. The fitness of the United States hovers close to the edge of
chaos, while that of Singapore teeters on the brink of rigidity.
Fitness Landscapes. Coevolution of units within a complex system can be
mapped as a rugged landscape in which the relative fitness of each organism is
shown as a peak rising or falling as a consequence of coevolution. As in an arms
race, the peaks of a predator and its prey may gain or decline according to
changes in their offensive and defensive capabilities. If attackers acquire more
lethal weapons, the fitness peak of the prey will drop. If individuals among the
prey population acquire characteristics that reduce their vulnerability, their
peaks will rise.
Self-organized Criticality. Balanced between order and chaos, a fit being is
like a sandpile that, if one more grain of sand is added, may collapse in an
avalanche. This fragile equilibrium is called self-organized criticality. The sand-
pile metaphor, however, is not universally accepted and is not essential to com-
plexity theory.
Punctuated Equilibrium. The concept of punctuated equilibrium under-
scores that evolution is often marked by surges of speciation and avalanches of
extinction (Gould 2002).4 Species often develop quickly, endure with little
change for a long time, and then die out suddenly—not gradually. Thanks to mu-
tation and self-organization, members of the species find their niche and hang
on to it. When their environment changes, they must adapt or disappear.
How long the system is stable and endures is difficult to predict—especially
in politics. Scientists in many fields noticed in the 1990s that critical events
occur more often—both earlier and later than forecast by the model of a bell-
shaped curve.
DIFFERENCES ACROSS EURASIA:

VARIATIONS THAT NEED EXPLANATION
The huge area to which we shall try to apply CAS is Eastern Europe and the for-
mer USSR. Adapting Snyder’s (2000) analysis, we identify four large domains
that took shape in Eurasia after the breakup of the USSR in 1991—each distin-
guished by the way it dealt with ethnic and development issues. In zone A was a
set of countries that benefited from ethnic calm and enjoyed gradual economic
and political development; in zone B, a shatterbelt of ethnic conflict and mater-
ial regress; in zone C, a region virtually frozen in time—with little ethnic conflict
and stagnant economic life (except in countries where the promise of carbon
fuels brought injections of outside capital). Finally, we may distinguish a hybrid
zone D where major countries—Russia and Ukraine—shared some but not all
characteristics of the other regions.
Zone A consists of societies and states that have experienced almost no
ethnic violence and have made strong progress toward democratic institutions
and economic development through market economics. From the former Yu-
goslavia, the exemplar is Slovenia. From erstwhile Soviet allies in Eastern Eu-
rope, the leaders are the Czech Republic, Poland, Hungary, and a late rising star,
Slovakia.5 Of former Soviet republics, only Estonia, Latvia, and Lithuania belong
in zone A (Clemens 2001).
Zone B comprises societies that became embroiled in severe ethnic fight-
ing in the 1990s—Chechnya, most of former Yugoslavia, and the erstwhile Soviet
republics of Armenia, Azerbaijan, Georgia, and Moldova. Each showed a very
low capacity for coping with ethnic differences and the problems of establishing
a viable economy and a stable democracy. In each case, as Snyder says, partial de-
mocratization probably aggravated ethnic tensions. Thus, “democracy” made it
harder for Armenia’s leaders to negotiate any kind of compromise with Azerbai-
jan over Nagorno-Karabakh, because nationalist firebrands could mobilize votes
against them.6
Zone C refers to Central Asia and Belarus, where dictators suppressed eth-
nic or other challenges to their rule. In the 1990s Tajikistan experienced much
fighting between political rivals, but ethnic differences were not at issue. In the
former Soviet republics of Central Asia, erstwhile Communist leaders became dic-
tators claiming to be both nationalist and democratic. Kyrgystan had a free press
for a time, but this ingredient of a true democracy disappeared in the mid-1990s.7
President Aleksandr Lukashenko tried to russify Belarus and negotiate its
union with the Russian Federation. His opponents sought to establish and
maintain a clear Belarusian identity, but Lukashenko repressed them with lit-
tle overt violence.
Where to place the other states not clearly in one of these three zones? By
the early twenty-first century Slovakia had clearly moved into zone A. There were
signs that Bulgaria, Romania, Croatia, Montenegro, and perhaps even Serbia
might follow suit. But the scales teetered. Each of these countries could readily
drop into zone B or C. Thus, Serbia made major strides toward real democracy
and peace with Montenegro in 2001–2, but could still become embroiled in
more ethnic warfare with Kosovars or the Hungarian-speakers of Vojvodina. By

2005 Bulgaria and Romania were becoming difficult to classify. Neither had suf-
fered much ethnic violence in the previous fifteen years, but each had a low HDI
ranking compared to, say, Slovenia or Slovakia. Each had been admitted to
NATO, on the hope that they would contribute to George W. Bush’s War on
Terrorism, but neither came close to qualifying for membership in the EU. As
this chapter is designed to illustrate the uses of CAS for understanding ethnic
conflict and development in post-Soviet Eurasia, the precise allocation of these
countries to group B or group C is not crucial. Their location on the A-to-D spec-
trum may well change, influenced, for example, by accession to or distance from
the European Union.
The two largest Slavic states emerging from the USSR comprised the hy-
brid zone D. By the early twenty-first century neither Russia nor Ukraine had
achieved a real democracy or a strong market economy. But neither suffered
from outright ethnic violence, except for Russia’s wars against Chechnya
(1994–96 and again after 1999). Russia’s ethnic nationalism was qualified by
civic nationalism. Thus, Moscow recognized Tatarstan’s “sovereignty” within the
Russian Federation (Rossiskaia Federatsiia, where rossiskaia is more inclusive than
the term russkaia, as “British” takes in more diversity than “English”). There were
signs early in the century that the Russian Federation’s Duma and President
Putin might require that any would-be Russian citizen be fluent in Russian.
Ukraine achieved a kind of civic nationalism incorporating native Russian
and Ukrainian speakers. Kyiv avoided war with Russian irredentists in the Crimea
and with Moscow over its claims to ships and naval facilities in Sevastopol. Like
Russia, however, Ukraine failed to use effectively its vast natural resources and
highly educated work force (D’Anieri 1999). Transparency International placed
Russia and Ukraine among the world’s most corrupt countries in the late twenti-
eth and early twenty-first centuries. Ukraine’s “orange revolution” in 2005
promised fundamental changes. President Viktor A. Yushchenko’s administration
followed a Western orientation even as it labored to overcome the misgivings of
diffident Russian-speakers in Ukraine. By year’s end, however, many of the coun-
try’s old problems had reemerged, albeit with new faces.
APPLYING COMPLEX ADAPTIVE SYSTEMS THEORY

TO EXPLAIN PAST AND PRESENT FITNESS
Adopting the language of CAS, this chapter argues that countries such as Slove-
nia, the Czech Republic, and Estonia in the 1990s demonstrated a high level of
TABLE 4.1. THE HIGHEST-RANKING COUNTRIES IN “HUMAN DEVELOPMENT”
AND OTHER VALUES FROM VARIOUS CIVILIZATIONS
HDI GDI Economic
Rank Rank Freedom Honesty
2003, 2003, Freedom Rank 2004, Rank 2003, Cultural
Country n⫽175 n⫽175 Index 2003 n⫽153 n⫽133 Tradition
Norway 1 3 Free 28 (MF) 8 Protestant

Belgium 6 2 Free 19 (MF) 17 Catholic
United States 7 6 Free 6 (F) 18 Protestant
Canada 8 5 Free 18 (MF) 11 Catholic
Japan 9 11 Free 35 (MF) 21 Japanese
Israel 22 22 Free 33 (MF) 21 Israeli
Greece 24 25 Free 56 (MF) 50 Orthodox
Cyprus 25 25 Free 22(F) 27 Orthodox
Singapore 28 28 Partially Free 2(F) 5 Mixed
Slovenia 29 29 Free 62 (MF) 29 Catholic
Korea, Republic of 30 30 Free 52 (MF) 50 Asian
Brunei Darussalam 31 31 Not free n.a. n.a. Muslim
Argentina 34 34 Partly free 68 (MF) 92 Catholic
Estonia 41 38 Free 6 (F) 33 Protestant
Cuba 52 n.a. Not free 155 (RE) 43 Catholic
Belarus 53 48 Not free 151 (RE) 53 Orthodox
Malaysia 58 53 Partly free 72 (MU) 37 Muslim
(continued)
TABLE 4.1. (continued)
HDI GDI Economic
Rank Rank Freedom Honesty
2003, 2003, Freedom Rank 2004, Rank 2003, Cultural
Country n⫽175 n⫽175 Index 2003 n⫽153 n⫽133 Tradition
Thailand 74 61 Free 40 (MF) 70 Buddhist

Cape Verde 103 82 Free 89 (MU) n.a. African
China 104 83 Not free 127 (MU) 43 Chinese
Indonesia 112 91 Partly free 99 (MU) n.a. Muslim-Hindu
India 127 103 Free 119 (MU) 83 Hindu-Muslim
Swaziland 133 107 Not free 72 (MU) n.a. African
Code: For economic freedom, F ⫽ free; MF ⫽ mostly free; MU ⫽ mostly unfree; RE ⫽ repressed.
Sources: “HDI” and “GDI” are from U.N. Development Programme, Human Development Report 2003 (New York: Oxford University Press, 2003), tables 1 and 22. “Free-
dom Index” is from Freedom House at www.freedomhouse.org. “Economic Freedom” is from Heritage Foundation at www.heritage.org. “Honesty Rank” is from Trans-
parency International at www.transparency.org.
fitness. As we see in table 4.1, they scored much higher on the UN Human De-
velopment Index and in Freedom House ratings for political and civil liberty
than did comparable peers such as Serbia, Romania, and Belarus. Each country
in zone A joined both NATO and the European Union. In zones C and D, by
contrast, few countries showed much interest in or had much prospect of join-
ing NATO or the EU in the foreseeable future.
Societies in zone A achieved high levels of fitness on many fronts after the
demise of the Soviet empire. Success in one domain helped them cope with
problems in others. Ethnic peace made it easier to raise living standards, consol-
idate democracy, and nourish creativity. Economic advances in Estonia, for ex-
ample, make it easier for Tallinn to provide welfare benefits for Russian-speakers
residing in Estonia but who were not citizens.
Countries in zones B, C, and D displayed low levels of overall fitness even
though many possessed assets lacking in zone A. Thus, Azerbaijan, Kazakstan,
and Russia possess energy resources far superior to those in any lands in zone A.
Parts of Ukraine and Russia have better soil as well as much richer mineral
deposits than any country in zone A.
Georgia, Armenia, Ukraine, and Russia have evolved from states and cul-
tures dating back more than a thousand years. Slovenia, by contrast, was never
an independent state before 1992. Estonia and Latvia had only two decades of
independence between the two world wars.
Most countries in zones B, C, and D faced simpler ethnic challenges than
in many zone-A countries, because they were more homogeneous. Ethnic mi-
norities were very small in Belarus, Moldova, the South Caucasus, and in most
of Central Asia (except for Kazakstan). About four-fifths of the Russian Federa-
tion’s population was Russian but most other groups in the federation spoke
Russian. A million or so Chechens occupied only a dot on the federation’s pe-
riphery. Still, the governments in zones B, C, and D experienced great difficulty
in dealing with ethnic minorities. By contrast, Estonia and Latvia in the 1990s
faced minorities of Slavic speakers that made up more than one-third of the res-
ident population. Estonian and Latvian leaders espoused a kind of ethnic na-
tionalism tempered by civic moderation. They instituted a naturalization
process that required aspiring citizens to pass residency, language, and civic
tests. By the early twenty-first century—more than a decade since indepen-
dence—few of either country’s Slavic speakers had acquired a working knowl-
edge of the official state language. Children and young adults, of course,
learned Estonian or Latvian more readily than most of their elders. Still, eth-
nic tensions produced no deaths in the Baltic. Estonia even permitted nonciti-
zens to vote in local elections. Indeed, the city councils in Riga as well as Tallinn
were sometimes dominated by coalitions of old leftists and “unity” parties
devoted to the interests of Russian-speakers.
Self-Organization
Self-organization takes in more than democratic politics. It entails also a market

economy and a social system that, from the bottom up, produces innovation and
ways to meet needs and exploit opportunities. The centralized regimes in zones
B, C, and D attempted to direct economic and cultural life as well as politics
from the top down. As in Soviet times, they repressed newspapers and news
media that contradicted the official line. President V. V. Putin was designated
acting president by his predecessor before a snap election that confirmed the ap-
pointment—bolstered by a then popular war against ethnic aliens. Privatization
in Russia and most other countries in zones B, C, and D permitted privileged in-
siders to seize public resources at low cost. So great was the plunder that by 2004
there were more billionaires in Moscow than in New York.
Coevolution
This concept explains several features of post-Soviet Eurasia. Most countries close to
Western Europe have coevolved with the West more quickly and thoroughly than
those that are more distant. Thus, the Czech Republic is more “First World” than
is Kyrgyzstan. But if a country shuts itself off or is otherwise isolated from global
trends, its overall fitness will suffer. Thus, Albania abuts Greece, but its Communist
rulers sought autarky. Belarus abuts Poland and Lithuania, but the government’s
orientation toward Moscow serves to minimize productive exchanges with the
West. Kazakstan “coevolves” with foreign oil drillers, but this is a very limited facet
of coevolution. In many respects Kazakstan and other Central Asia states resemble
Communist Albania—cut off from the West by government fiat.
Emergence
Nowhere in the formerly Communist lands did there emerge strong patterns of
cooperation. Instead, it was more like “every state for itself”—indeed, “every na-
tional and subnational group for itself.” Even in zone A, each state focused on
joining Western Europe and NATO—not on cooperating for shared ends with
its immediate neighbors. Rivalries persisted in the Caucasus even though both
Georgia and Armenia needed the energy that Azerbaijan could provide and for
which it needed buyers.
Central Asian states proved unable, after as well as before independence,
even to find ways to stop the shrinkage of the Aral Sea—an environmental disas-
ter that affects the whole region. The Commonwealth of Independent States had
many accords registered on paper but never executed. Subgroups meant to either
to resist or to strengthen the commonwealth also achieved little.
Whatever the shortfalls of the European Union, it is a triumph of coop-

eration compared to the beggar-thy-neighbor behaviors of ex-Communist soci-
eties. Indeed, it was EU and NATO demands for settled borders and ethnic
peace that persuaded Hungary and Romania to patch up their differences and
convinced Estonia and Latvia to renounce some border regions seized by
Moscow in the 1940s.
Agent-Based Systems
In zone A individual agents are free to innovate and carry on their business with
a minimum of government control. The system is shaped by its members rather
than by a central command. This is not quite “order for free,” which Kauffman’s
version of CAS attributes to established ecosystems (such as coral reefs). Still, it
resembles the positive results that Adam Smith expected if individuals were
allowed to do what they do best, as if guided by an “invisible hand.”
Self-Organized Criticality
CAS warns that societies may be less fit than they appear. Fitness depends on
the harmony of many factors. Just as an extra grain of sand may cause a sandpile
to collapse, a new or heavier burden could seriously weaken an apparently fit so-
ciety. How would Lithuanians respond if a faulty nuclear reactor shut down
their energy supply or spread poison to the air and soil? Or if Russians for a pro-
longed time simply turned off the oil and gas flows on which Lithuania (and
many post-Soviet societies) depends? Each Baltic country endured severe
stresses in the 1990s, but one cannot be sure what grain of sand—what policy in-
novation or social change—may start an avalanche that radically changes a soci-
ety and its fitness.
Punctuated Equilibrium
The concept of punctuated equilibrium warns us not to expect steady progress in

fitness. West European unification did not emerge gradually but in sharp jumps
and with some steps backward. Meaningful social change often requires a period
of preparation. New generations can be educated. In Estonia and Latvia many
young persons who speak Russian at home are learning the official state lan-
guage. Accumulating experiences may tip even middle-aged Russian-speakers to-
ward integration with native Balts. Long plateaus without improvement may
drive some people to depart or to take drastic steps to effect change. But regress
is also possible. How long will displaced persons in Bosnia and other parts of the
former Yugoslavia wait until they return to their homes?
Fitness Landscapes
The relative fitness of a fruit fly and a frog population may be portrayed as
“peaks” that rise and fall with coevolution. Can we graph changing patterns of
fitness among the societies of post-Soviet Eurasia? This is not a simple task, if
only because fitness among humans is multidimensional. The UN Human De-
velopment Index provides a solid starting point to measure public health, edu-
cation, and material living standards.8
If we focus on ethnic problems, we would also study measures of ethnic
harmony and its opposite—injury, dislocations, and deaths caused by ethnic un-
rest. We expect that low fitness in this domain will tend to correlate with low
scores in overall human development, lack of political and civil liberties, low
technological achievement, and corruption. Though it is difficult to show all
these variables in a single peak, a cobweb graphic could illustrate the correlations
suggested here.9
PREDICTING ETHNIC VIOLENCE

AND PRESCRIBING REMEDIES
Theories of complex adaptive systems provide useful concepts for analyzing ethnic
issues and other ingredients of societal fitness. But they offer only general princi-
ples for anticipating future outcomes or prescribing constructive policies. In this re-
gard, however, it does no worse than most competing theories—few of which
provide useful handles for predicting or shaping the future. Indeed, if CAS is cor-
rect about the role of self-organization in fitness, social Darwinists and ultrareal-
ists are wrong: success in politics does not derive from raw power plus cunning.
The fundamental insight of CAS is its prediction that fitness will be found
along the middle of the bell curve ranging from rigid order to random instabil-
ity, though high creativity is most frequently found close to the edge of chaos.
This insight helps explain why Central Asia is frozen in time, why the Caucasus
explodes, and why Russia resorts to an iron fist to overcome chaos, and why
Slovenia and Estonia adapt well to their new freedoms.
This insight has clear policy implications: avoid the extremes of dictator-
ship and anarchy. To generate a healthy and innovative society, cultivate self-
organization—not a system steered from on high. Western policymakers and in-
vestors should not count on authoritarian regimes in Kazakstan or Azerbaijan to
maintain order forever. They should not prop up local dynasties in the hope of
securing privileged access to oil and gas. Outsiders cannot compel internal re-
forms but should do what they can to nudge these societies toward greater self-
organization. Countries such as Azerbaijan suffer not only from top-down
controls but also from a rent-seeking mentality among many well-educated per-
sons who will eventually play major roles in business and politics. Their attitudes
as well as formal structures will determine whether Azerbaijan and Kazakstan use
their petrodollars to create values for the entire community (as in Norway) or fol-
low more closely the Saudi Arabian or Nigerian models.
CAS attention to independent actors agrees with the growing conviction
among political scientists that formal and informal institutions of civil society
help to buffer the ravages of free markets and curb the excesses of willful govern-
ments. The stronger and more diverse the independent agents shaping the for-
merly Communist societies, the healthier and fitter they will be. Constructive
policies will cultivate creative individuals, businesses, and NGOs that enlarge pub-
lic goods and are not dominated by government. These independent agents face
a difficult struggle against the moral legacies of Communism—corruption, group-
think, and a welfare mentality that discourage initiatives from the bottom up.
Even if the goal of self-organization seems clear, questions arise about the
road to this goal. What if democracy terminates democracy—as happened in Ger-
many in the 1930s? Is self-organization desirable if the majority votes against the
minority, as happened in Sri Lanka and as Serbs feared would happen in a ma-
joritarian Bosnia? And what if the majority brings in a government that imposes
the laws and mores of one religion, as in parts of Nigeria?
HOW TO ACQUIRE AND NURTURE FITNESS
Culture matters.10 All the societies in zone A became oriented toward universal
literacy, free thought, and open debate (relative to most other societies) long ago.
The societies in zones B, C, and D moved toward universal literacy only in the
past 100 years.11 Many regimes in these zones still discourage or try to prevent
open debate on policy and other important issues.
Following the leads of John Wycliffe, Jan Hus, Martin Luther, and other re-
formers, each society in zone A acquired its sacred religious texts in the vernacular
between the fifteenth and seventeenth centuries. For the first time in history, some
princes and religious leaders also urged individuals—female as well as male—to read
and interpret sacred texts on their own. This twin revolution helped to liberate all
who experienced it (Clemens 2005). After the Peasants’ Revolt, however, Luther
feared that he was provoking chaos. He then wrote his Short Catechism instructing
people what to believe. But Luther could not stop the transformation he had
Figure 4.1. Date the Bible Published in Vernacular

Correlated with HDI Rank
unleashed. The synergies of literacy and individualist thinking were empowered by

the printing press, the Renaissance, the discovery of the New World, and growing
refinement of scientific methods. Catholic France and Italy had Bibles in the ver-
nacular even before Luther’s challenge to Rome. In the seventeenth century Swe-
den’s monarchy and state church wanted their subjects—even servant girls—to read
and discuss the Bible. Bibles in the vernacular also helped cultivate a sense of na-
tional identity (Hastings 1997; Lepore 2002).
Certainly many factors shape human development, but figure 4.1 shows
a strong correlation between high HDI scores and early publication of Bibles in
the vernacular. Where Orthodox Christianity prevailed, Bibles in the vernacu-
lar were not widely published until the late nineteenth century or the twentieth
century. (The sole exception was Romania, which published both the New and
Old Testaments in the seventeenth century.)12 Wide-scale literacy came to the
Orthodox countries much later than in Protestant and Catholic countries or in
Jewish communities.
Unlike the Christian Bible, the language in which the Qur’an was first
written is regarded by Muslims as sacred—the only truly accurate way to express
God’s message. Islamic societies did not encourage mass literacy or, as a rule,
individual interpretation of sacred texts. For Arabs as well as non-Arab Muslims,
memorization and recitation of the Qur’an have been far more important than
discussion. Few Bosnians, Azeris, or Central Asians have been able to read clas-
sical Arabic.13 Translations of the Qur’an into Persian, Turkish, and Chinese
were for many years largely in the form of paraphrase and commentary.14
By the 1950s Communism had brought near universal literacy to the
USSR and Eastern Europe—even to Albania. But Communist regimes and
schools discouraged freethinking. Centralized controls channeled thought and
discouraged debate. Even when Communist regimes sought to foster technolog-
ical innovation, this proved difficult, because of state secrecy and communica-
tions networks that ran vertically but not horizontally. The Soviet dissident
Andrei Sakharov lost his security clearance and was sent into internal exile;
many other dissidents suffered worse fates.
CONCLUSION
In the early twenty-first century most governments in zones B, C, and D still do

not encourage free thought and debate. Until they do, they will not possess a nec-
essary ingredient of social fitness. Comparatively unfit, they will lag their more
westernized neighbors in many ways. In the language of CAS, these countries—
even erstwhile superpower Russia—will wander in the valleys of a fitness land-
scape, looking for ways to propel their peak(s) upward. Lacking self-organized
economies and polities, they will have great difficulty dealing with ethnic issues
within and across borders. Democratic in form but authoritarian in substance,
they will tend to repress dissent rather than create solutions for mutual gain.
Elections held in 2000 and 2004 suggested that most Russians still hoped
that a vigorous leader, Vladimir Putin, like a legendary vozhd, would unite and
mobilize the people for a better life. Having won many votes by intensifying the
war against Chechens, Putin proceeded to silence independent media, jail the
country’s richest man when he sought to shape political life, and pushed
through reforms permitting the Kremlin to appoint regional governors instead
of having them directly elected by their subjects.
In the mid-1990s Georgians welcomed Eduard Shevardnadze back from
Moscow to Tbilisi, counting on him to end a reign of chaos. But reliance on top-
down leadership did not end turmoil in Georgia. Rather, it added to the already
heavy burdens of corruption at the center. Shevardnadze was ousted in 2003 in
a popular revolt led by an American-trained lawyer, Mikheil Saakashvili, who
promised to replace corruption and chaos with a rule of law. By 2004, however,
he was promoting his own cult of personality. Saakashvili had learned the
rhetoric of democracy, but—in a society that wants a strong, charismatic leader—
he gravitated toward the national norm.
Would closer ties with America improve fitness in former Soviet republics?
For countries such as Georgia and Uzbekistan, closer ties with the U.S. hyper-
power might bring material gains but could also weaken self-organized fitness.
Lacking internal strength, each people’s capacity to cope with ethnic diversity
might then decline—especially if exploited by political entrepreneurs hoping to
gain power and wealth from others’ differences (Singer 1999, 57).
Where Can We Go from Here?
Complexity cannot generate precise algorithms for analyzing ethnic conflict. But
it does provide valuable conceptual tools for this task—principles, metaphors,
models. Thus, a major insight of CAS is the concept of societal fitness. Unlike
neorealists who believe that relative material power—missiles and GDP—is the
best guide to world politics, CAS suggests each actor’s most basic need is a ca-
pacity to cope with challenges at home and abroad, including ethnic diversity.
How could we operationalize these concepts? Let us assume that HDI rank
is an approximate indication of societal fitness, and that societal fitness in a large,
modern society depends on universal literacy and free expression. Let us assume
also that the onset of universal literacy and free thought can be traced to the dates
when the most sacred books of that society were published and when conditions
were established in which they could be subjected to individual interpretation.
The graphics in this essay suggest a correlation between HDI rank and the
date when the Bible was published in the vernacular of the countries that later
became units of the Soviet bloc and Yugoslavia. To really understand these rela-
tionships, however, each variable would have to be studied in depth. Here are a
few of the tasks:
1. Identify the conditions in which the Bible was rendered in the

vernacular and published in each country. When? By whom?
Why? How many copies? Were they repressed (as in Russia in the
1820s)? Did they sell? When did subsequent printings take place?
2. Trace the evolution of literacy in these countries. Develop a

common standard for measuring literacy. How much literacy
was there before the printing press and Martin Luther? How did
it evolve in the decades and centuries after Luther? What forces
and institutions resisted or facilitated the growth of literacy?
3. Trace the growth of independent thinking. What indicators—

in science, the arts, politics, and economics—show indepen-
dent thinking? Such indicators are more evident in a relatively
open metropole such as England than in a repressed depen-
dency such as the places we now call Estonia or Slovakia.
4. Adapt the approach used to study traditionally Christian coun-
tries to those in the Muslim and other religious traditions.
5. Determine a way to prove causation rather than mere correla-
tion. How should we weigh the contribution of one factor,
such as literacy, against others, such as growing wealth?
6. Analyze the chicken-and-egg. Ask which came first: individual
freedom or the twin revolution? Long before Luther, condi-
tions favoring individual freedom were stronger in some re-
gions (such as Bohemia) than in Byzantium or Russia.
7. Distinguish the kinds of ethnic/national consciousness that
existed in previous centuries (for example, among Bohemia’s
Hussites) and that of the last century or two.
8. Distinguish technology from the culture where it is applied.
Why did most Europeans respond with alacrity to the print-
ing press while Islamic cultures did not?
9. Learn from outliers: Romania (Orthodox but closer to Rome
than to Russia) had the Bible relatively early but nonetheless
has a low HDI ranking. Slovaks got the Bible relatively late
but achieved a fairly high HDI score in the late 1990s. Rich
data on individual countries is available in the human devel-
opment reports produced by local social scientists in many
East European and former Soviet states; they are accessible
online from the U.N. Development Programme.
10. Consider the shortcomings of the twin revolution. If literacy
and independent thinking conduced to human development,
why has the West shown so much intolerance, violent nation-
alism, and war? Perhaps the twin revolutions were necessary
but not sufficient for overall fitness.
11. Consider the policy implications: If high levels of human devel-
opment may be traced to a twin revolution that began five hun-
dred years ago, how can they be fostered in societies that have
experienced one or both revolutions only in recent decades?
12. Can the benefits of the twin revolution be nullified by ma-
nipulation of mass media and government power by authori-
ties seeking to create their version of a Brave New World?
The argument here is that to understand present trends and develop construc-
tive ways to deal with ethnic diversity, we must review not just decades but cen-
turies of history—cultural, political, economic. The tasks in such work are
complex and vast, but can be made more manageable by using the conceptual
tools developed by CAS for studying societal fitness.
NOTES
1. The following interpretation of complexity theory is based largely on

the work of Stuart A. Kauffman (1993, 1995, 2000) and other scholars—from the
Nobel physics laureate Murray Gell-Mann to the Nobel economics laureate Ken-
neth Arrow—who have interacted at the Santa Fe Institute. For early work at the
Santa Fe Institute, see Roger Lewin (1992). The Santa Fe Institute publishes the
journal Complexity and working papers such as Martin Shubik, “Game Theory,
Complexity, and Simplicity Part I: A Tutorial” (98-04-027); and Melisa Savage
and Manor Askenazi, “Arborscapes: A Swarm-based Multi-agent Ecological Dis-
turbance Model” (98-06-056). Robert M. Axelrod (1997a, 1997b) and Axelrod
and Michael D. Cohen (1999) have used a variety of methods to resolve complex
problems. For an application of complexity theory by a former student of Axel-
rod, see Lars-Erik Cederman (1997). For a book that blends historical analysis,
international relations theory, and systems analysis, see Hendrik Spruyt (1994).
In the same vein, Robert Jervis (1997) examines the complex interactions of so-
cial units, but says little about self-organization. Compare with James N. Rose-
nau (1990). For related work by IR specialists, see papers given by Michael
Lipson (1996) and Matthew J. Hoffmann (1999). For applications to manage-
ment, see Roger Lewin and Birute Regine (2000). The utility of complexity the-
ory is assessed by Hayward R. Alker and Simon Fraser (1996) and continued in
Alker (1996). For a skeptical view of complexity theory, see John Horgan (1996).
For a more balanced appraisal, see “Edge of Chaos” and many relevant entries in
Ian Marshall and Danah Zohar (1997).
2. Many aspects of nonlinearity are examined in Diana Richards (2000),
where nonlinear models are applied to federalism, alliance formation, epochs
in political economy, environmental regimes, and outbreaks of war. Richards
says that nonlinear modeling can build directly on existing economic theory. In
political science, however, nonlinear modeling must invent a method unique to
the problem at hand—from dynamical systems to spatial voting models to time-
series analysis.
3. On coevolution, see also Charles J. Lumsden and Edward O. Wilson
(1981); Martin A. Nowak et al. (1995); and Edward O. Wilson (1998).
4. But “punctuation” may result from an incomplete fossil record;
also, Gould may have confused “individual,” “class,” and “species.” See Mark
Ridley 2002, 11.
5. See the studies done for Freedom House by Adrian Karatnycky,

Alexander Motyl, and Aili Piano (2001) and the country reports published regu-
larly by Human Rights Watch, Amnesty International, Transparency Interna-
tional, the United Nations Development Programme, and the U.S. Department
of State—all accessible on the Internet.
6. Citing several studies, Snyder concludes that the ethnic content of the
Moldovan conflict was ambiguous. The Moldovan government in the early
1990s was nationalistic, but Russian-speakers in the breakaway “Transdniestr Re-
public” were driven more by nostalgia for the Soviet empire than by nationalism
(Snyder 2000, 250–51).
7. Abutting the former USSR, Moscow’s one-time client state Mongolia
is a special case. In the 1990s Mongolia moved quickly toward democracy, even
though it was poorer than most parts of the USSR and had a weak infrastructure
for education and communication. The country had few internal ethnic prob-
lems (90 percent of the population is Mongolian; 4 percent Kazak; 2 percent
Russian; 2 percent Chinese; 2 percent other) and did not clash with China
despite the potential for expansionist claims by each side.
8. For discrepancies between the UNDP Human Development Report
published annually in New York and country reports published by UNDP
offices in Baltic and East European capitals, see Clemens 2001, 110–11.
9. For a model, see Maruca 2000, 24.
10. For a range of viewpoints, see the essays in Lawrence E. Harrison and
Samuel P. Huntington (2000); Mariano Grondona (1996); and Dominique
Jacquin-Berdal et al. (1998).
11. Literacy rates are difficult to track and measure, but estimates for
many formerly Communist countries are at Snyder 2000, 200–202.
12. Dates of Bible publication in many languages, including those
native peoples of Siberia and North America, are given in www.world
scriptures.org, which also reproduces the opening lines of St. John’s Gospel
in each language. This survey is so detailed that it notes the very different
years for Bible publications in Tartu Estonian (no longer spoken) and Tallinn
Estonian and for Eastern and Western Livonian, each spoken now by only a
hundred or so persons.
13. An Azeri in Moscow showed me his family Qur’an written in Arabic,
for him a completely unknown tongue.
14. Some translations into Malay have been so literal that they were not
intelligible without prior knowledge of Arabic. But Bosnians could have read a
translation into Serbo-Croat in 1875 (Swartz 2004).
REFERENCES
Alker, Hayward R. 1996. Rediscoveries and Reformulations: Humanistic Methodolo-

gies for International Studies. Cambridge: Cambridge University Press.
Alker, Hayward R., and Simon Fraser. 1996. “On Historical Complexity: ‘Natu-
ralistic’ Modeling Approaches from the Santa Fe Institute.” Paper de-
livered at the American Political Science Association Annual Meeting,
San Francisco.
Axelrod, Robert M. 1997a. “Advancing the Art of Simulation in the Social Sci-
ences.” Paper delivered at the International Conference on Computer
Simulation and the Social Sciences, Cortona, Italy.
———. 1997b. The Complexity of Cooperation: Agent-Based Models of Competition and
Cooperation. Princeton, NJ: Princeton University Press.
Axelrod, Robert M., and Michael D. Cohen. 1999. Harnessing Complexity: Orga-
nizational Implications of a Scientific Frontier. New York: Free Press.
Brown, Michael E., ed. 1996. Debating the Democratic Peace: An International
Security Reader. Cambridge, MA: MIT Press.
Bunce, Valerie. 2004. “Status Quo, Reformist, or Secessionist Politics: Explain-
ing Minority Behavior in Multinational States.” Paper prepared for the
Workshop on Nationalism, Secession and Inter-Ethnic Cooperation
and Conflict, Cornell University, April 23–24.
Cederman, Lars-Erik. 1997. Emergent Actors in World Politics: How States and
Nations Develop and Dissolve. Princeton, NJ: Princeton University Press.
Clemens, Walter C., Jr. 2001. The Baltic Transformed: Complexity Theory and Euro-
pean Security. Lanham, MD: Rowman & Littlefield.
———. 2004. Dynamics of International Relations: Conflict and Mutual Gain in an Era
of Global Interdependence. 2nd ed. Lanham, MD: Rowman & Littlefield.
———. 2005 “The Culture of Democracy: Literacy, Individual Freedom, and
Human Development.” Work-in-progress at Boston University and
Harvard University.
D’Anieri, Paul J. 1999. Economic Interdependence in Ukrainian-Russian Relations.
Albany: State University of New York Press.
Doyle, Michael W. 1997. Ways of War and Peace: Realism, Liberalism, and Social-
ism. New York: W. W. Norton.
Elman, Miriam Fendius, ed. 1997. Paths to Peace: Is Democracy the Answer?
Cambridge, MA: MIT Press.
Epstein, Joshua M., and Robert Axtell. 1996. Growing Artificial Societies: Social Sci-
ence from the Bottom Up. Cambridge, MA: MIT Press.
Friedman, Edward, and Barrett L. McCormick, eds. 2000. What If China Doesn’t
Democratize? Implications for War and Peace. Armonk, NY: M. E. Sharpe.
Gould, Stephen Jay. 2002. The Structure of Evolutionary Theory. Cambridge, MA:
Harvard University Press.
Grodona, Mariano. 1996. El mundo en clave. Buenos Aires: Planeta.
Harrison, Lawrence E., and Samuel P. Huntington, eds. 2000. Culture Matters:
How Values Shape Human Progress. New York: Basic Books.
Hastings, Adrian. 1997. The Construction of Nationhood: Ethnicity, Religion, and
Nationalism. New York: Cambridge University Press.
Hoffmann, Matthew J. 1999. “Constructivism and Complexity Science: Theo-
retical Links and Empirical Justification.” Paper presented at the An-
nual Meeting, International Studies Association, Washington, DC.
Horgan, John. 1996. The End of Science: Facing the Limits of Knowledge in the Twi-
light of the Scientific Age. Reading, MA: Addison-Wesley.
Jacquin-Berdal, Dominique, et al., eds. 1998. Culture in World Politics. New York:
St. Martin’s.
Jervis, Robert. 1997. System Effects: Complexity in Political and Social Life. Prince-
Karatnycky, Adrian, Alexander Motyl, and Aili Piano, eds. 2001. Nations in Tran-
sit, 1999–2000: Civil Society, Democracy, and Markets in East Central
Europe and the Newly Independent States. New York: Freedom House.
Kauffman, Stuart A. 1993. The Origins of Order: Self-Organization and Selection in
Evolution. New York: Oxford University Press.
———. 1995. At Home in the Universe: The Search for the Laws of Self-Organization and
Complexity. New York: Oxford University Press.
———. 2000. Investigations. New York: Oxford University Press.
Lepore, Jill. 2002. A Is for American: Letters and Other Characters in the Newly
United States. New York: A. A. Knopf.
Lewin, Roger. 1992. Complexity: Life at the Edge of Chaos. New York: Macmillan.
Lewin, Roger, and Birute Regine. 2000. The Soul at Work: Listen, Respond, Let Go:
Embracing Complexity Science for Business Success. New York: Simon &
Schuster.
Lipson, Michael. 1996. “Nonlinearity, Constructivism, and International Rela-
tions: Or, Changing the Rules by Playing the Game.” Paper presented at
the Annual Meeting, American Political Science Association, Chicago.
Lumsden, Charles J., and Edward O. Wilson. 1981. Genes, Mind, and Culture:
The Coevolutionary Process. Cambridge, MA: Harvard University Press.
Marshall, Ian, and Danah Zohar. 1997. Who’s Afraid of Schrödinger’s Cat: All the Sci-
ence Ideas You Need to Keep Up with the New Thinking. New York: Morrow.
Maruca, Regina Fazio. 2000. “Competitive Fitness.” Harvard Business Review 78
(July–August): 24.
Nowak, Martin A., et al. 1995. “The Arithmetics of Mutual Help.” Scientific
American 272:76–81.
Richards, Diana, ed. 2000. Political Complexity: Nonlinear Models of Politics. Ann
Ridley, Mark. 2002. “The Evolution Revolution.” New York Times Book Review,
March 17, 11.
Rosenau, James N. 1990. Turbulence in World Politics: A Theory of Change and Con-
tinuity. Princeton, NJ: Princeton University Press.
Rousseau, David L., Christopher Gelpi, and Dan Reiter. 1996. “Assessing the
Dyadic Nature of the Democratic Peace, 1918–1988.” American Political
Science Review 90:512–33.
Russett, Bruce, John Oneal, and David R. Davis. 1998. “The Third Leg of the
Kantian Tripod for Peace: International Organizations and Militarized
Disputes, 1950–1985.” International Organization 52:441–67.
Singer, J. David. 1999. “Prediction, Explanation, and the Soviet Exit from the
Cold War.” International Journal of Peace Studies 4, no. 2:47–59.
Snyder, Jack. 2000. From Voting to Violence: Democratization and Nationalist Con-
flict. New York: W. W. Norton.
Spruyt, Hendrik Spruyt. 1994. The Sovereign State and Its Competitors: An Analysis
of Systems Change. Princeton, NJ: Princeton University Press.
Swartz, Merlin. 2004. “Translations of the Quran.” Work-in-progress at Boston
University.
United Nations Development Programme. 2003. Human Development Report.
New York: Oxford University Press.
Waert, Spencer R. 1998. Never at War: Why Democracies Will Not Fight One
Another. New Haven, CT: Yale University Press.
Wilson, Edward O. 1998. Consilience: The Unity of Knowledge. New York:
A. A. Knopf. http://www.worldscriptures.org.
CHAPTER 5
Beyond Regime Theory

Complex Adaptation and the
Ozone Depletion Regime
Matthew J. Hoffmann
This chapter undertakes a deceivingly simple endeavor—to show how insights

from complexity concepts (drawn from the complex systems taxonomy described
in chapters 1 and 2) can aid our understanding of environmental regimes. This
is not a wholesale indictment of regime theory. As there is no single regime the-
ory, it is not possible to indict the whole enterprise (Hasenclever, Meyer, and Ritt-
berger, 1997). Rather, I focus on one aspect of a complexity approach that is
missed by traditional (especially neoliberal) regime theory—namely, how actors co-
evolve with their political context and how adaptive actors come to understand
both the environmental problems that they face and the potential solutions to
those problems. I illustrate how complexity concepts can help us to move beyond
regime theory with a brief examination of the ozone depletion regime.
In its traditional, neoliberal instantiation, regime theory is mostly con-
cerned with bargaining. How can states (for the most part) come up with a set of
rules/institutions to help them achieve/exploit common interests in the absence
of a hegemonic authority? This is the driving question for much of the regime
theory enterprise, and a good deal of time and expertise has been spent re-
searching the ways in which such bargaining has and can take place across a
number of issues (Keohane 1984; Krasner 1983; Hasenclever, Meyer, and Ritt-
berger, 1997; Koremenos, Lipson, and Snidal 2001). The goal of regime theory
is simple: to explain when rules, procedures, principles and norms (Krasner
1983) are likely to occur in a particular issue-area.
95
96 MATTHEW J. HOFFMANN
In the best social scientific tradition, you have a dependent variable—

regime emergence—and a number of independent variables: problem type,
situation type, extant institutions, norms, exogenous shocks, and so forth.
Hypotheses follow on when to expect regimes to form and function effectively
(List and Rittberger 1992). Rational choice is often the crucial behavioral
assumption for explaining how the independent variables come to determine
whether or not a regime is likely.1 Beyond rational choice, Oran Young (1994,
1997, 1999) provides a regime life cycle, so to speak, in his institutional bar-
gaining approach that parses regime dynamics into prenegotiation, negotia-
tion, and postnegotiation phases. Others focus on social practice models
(Young 2002).
What’s wrong with this? Nothing per se. However, taken as a diverse
whole, regime theory has faced a number of challenges that have yet to be re-
solved. First, there is the challenge of complexity. The list of independent vari-
ables is long, and how they fit together is not self-evident. Second, there is the
challenge of mechanisms. Most work in regime theory has been dedicated to dis-
cerning factors and ascertaining if actual bargaining matches our hypotheses—
especially those drawn from rational choice. Less time, with the laudable
exception of Oran Young (1997, 1999), has been spent on exploring the mecha-
nisms that produce the rules, norms, principles, and decision-making procedures
that characterize regimes.
A complexity approach has a great deal to add to the study of regimes and
to move us beyond (neoliberal) regime theory. Models of evolutionary coopera-
tion and bounded rationality from the complexity literature can inform sophis-
ticated regime bargaining studies (Axelrod 1997; Arthur 1994b). Studies of
increasing returns in politics and economics (Arthur 1994a; Pierson 2000) allow
us to understand why regimes, once formed, stick and have influence (Ikenberry
2001). In this chapter, I explore an additional aspect of complexity that moves us
beyond regime theory—complex adaptation. I discuss how treating states as adap-
tive actors ensconced in a coevolutionary process with their context helps us to
understand the ozone depletion regime.
A complexity approach considers that regime theory as traditionally
conceived freezes too much of the dynamism in environmental regimes.
Regime theory generally conceives of the actors and their context to be rela-
tively static and independent. In other words, states know what they want
and can relatively easily perceive the problems they face and the rules of the
game. A complexity approach calls this into question and claims that the sys-
tem rules and the actors are entwined in a coevolutionary relationship. In
other words, actors are constantly adapting to the system rules, and system
BEYOND REGIME THEORY 97
rules (both the specific regime rules and the larger constitutive rules of the sys-
tem) are dynamically changing through the self-organized actions and inter-
actions of actors.
A complexity approach provides insight into how this coevolution takes
place, thus providing insight into the formation of regimes and their evolution
through time. Through complexity, it is possible to understand actors’ percep-
tions of the problems and the solutions deemed possible—crucial prior informa-
tion for explaining the bargaining that is the usual fodder for regime analysis.
Young (1999) notes that such regime foundations or “discourses” “not only pro-
vide a way of framing and addressing problems and the behavioral complexes
within which they are embedded but also contain normative perspectives on the
importance of the problems and appropriate ways to resolve them” (206–7).
To demonstrate such an approach, I turn to the ozone depletion regime.
This oft-studied regime (see, e.g., Litfin 1994; Haas 1992; Benedick 1991; Tolba
1998; Parson 1993; Rowlands 1995) has been hailed as a singular success in the
realm of environmental politics. Yet there are puzzling aspects of the regime
that have been overlooked with serious consequences for how we understand
the response to ozone depletion itself as well as the “lessons learned” for regime
analysis in environmental politics more generally. Specifically, I explore two as-
pects of this complicated regime. First, how did the system rules governing par-
ticipation in the regime change from calling for North-only negotiations in the
early 1980s to the universal negotiations of the late 1980s? The international
community’s understanding of who should participate in the ozone depletion
regime changed over time and influenced what issues were addressed, and what
rules would be encompassed in the regime itself. It is impossible to understand
the ozone depletion regime without understanding the underlying requirement
for universal participation that developed. Second, why did the United States,
which could have ignored this change, adapt its understanding of the system
rules to fit the change?.
I contend that one way to understand both of these puzzles is to treat
states as complex adaptive actors and to consider that universal participation
emerged through the process of complex adaptation. I demonstrate the plausi-
bility of this contention by tracing the coevolution of the dominant state in the
ozone depletion regime (the United States) and the system rules within which
the United States was embedded. First, I discuss the aspects of complexity used
to analyze the ozone depletion regime. I then give a brief overview of the ozone
depletion regime for background purposes. The discussion of the two main puz-
zles is next, and I conclude with some thoughts about further empirical testing
of the hypotheses presented.
COEVOLUTION AND COMPLEX ADAPTATION
Complexity provides a process that links agents and a broader (social) system. In
complexity, the two are inextricably entwined, though, crucially, neither is re-
ducible to the characteristics of the other. System characteristics cannot be di-
rectly derived from knowledge about agent characteristics. Similarly, agent
characteristics are not deducible from knowledge of the system. This irreducibil-
ity is characteristic of complex systems whereby the system characteristics emerge
from the actions and interactions of the agents and those actions and interactions
are shaped by the system (see, e.g., Holland 1995; 1998; and Arthur, Durlauf, and
Lane 1997). Explaining the system-agent linkage is done analytically by positing
the process of complex adaptation. This process links adaptive actors in a coevo-
lutionary relationship with each other and with the larger system.
Adaptive agents are defined by internal rule models or schema (Holland
1995; Gell-Mann 1994). These rule models represent the agent’s internal (or sub-
jective) understanding of the world (the larger system) around them. They allow
the agents to perceive and define their situation, predict the consequences of ac-
tion, and act. In most applications of adaptive agents, the rules are behavioral,
but they can also represent identities, interests, and goals.
The actions that adaptive agents undertake and the interactions in which
they participate reproduce or alter the larger system. The system rules that define
the agents’ context emerge from the actions and interactions of the agents,2
while in turn shaping those same actions and interactions. In a complex system,
the system rules influence agents’ internal rule models through coevolutionary
processes. When some agents change their behavior, this alters the system for the
other agents. A new context “forces” agents to alter their rule models as the con-
text determines what goals, interests, and behaviors are appropriate or fit. Adap-
tive agents are always trying to “fit” with their context. When their internal rule
models fit their context, the agents are successful. When their rules do not fit,
the agents are not successful. System change results when innovation on the part
of a subset of agents throws the system rules into flux and other agents then
adapt their rule models and therefore their actions and interactions.
At the agent level this adaptation is facilitated by self-evaluation of behav-
ior. Agents evaluate the results of actions and assess the ‘fitness’ of their rule
models.3 Internal rule models are strengthened, weakened, changed, or kept in
relation to the evaluations. The system rules, produced by agent actions and
interactions, do more than constrain potential actions; they become incorpo-
rated, through the evaluation process, into the agents’ rule models.4 In this way,
an agent’s system shapes its internal rule models—its interests, identity, and
behavior—while the agent’s actions feedback and affect those same system rules.
A complexity perspective thus provides a model of agent behavior—

adaptive—and a feedback process that dynamically links the internal under-
standings that agents have of their system rules with the system rules
themselves. Very simply, the logic of complexity entails the coevolution of
actor understandings and the political context. States are embedded in a sys-
tem with rules, and each state has an internal understanding of the system
rules. At any time states “know” what the global response to environmental
problems like ozone depletion should be, though not all states will share the
same understanding. States act on the basis of these internal rule models.
Their actions (bargaining, for instance) shape the system rules, reifying or
changing the rules that dominate in an issue-area. There is, thus, inherent dy-
namism in the internal rule model/system rules relationship—adaptive actors
shape and are shaped by their political context through their actions and in-
teractions. Change is driven by interpretation of scientific information, do-
mestic politics, and social interaction with other states—the evaluation of
actions and the altering of rule models.5
The ozone depletion regime provides fertile ground for a demonstration
of the empirical utility of a complexity perspective. After a brief background de-
scription of the ozone depletion regime, I trace the process of complex adapta-
tion and discuss how universal participation emerged, what impact it had on
the ozone depletion regime, and why the United States adapted to this emer-
gent participation rule.
THE FORMATION OF THE OZONE DEPLETION REGIME
Ozone depletion was recognized as a potential problem in 1974 when two scien-
tists put forward the hypothesis that chlorofluorocarbons (CFCs) could destroy
the ozone molecules in the stratosphere that protect the Earth from UV radiation
(Rowlands 1995, chap. 2).6 Politically, this “global” problem entailed varied re-
sponses. Some states, notably the United States and the Nordic states, took uni-
lateral domestic action and began to regulate CFCs. International activity began in
the late 1970s, and negotiations resulted in a succession of agreements (Vienna
Convention, 1985; Montreal Protocol, 1987; London Amendments, 1990;
Copenhagen Amendments, 1992) that moved the state of the ozone depletion
regime from a call for research through a complete phaseout of CFCs.
The initial regime activity and negotiations that produced the Vienna
Convention of 1985 consisted mainly of bargaining between the United States
and the EU states. At first the EU was opposed to reductions in ozone-depleting
chemicals, because collectively the European states had come to dominate the
worldwide CFC market when US domestic actions served to decrease its output.
With the EU and US at loggerheads, the Vienna Convention accomplished
little, except for an important call for further negotiations and a statement of
guiding principles. The discovery of the ozone “hole” in 1986 dramatically
heightened the sense of urgency surrounding the ozone depletion problem, and
though the international community lacked scientific proof of the connection
between CFCs and ozone depletion, they forged ahead with negotiations to re-
duce CFC emissions.
These negotiations culminated in the Montreal Protocol and were again
mainly comprised of US/EU bargaining. The negotiations were larger in num-
ber (up to about sixty states from twenty-five) and the Global South took part in
significant numbers for the first time. The Montreal Protocol itself laid out a
compromise on binding reductions that would see CFC emissions decrease by
50 percent by the year 2000. The Montreal Protocol is still hailed as an exemplar
of environmental negotiating and the ability of the international community to
take decisive action on an urgent environmental problem. The signing of the
Montreal Protocol is generally conceived of as the emergence of the ozone de-
pletion regime—when the rules fostering cooperation to solve the ozone deple-
tion problem were put into place.
Regime dynamics did not end with the Montreal Protocol, however. After
the protocol was signed, new scientific findings solidified the proof of the CFC-
ozone depletion connection, and the urgency surrounding the issue ramped up
once again. The move to amend the Montreal Protocol and strengthen the
ozone depletion regime was complicated by the now universal attendance at the
ozone negotiations. Post-Montreal negotiating sessions routinely included up-
wards of one hundred states, and the North-South dimension of these negotia-
tions altered the regime dynamics significantly.
At the London meeting of the Parties to the Montreal Protocol bargaining
was much more North-South than it was United States–European Union. The
compromise or “Grand Bargain” (Tolba 1998) struck in London was an acceler-
ated CFC phaseout (100 percent emission reductions by 2000) combined with a
pledge by Northern states to compensate Southern states for non-CFC develop-
ment paths and their accession to the ozone depletion regime. Thus, in five short
years, the international community went from calls for research coming from a
negotiation with mostly Northern states to a full-fledged phaseout of CFCs and
a regime encompassing most of the globe.
Two parts of this regime formation story demand our attention. First, in
1987 we see a breakpoint change in how the international community perceived
the proper response to ozone depletion. Before this time, the proper response
consisted for the most part of Northern states alone. The system rule defined
ozone depletion as a Northern problem and deemed that solutions would be

formed and adhered to by Northern states. After 1987, universal participation in
the ozone depletion regime has been the rule. The definition of the ozone de-
pletion problem and solutions had changed. Why did the ozone depletion
regime go global?7
Second, in 1987, the US internal rule model clearly defined ozone deple-
tion as a North-only problem. This rule model drove US actions in the Vienna
Convention and Montreal Protocol negotiations. After Montreal, the US rule
model changed. This would not be so puzzling if the United States were not the
hegemon, and the single most important player in the ozone depletion regime.
According to conventional wisdom (and conventional international relations
theory), the United States, of all states, did not have to change its perceptions of
ozone depletion. Why did the United States change its internal rule model to
fit the altered system rules?
These puzzles are more than academic nitpicking. Understanding the
emergence, functioning, and lasting impact of the ozone depletion regime re-
quires that we understand how ozone depletion came to require universal par-
ticipation and, further, how even the United States came to adapt to a particular
set of system rules. The altered system rules determined how the bargaining to
amend the Montreal Protocol would take place. Further, the altered system rules
in ozone depletion would come to have an impact beyond the ozone depletion
regime, influencing how other environmental problems, like global warming,
came to be defined.
EXAMINING PARTICIPATION AND REGIME FORMATION 8
A complexity perspective tells a simple, though profound, story about the evolu-
tion of the participation system rule. The story is simple, because the transition
can be described in three straightforward stages. In stage 1, the system rule and in-
ternal rule models of the states match—states’ behaviors are driven by internal
models that are adapted to and reify the system rule for North-only participation.
Both Northern and Southern states define ozone depletion as a problem requir-
ing North-only participation. In stage 2, some actors change their rule models
when they negatively evaluate the outcomes of their actions (or inaction, in this
case). Southern states come to define the ozone depletion problem differently and
perceive that it requires their participation as well. When they act on this new in-
ternal model and participate in the regime negotiations, the stability of the North-
only rule erodes. The participation system rule goes into flux. In stage 3, other
actors adapt to the flux in the system rule. Northern states change their rule mod-
els and begin advocating universal participation. The actions of Northern and
Southern states, now operating with similar rule models again, instantiate a newly
stable system rule for universal participation.
This is quintessential complex adaptation. Agents are driven by rule models
and their actions (re-)create the system in which they are embedded. Innovation in
a subset of the population of agents disrupts the system rules and leads to adapta-
tion across the whole system. The microprocess of agent-adaptation and internal
rule model change is linked to the macroprocess of evolution in system rules.
The story is profound because of the implications of the transitions. The
system rules can be considered the political context (or “episteme”—see Johnston
2005) within which regimes are formed; they are the rules of the game and struc-
ture the process of regime formation. Understanding the system rules is a key
aspect of analyzing regimes, because they determine what gets bargained over
and how such bargaining takes place. In traditional regime theory, the system
rules are assumed to be static and unproblematic. This is not always a poor set of
assumptions, but it reifies a misleading understanding of politics as relatively sta-
ble. It blinds us to the dynamic (even if slowly changing) nature of system rules
and to how system rules shape and constrain the bargaining that is the main
focus of regime theory.
With a complexity perspective we can account for change in system rules
and more fully explain the process of regime formation. In the following discus-
sion, I trace in greater detail the transitions in the participation system rule and
in US rule models for dealing with ozone depletion through the three stages out-
lined above.9
Stage 1: North-Only Participation
In 1985 there was a stable system rule calling for North-only participation. Both
Northern and Southern states had internal rule models that defined the ozone
depletion problem as one requiring a Northern negotiated regime, and the ac-
tions of both Northern and Southern states reified the North-only system rule.
It was obvious to both Northern and Southern states that ozone depletion
was a Northern problem (Sims 1996, 201–14). Only four of the original twenty-
one signers of the Vienna Convention were Southern states, and by 1987, only
two developing nations had ratified the convention (Benedick 1991, 265–69).
Paul Horwitz of the US Environmental Protection Agency (EPA) and UN Envi-
ronment Program (UNEP) noted that “decisions were being made by countries
that were the problem—they believed they could get a hold of the problem. [The]
group thought they owned the issue.” He also noted that in the beginning there
was “less of a stress on global participation” and that it “didn’t make sense to ne-
gotiate a global agreement.”10 A global agreement did not make sense to South-
ern states, either; as Stephen Seidel of the US EPA observed, “[T]he developing
countries didn’t want a big role. There was the Toronto group [the United
States, Canada, and the Nordic countries] and there was the EU[;] the key nego-
tiations took place between them.”11 Southern disinterest in participating
matched Northern indifference to Southern participation and would continue
past the Vienna negotiations. India’s position was broadly representative:
In the post-Vienna period, Indian policy makers continued to believe that

ozone depletion was mainly the concern of the developed countries. They
saw little change in the situation—the scientific uncertainties about ozone
depletion continued, there was no proof of any threat to India, and India’s
CFC production remained marginal to world production. India, there-
fore, did not participate in the preparatory meetings for the Montreal
Conference. (Rajan 1997, 59)
This stability around North-only participation pervaded the negotiations that

culminated in the Vienna Convention of 1985. The main bargaining was under-
taken between the United States and the European Union, and the main issues in
contention were whether to undertake CFC emission reductions before unassail-
able scientific proof of the CFC–ozone depletion link was available and how to deal
with technology transfer among Northern states once substitutes for CFCs were de-
veloped. The bargaining at this stage was shaped by the participation rule. Because
everyone knew that ozone depletion was a Northern problem, everyone also knew
what issues to discuss. Thus, it is no surprise that the issues that traditionally engage
the South—economic development, financial resources, and technology transfer—
were not of primary concern. The Vienna Convention only discussed the need to
take into account the “circumstances and particular requirements of developing
countries” and only vaguely discussed technological cooperation.12
The North-only rule continued to shape the formation of the ozone de-
pletion regime as the international community responded to the discovery of the
ozone hole and the associated increased urgency of ozone depletion with negoti-
ations toward an ozone protocol. In the four meetings in 1986–87 the main bar-
gaining continued to occur between the United States and the European Union,
with the United States pushing for deep CFC cutbacks quickly and the Euro-
pean Union urging slower action (see UNEP 1986, 1987a, 1987b). Southern
participation through the summer of 1987 and discussion of Southern issues
were sparse. Essentially, the provisions for development assistance remained
nearly unchanged from what was included in the Vienna Convention.
In Montreal in September 1987, the United States remained focused on
North-only negotiations and set its sights on Japan and the Soviet Union in
addition to the European Union (Boston Globe, September 8, 1987, 1). The
United States still considered the ozone depletion problem to be a Northern
issue. While the United States was “looking for an effective agreement involv-
ing as many nations as possible” (San Diego Union-Tribune, September 10, 1987,
A17), it is clear that the nations that the United States wanted to involve were
Northern states.
Throughout the early regime formation process, Northern and Southern
actions reified the North-only participation rule. The United States and the rest
of the Northern states were not trying to convince Southern states to participate.
Similarly, Southern states were not pining to be included. Southern states were
not barred from the ozone depletion regime. On the contrary, the negotiations
were held under the auspices of the UNEP, a universal membership organiza-
tion. It was simply the case that both Northern and Southern states had inter-
nal rule models that defined the ozone depletion problem as one that required
North-only participation. The system rule determined what states were present
and what was bargained over—important parameters of regime formation.
Stage 2: Instability in the System Rule
But the stability of the North-only participation rule was not destined to
endure. In the summer and fall of 1987, Southern rule models underwent a
transition, evident in changing Southern behavior leading up to and including
the Montreal negotiations in September 1987. Very simply, Southern states
started to participate. Whereas previous negotiating sessions had been attended
by twenty-to-thirty, states with less than a third from the South, at Montreal 65
percent of the fifty-seven participants were Southern states (UNEP 1987c).
In-depth explanation of the transition in Southern states’ rule models is
beyond the scope of this chapter, but briefly, a number of reasons are evident
for why Southern states would weaken the North-only rule model and come to
feel that their own participation in the ozone depletion regime was necessary.13
First, UNEP and its executive director, Mostafa Tolba, aggressively advocated
universal participation (Hoffmann 2002). Keeping Southern states informed
of the process contributed to the negative evaluation that Southern states had
of the ozone depletion regime and their own North-only rule models. In addi-
tion, there were incentives to change rule models based on traditional notions
of self-interest (the desire to have continued access to cheap, useful chemicals)
and environmentalism (the desire to be seen as environmentally friendly).
However, for this discussion, why the Southern states changed their rule mod-
els is in some ways less important than the fact of their changed rule models
and behavior.
In participating, Southern states altered the stability of the system rule call-
ing for North-only participation. With Northern states still following the North-
only rule and Southern states following a new internal model that called for their
own participation, what emerged was a participation rule in flux.
The instability in the system rule paradoxically had both relatively minor
as well as extremely significant consequences. The flux had little impact on the
Montreal negotiations themselves. There was a great deal of inertia behind the
proceedings to this point, and Southern participation at Montreal did not
greatly alter the work that had been accomplished over the previous two years. In
addition, as the Northern states were still operating on a North-only participa-
tion rule, the substance of the discussions at Montreal remained focused for the
most part on Northern issues of CFC reductions and substitutes. While South-
ern issues were raised, the South started ‘playing’ too late to be a major factor at
Montreal, as most of the provisions of the MP had already been worked out
(Miller 1995, 78–79).
On the other hand, the transition in Southern rule models was significant
in that it changed the post-Montreal negotiating landscape entirely and drove
the adaptation of American and other Northern states’ rule models. The change
in Southern rule models and behavior altered the system rules and presented a
new context to which the Northern states had to adapt. As Northern states
would soon discover, a regime negotiated by states driven by North-only partici-
pation rule models would not be effective in a context where the North-only sys-
tem rule had eroded. The flux in the participation system rule would soon come
to alter how states perceived the ozone depletion problem and the solutions nec-
essary to solve it.
Stage 3: US Adaptation and the Emergence of

Universal Participation
Southern states were not satisfied with a protocol that was negotiated es-
sentially without them, and they signaled this dissatisfaction by declining to sign
the Montreal Protocol. This action, while further eroding the North-only system
rule, blindsided the United States and other Northern states. Most Northern ne-
gotiators thought “the terms [of the Montreal Protocol] were attractive enough
to encourage other developing countries to sign onto the document” (Rowlands
1995, 169–70). This thinking was a vestige of the understanding of the ozone
problem as one to be dealt with through North-only measures—a rapidly deteri-
orating vision of a global response for ozone depletion.
Yet Northern adaptation to the flux in the participation system rule was
not inevitable. If Northern states deemed the Montreal Protocol sufficient to
solve the ozone depletion problem, then Southern participation may never have
been an issue of concern. However, immediately after the Montreal Protocol was
signed, the urgency surrounding the ozone depletion problem dramatically in-
creased. With scientific proof of the CFC–ozone depletion link in hand and
more information on the potentially catastrophic effects of ozone depletion, the
international community came to realize that the Montreal Protocol was not
enough to solve the problem (see, e.g., San Francisco Chronicle, October 28, 1987,
A24; Benedick 1991; Washington Post, September 27, 1988, A3; Congressional
Quarterly Weekly Report, March 19, 1988, 706). The regime formation and/or
strengthening process thus continued in the late 1980s, but it continued with a
different focus.
The United States negotiated the Montreal Protocol operating under a
rule model (and political context) that told it that Northern states had control of
the problem. This was not an outlandish assumption. The South produced and
consumed relatively small amounts of CFCs, and Southern production to this
point was primarily in joint ventures with Northern companies.14 In 1988–89,
however, the ignored or forgotten potential of the South to produce/consume
CFCs and thus contribute to the problem became critical to Northern states
with a changing perception of the ozone depletion problem. CFC technology is
relatively simple and was widely available. In addition, certain Southern states
(Brazil, China, India, and perhaps Indonesia) had large enough domestic mar-
kets to create a viable CFC industry. Crucially, this potential to contribute to the
problem was as evident in 1986 as it was in 1988. However, the importance
attached to the Southern potential and the expectation that the South would
comply with the Montreal Protocol changed significantly in 1988–89. In 1988,
Southern states no longer assumed that ozone was a North-only problem, with
the North responsible for devising solutions. The South claimed a voice and
demanded to be a part of the decision-making process that promised to alter
development paths.
Initial US evaluations of the protocol were enthusiastic, and the imminence
of a rule model change was not evident in the United States in the months im-
mediately following its signing. In late 1987, the United States was very pleased
with the substantive results of the protocol, especially the 50 percent reduction in
some CFCs. However, as the science became more certain and the reality of the
lack of Southern signers became more evident, the enthusiasm for the protocol
waned. Even as the United States prepared to ratify the protocol, the US Senate
prepared to consider further, unilateral action, and the EPA began calling for a re-
assessment and accelerated cutbacks (Mills 1988, 370). Environmental Non-
governmental Organizations (NGO) representatives were even blunter. David
Doniger of the Natural Resources Defense Council (NRDC) argued that “There
is virtually no chance that the current protocol will be sufficient to solve the prob-
lem” (International Environment Reporter 1988a, 111), and the NRDC prepared to
force the EPA to take action through the courts once again (Global Climate Change
Digest 1988). Even industry joined the calls for further international action, hop-
ing to delay or defeat unilateral measures (Carnevale 1988). It is hard to overesti-
mate the importance of industry’s role in pushing universal participation in ozone
depletion. Though industry did not create this vision of a global response, once it
was in place industry played a crucial role in fostering the evolution of US rule
models toward universal participation.15
These evaluations of the Montreal Protocol significantly weakened the
US internal rule model calling for a North-only response to ozone depletion.
The North-only rule had driven US behavior in the negotiations of a land-
mark agreement, but the Montreal Protocol was a landmark only as a bare be-
ginning in the fight against ozone depletion. With the participation rule
weakened by negative evaluation, the United States adopted a universal par-
ticipation rule. The United States adapted to the new system rule and began
advocating universal participation.
The new consensus on the system rule calling for universal participation
made a significant difference in how the bargaining proceeded. Participating in
earnest in the negotiating process for the first time in large numbers, the South
began pushing for developmental assistance at a 1989 London conference on the
ozone depletion problem attended by 123 states (UNEP 1989, paragraph 11). As
Litfin observed, “[T]he treatment of developing countries, which hitherto had
been considered a minor issue, became a central concern” (Litfin 1994, 129). It
was at this conference that the South became vocal. Daniel Arap Moi of Kenya
voiced the generic Southern concern when he stated, “Some nations will not
find it easy to forego the use of CFCs in their quest for industrialization” (Inter-
national Environment Reporter 1989a, 106). Representatives from India and China
were more direct. India’s spokesperson praised the “polluter pays” principle and
“made known the Third World’s doubts about the industrialized countries’ po-
litical will to come up with the required financial aid and technology transfers
for CFC technology” (International Environment Reporter 1989b, 169). A Chinese
official protested that Southern nations “resented the rich ‘telling them what to
do and not to do’” (International Environment Reporter 1989b, 169).
The work ahead was clear-cut. As Richard Smith of the US State Depart-
ment reported: “At Helsinki [the first meeting of the parties to the Montreal Pro-
tocol], it was clear that many developing countries want to participate but are
understandably concerned about the potential costs to their economies” (US
House of Representatives 1989, 77). It was clear that if the Northern states were
to achieve Southern cooperation, they would have to accede to an international
fund and technology transfer mechanisms—they would have to make side pay-
ments to the South. This was the bargain that the South, now fully engaged in
the negotiating process, was driving.
The United States was loath to submit to Southern demands, and the im-
passe threatened to scuttle the ozone negotiations. The South made it clear that
they would not participate without a fund, and the United States, in particular,
held on to its opposition to a new fund, advocating instead the use of the World
Bank and other existing institutions (Weisskopf 1990a, A21). Just prior to the
London meeting of the parties to the Montreal Protocol, the United States re-
lented and agreed to an independent fund with the caveat that any ozone fund
set up would set no precedent for other environmental issues (Weisskopf 1990b,
A1). At London, compromise ruled the day, and by June 29, 1990, a deal had
been struck (UNEP 1990).
The London Amendment to the Montreal Protocol contained the ulti-
mate phaseout by 2000 desired by all parties (and most nonparties), but the big-
ger accomplishment of the London negotiations was the establishment of the
Multilateral Fund and technology transfer mechanisms. The United States and
other Northern states agreed to pay the full incremental costs of the transition
away from CFCs incurred by the South, and more importantly, they agreed to
do this with new and additional funds administered by a new institution. This
was a huge victory for the South, because the new institution was to be jointly
controlled, rather than solely administered by donor states. With the funding
provision in place, the South (crucially India and China) agreed to the Montreal
Protocol, and the fight against ozone depletion became a truly global affair.
Transition through Complex Adaptation
Universal participation was not a natural or inevitable rule for the ozone deple-
tion regime. Nothing about ozone depletion or the interests of the major actors
radically changed between 1986 and 1988. What did change were the internal
rule models of the Southern states—how they perceived and acted toward the
ozone depletion problem. This change set off a coevolutionary transition in both
the internal rule models of the Northern states and the system rule for partici-
pation. Regime theory misses these dynamics and is thus hampered from under-
standing the bargaining that takes place in regime formation and beyond.
BEYOND REGIME THEORY?
Three questions remain, however. First, why should we care about the emergence
of universal participation and the US adaptation to this new rule? The second is
more fundamental to this chapter: why do we need complexity concepts to ad-

dress the puzzles of universal participation? Finally, how can the complexity-based
claims about regime transformation put forward here be rigorously assessed?
Why Study the Transformation of System Rules?
Substantively, the story of the emergence of universal participation is crucial, be-

cause the system rule for participation has an enormous influence over the sub-
stance of regime negotiations themselves. It shapes what issues are discussed and
even considered. Before Montreal, Northern issues of CFC reductions and re-
placement technologies dominated negotiating agendas. Southern issues of de-
velopment and technology transfer were virtually ignored. After Montreal, the
altered system rule for participation ushered in a transformed set of negotiations.
North-South issues came to dominate the agenda. Regime theory, which too
often takes the number and identity of the actors for granted, cannot capture this
dynamic transition, to the detriment of its ability to explain regime outcomes.
Further, system rules tend to lock in (Arthur 1994a; Ikenberry 2001) and,
through increasing returns (Pierson 2000), tend to be relatively stable. The uni-
versal participation rule that emerged in the ozone depletion regime ushered in
an era of global responses to environmental problems. Once this happened in
ozone depletion, other issues were also seen as requiring universal participation.
Examining and explaining the emergence of universal participation in the ozone
depletion regime aids our explanations of the formation of regimes beyond
ozone depletion, especially climate change (Hoffmann 2005).
Do We Need Complexity Concepts to Comprehend the

Emergence of Universal Participation?
It is not immediately clear that complexity is the only perspective that can ex-
plain the emergence of universal participation. Perhaps states were making ob-
jectively rational choices in response to the incentives and constraints they faced
in the ozone depletion issue. For example, the transition in US thinking could
be characterized as a rational updating of strategies/beliefs in light of new scien-
tific information or new understanding of economic interests. The United States
wanted to solve the ozone depletion problem, and a commitment to North-only
participation did not facilitate the accomplishment of this goal. Universal par-
ticipation was therefore the obvious choice: involve the South in negotiations
and actions toward a solution. Similarly, the Southern states’ decision could
have been a rational reaction to the imminent regulation of an important class
of chemicals.
However, close scrutiny reveals that a rational updating explanation falls

short. First, instrumental acceptance of universal participation would not di-
minish the influence of the nascent system rule for universal participation (see
Risse, Ropp, and Sikkink 1999 for a similar argument about norm acceptance).
In hindsight, Southern states certainly had good (rational) reasons to participate,
but those reasons were constant in the 1985–88 period when their behavior
changed. It took a transition in how they perceived the ozone depletion prob-
lem, not a rational calculation, to catalyze their participation. In addition, the
US “instrumental” acceptance can also be seen as adapting to a new context. The
US understanding no longer fit with the prevailing system rule, and so US un-
derstandings had to change (Bernstein 2001). Without the appropriate under-
standing of the foundations of the ozone depletion regime, the United States
could not hope to actively or effectively participate.
Second, the specific direction of the change in US definitions of ozone de-
pletion relied on the existence of a system rule requiring universal participation.
The United States did not calculate from a number of potential options that
committing to universal participation was the way to maximize its utility. In-
stead, the United States committed to universal participation because that re-
quirement had come to pervade the system. Multiple actors with widely varying
motivations all arrived at the same conclusion: the ozone depletion regime re-
quired universal participation. This was a coevolutionary process whereby the
system rules, altered by the participating Southern states, contained a new un-
derstanding of the ozone depletion problem.
No options beyond universal participation were considered, though plau-
sible alternatives are imaginable in hindsight. It was taken as natural for the
United States and other Northern states to pursue universal participation and
extend the ozone agreement to the Southern states to eliminate future damage
to the ozone layer. Crucially, however, rather than pursuing universal participa-
tion, the United States could have stayed the course of the Montreal Protocol
provisions or could have advocated limited Southern participation. Either
choice may have been a more effective way to meet its goals, especially given the
hegemonic position of the United States both in terms of traditional notions of
power and power in the ozone depletion issue itself.
First, the United States and Europe could have remained committed to a
North-only negotiated regime and pursued a coercive strategy to force Southern
states to accept the ozone depletion regime.16 It is not clear that the Southern
states had as much bargaining leverage as they are usually credited with, and it is
certainly not clear that the trade restrictions in the Montreal Protocol would not
have eventually brought the South on board anyway. In fact, this was the think-
ing at the EPA in the immediate aftermath of the Montreal Protocol negotia-
tions (International Environment Reporter 1988b, 226). It might have been a ratio-
nal choice to retain a North-only vision of the ozone depletion problem and let
the trade sanctions and fear of technological obsolescence “force” the South to
comply with the Montreal Protocol. In other words, why bother negotiating with
Southern countries if the provisions of the protocol combined with the power of
Northern nations would make Southern ascension to the regime inevitable?
Second, even if it was objectively rational to include Southern countries in
the regime formation process, this does not necessitate universal participation.
The United States and other Northern states could have worked to involve the
large Southern states. It could have been perfectly plausible to entice China,
India, and Brazil to agree to the Montreal Protocol and join the process. The
large Southern states were the main concern. According to Irving Mintzer of the
World Resources Institute, “if just four developing countries—China, India, In-
donesia, and Brazil—increase their domestic consumption of CFCs to the levels
allowed by the protocol, CFC production on a worldwide basis would double
from the 1986 base level” (quoted in the Los Angeles Times, March 6, 1). Dealing
with just these states, the ones with real leverage, likely would have been less ex-
pensive for the United States in terms of necessary development concessions and
perhaps would have been more efficient, avoiding the problems associated with
large negotiations.
Neither of these potential alternatives to universal participation was con-
sidered by the United States or other states, because once its original definition
of the ozone depletion problem was weakened, the United States (and other
Northern states) came to understand the problem as universal—it came to accept
the current understanding of the problem. The international community had al-
ready instantiated the system rule for universal participation, thus constraining
possible choices for US definitions/strategies. In addition to the growing inter-
est of Southern states, the system rule was also already enshrined at UNEP, and
the new participation rule was ensconced in the structure of the negotiations.17
The emergence of universal participation and US adaptation to it was not
rational in the traditional sense. Instead, it was a result of complex adaptive
processes. By conceiving of states as adaptive actors and tracing the dynamics of
complex systems, we gain greater understanding of the foundation of the ozone
depletion regime.
Empirically Testing the Complexity Explanation?
In this chapter, I have endeavored to demonstrate two points. First, a complex-

ity approach provides concepts and tools useful in meeting the challenges faced
in the study of global environmental regimes. Complex adaptation, internal rule
models, and coevolution form the foundation of micro- and micro/macro-

processes needed to frame our analyses. Second, these concepts and tools can
be successfully employed in empirical assessment, enhancing understanding of
environmental regimes. This is, however, a bare beginning. In this concluding
section, I briefly discuss directions for future and comprehensive complexity
work—a research program for rigorously assessing complexity explanations of en-
vironmental regime transformation.
Going beyond (neoliberal) regime theory with a complexity approach for
studying environmental regime emergence and transformation is not obviously
necessary. The brief empirical discussion above hints at but does not fully
demonstrate the superiority of a complexity approach over neoliberal regime the-
ory. The question that remains is how to craft a research agenda that would pro-
vide such a demonstration.
The two approaches have clear and quite distinct sets of assumptions.
Regime theory provides explanations of bargaining (and bargaining outcomes)
among static, rational actors in a static context. Environmental problems and their
solutions are assumed known, as are the interests of the states facing the problems.
A complexity approach to environmental regimes provides explanations of bar-
gaining among dynamic, adaptive actors in a coevolving context. Environmental
problems and their solutions are actively created and re-created as actors coevolve
with their context. These different foundations obviously provide for different em-
pirical expectations. Indeed, I chose the transformation in the ozone depletion
regime from North-only participation to universal participation precisely because
the two approaches differ in their expectations and explanations.
Direct comparison is somewhat difficult, because regime theory and a
complexity approach do not agree on what can change. For regime theory, system
rules are assumed and exogenous to the analysis. However, we can discuss ex-
pectations about participation (Koremenos, Lipson, and Snidal 2001) and why
they might change. Regime theory says such a transformation should come at the
behest of powerful actors working with new information—rational updating—and
the explanation is found in rational bargaining. A complexity approach views sys-
tem rules as malleable and endogenous to the analysis, arguing that the trans-
formation of participation rules emerged through the coevolutionary actions and
interactions of states.
So does the complexity approach go beyond regime theory? What steps are
required to test the complexity explanation? First, it is at least plausible on its
face that a complexity approach can provide a superior explanation of regime dy-
namics. As noted above, the transformation of the regime from requiring North-
only to universal participation is an (important) anomaly for regime theory.
Without this anomaly, we would not need to consider complexity; and if, as in
this case, the complexity explanation can accommodate the anomaly, we have
gone beyond regime theory.
A rigorous test of the complexity explanation begins with a method not
discussed explicitly in this chapter: formal modeling. Proponents of a complex-
ity approach (in the social sciences and beyond) have developed a set of com-
puter simulation tools that offer a “laboratory” traditionally denied to social
science (see, e.g., Epstein and Axtell 1996; Hoffmann 2005; Axelrod 1997; Ce-
derman 1997). These agent-based models allow proponents of a complexity par-
adigm to rigorously assess the logic of their arguments—putting a proposed
explanation onto the computer forces one to explicitly define critical assump-
tions. Such simulation analysis facilitates the discovery of boundary conditions,
unexpected hypotheses, and perhaps most importantly, understanding of under
what conditions we can expect to see certain outcomes (expected or otherwise).
Joshua Epstein (1999) claims that computer simulation experiments provide re-
searchers with rigorously arrived-at “candidate explanations” for social phenom-
ena. Where regime theory uses game theory to provide a rigorous foundation for
its explanations of bargaining, a complexity approach uses agent-based modeling.
Thus, the logic of a complexity explanation can be assessed formally and at least
as rigorously as the regime theory explanation.
Regarding the ozone depletion regime, I have in other places (Hoffmann
2002, 2005) reported on agent-based modeling experiments designed to explore
the emergence and evolution of system rules (or norms) through the interactions
of coevolving agents. These experiments demonstrated the abstract plausibility,
though not the empirical validity, of the explanation for regime transformation
developed in this chapter.
With confidence in the logical soundness of the explanation based on com-
plexity concepts, empirical assessment of the complexity explanation of the ozone
depletion regime requires further and detailed process tracing (see Hoffmann
2005 for an effort in this direction). This is a nontrivial task, and it is no wonder
that, with laudable exceptions (Brunk 2002; Cederman 2003; Jervis 1997), com-
plexity scholars have shied away from empirical work. A full account of the emer-
gence and influence of universal participation requires an analysis of the
coevolution of multiple actors’ rule models with the participation requirements,
as well as the coevolution of the actors themselves. Complexity processes are far
from parsimonious, and a full comprehension of a complex system requires thick
description and rich empirical detail. Specifically, rigorous testing of the explana-
tion proposed here would entail tracing the development and adaptation of the
European Union’s and Southern states’ rule models in addition to that of the
United States. In addition, more attention needs to be paid to the dynamics of
the rule models themselves, exploring the domestic and global political processes
through which actors define their rule models, evaluate their behavior, and alter
their rule models. As discussed above, complex adaptation is an abstract model of
actor behavior that needs to be fleshed out with significant empirical detail.
Combining agent-based modeling experiments and empirical process trac-
ing (other empirical methods can be used as well, depending on the research
question) in a recursive process provides the most rigorous test of the explana-
tion. The simulation experiments inform the process-tracing case studies, and
the empirical work feeds back to inform further modeling. With such testing we
can assess whether or not the complexity explanation asserted and initially ex-
plored above actually addresses the anomaly of universal participation, and we
can demonstrate that a complexity approach takes us beyond regime theory.
NOTES
1. A special issue of International Organization from 2001 (vol. 55, no. 4) is

entirely dedicated to exploring rational choice mechanisms in regime emergence
and design.
2. Emergence is an oft-debated and imprecise concept. For an introduc-
tion to emergence and emergent processes, see Holland 1998.
3. Agents are not always or even usually treated as unitary. When dealing
with meta-agents (agents composed of other agents), subagents within the agent
do evaluation. For instance, environmental groups evaluate the outcome of ne-
gotiations that the United States participates in.
4. The evaluation stage of complex adaptation is crucial. At this stage,
agents alter their rule models, which is key for understanding how ideas become
an ingrained part of internal rule models. In addition, however, this stage adds
variation in a population of agents, because different agents may have different
evaluation processes and different criteria for fitness.
5. Internal rule models can be difficult to operationalize empirically, be-
cause they are inherently unobservable. However, one advantage to studying
large, corporate actors such as states is that such agents usually write down the
understandings that comprise their rule models. As a proxy for the actual (un-
observed) rule models of the United States, for instance, I treat the negotiating
positions that the executive branch used in the various negotiations as the rule
models. These negotiating positions reveal how the United States defined the
problems over time as well as what it desired out of agreements and what it was
willing to commit to. In order to get a full sense of the negotiating position
for any particular set of negotiations, I utilized several sources in an attempt to
“triangulate” and get a true picture (US Congressional documents, UN meeting
reports, newspaper accounts, and interviews). The system rules are equally
difficult to directly observe. They are essentially the norms/structures or, in the
broadest sense, rules that are external to the United States. I also triangulate
around these system rules from several sources (UN documents, newspaper ac-
counts, and interviews).
6. For in-depth analysis of the ozone depletion regime, see Tolba 1998;
Benedick 1991; Litfin 1994; Rowlands 1995; and Hoffmann 2005.
7. The dominant way to approach this change in the literature is to ana-
lyze the ozone depletion regime before and after Southern states joined the pro-
ceedings. Unfortunately, almost no attention is paid to the transition. This is
likely an artifact of the rational choice approach’s propensity to downplay history
and to treat each bargaining situation as if no interactions occurred before the
current negotiation. See Hoffmann 2005; and Mitchell and Keilbach 2001.
8. For an elaboration on this discussion of the formation of the ozone
depletion regime, see Hoffmann 2005.
9. Participation is obviously just one of a number of important sys-
tem rules.
10. Interview with Paul Horwitz, US EPA.
11. Interview with Stephen Seidel of US EPA. The Toronto Group con-
sisted of the United States, Canada, and the Scandinavian countries.
12. Text of Vienna Convention—reprinted in Benedick 1991, 218–29.
13. For more on the Southern transition to participation, see Hoffmann
2005; Morrisette et al. 1991; and Downie 1995.
14. Interview with Paul Horwitz.
15. Competitiveness was the main concern of industry, but this nonethe-
less led them to advocate universal participation as the solution for the ozone
depletion problem.
16 Such coercive strategies are well within the bounds of rational choice
predictions—see Mitchell and Keilbach 2001.
17. UNEP report printed in US House of Representatives 1989, 1050.
REFERENCES
Arthur, Brian. 1994a. Increasing Returns and Path Dependence in the Economy. Ann
———. 1994b. “Inductive Reasoning and Bounded Rationality.” AEA Papers and
Proceedings 84, no. 2:406–11.
Arthur, W. Brian, Steven N. Durlauf, and David A. Lane, eds. 1997. The Econ-
omy As an Evolving Complex System II. Reading, MA: Addison-Wesley.
Axelrod, Robert. 1997. The Complexity of Cooperation: Agent-Based Models of Com-
petition and Collaboration. Princeton, NJ: Princeton University Press.
Benedick, Richard. 1991. Ozone Diplomacy: New Directions in Safeguarding the
Planet. Cambridge, MA: Harvard University Press.
Bernstein, Steven. 2001. The Compromise of Liberal Environmentalism. New York:
Columbia University Press.
Brunk, Gregory. 2002. “Why Do Societies Collapse? A Theory Based on Self-
Organized Criticality.” Journal of Theoretical Politics 14, no. 2 (April):
195–230.
Carnevale, Mary Lu. 1988. “Du Pont Plans to Phase Out CFC Output.” Wall
Street Journal, March 25.
Cederman, Lars-Erik. 1997. Emergent Actors in World Politics. Princeton, NJ:
Princeton University Press.
———. 2003. “Modeling the Size of Wars: From Billiard Balls to Sandpiles.” Amer-
ican Political Science Review 97, no. 1 (February): 135–50.
Congressional Quarterly Weekly Report. 1987. “Members Hail Ozone Agreement.”
September 19, 2283.
Downie, David Leonard. 1995. “UNEP and the Montreal Protocol.” In Interna-
tional Organizations and Environmental Policy, ed. Robert Barlett, Priy A.
Kurian, and Madhu Malik, 171–86. London: Greenwood Press.
Epstein, Joshua. 1999. “Agent-based Models and Generative Social Science.”
Complexity 4, no. 5:41–60.
Epstein, Joshua, and Robert Axtell. 1996. Growing Artificial Societies: Social Science
from the Bottom Up. Washington, DC: Brookings Institution Press.
Gell-Mann, Murray. 1994. “Complex Adaptive Systems.” In Complexity: Meta-
phors, Models, and Reality, ed. George Cowan, David Pines, and David
Melzer. New York: Addison-Wesley.
Global Climate Change Digest. 1988. “NRDC Opposition to CFC Ruling.” Vol. 1,
no. 3 (September), www.globalchange.org/gccd/gcc-digest/d88sep3.htm.
Haas, Peter. 1992. “Banning Chlorofluorocarbons: Efforts to Protect Stratos-
pheric Ozone.” International Organization 46, no. 1:187–224.
Hasenclever, Andreas, Peter Meyer, and Volker Rittberger. 1997. Theories of Inter-
national Regimes. Cambridge: Cambridge University Press.
Hoffmann. Matthew J. 2002. “Entrepreneurs and the Emergence and Evolu-

tion of Social Norms.” In Third Workshop on Agent-Based Simulation,
ed. Christoph Urban, 32–37. Ghent, Belgium: SCS-European Pub-
lishing House.
———. 2005. Ozone Depletion and Climate Change: Constructing a Global Response.
Albany: SUNY Press.
Holland, John. 1995. Hidden Order. New York: Addison-Wesley.
———. 1998. Emergence: From Chaos to Order. Cambridge: Perseus Books.
Ikenberry, G. John. 2001. After Victory: Institutions, Strategic Restraint, and the
Rebuilding of Order After Major Wars. Princeton, NJ: Princeton Univer-
sity Press.
International Environment Reporter. 1988a. “Companies Should Plan on Bigger
Cuts in CFC Production, Consumption.” Vol. 11, no. 2 (February):
111.
———. 1988b. “Plans for Assessments Under Protocol Should Begin in Fall,
Thomas Writes Tolba.” Vol. 11, no. 4 (April): 210–11.
———. 1989a. “Over One-Hundred Nations Gather in London to Focus on Need
for Tighter Limits on CFCs.” Vol. 12, no. 3 (March): 106.
———. 1989b. “Nations Agree on Risks From CFC Use, But Not on Speed of Ad-
ditional Control Steps.” Vol. 12, no. 4 (April): 169.
Johnston, Iain. 2005. “The Power of Interpretive Communities.” In Power and
Global Governance, ed. Robert Duvall and Michael Barnett. Cambridge:
Cambridge University Press.
Jervis, Robert. 1997. System Effects. Princeton, NJ: Princeton University Press.
Keohane, Robert. 1984. After Hegemony: Cooperation and Discord in the World
Political Economy. Princeton, NJ: Princeton University Press.
Koremenos, Barbara, Charles Lipson, and Duncan Snidal. 2001. “The Rational
Design of Institutions.” International Organization 55, no. 4: 761–99.
Krasner, Stephen. 1983. International Regimes. Ithaca, NY: Cornell University
Press.
List, Martin, and Volker Rittberger. 1992. “Regime Theory and International
Environmental Management.” In The International Politics of the Envi-
ronment, ed. Andrew Hurrell and Benedict Kingsbury. New York:
Oxford University Press.
Litfin, Karen. 1994. Ozone Discourses. New York: Columbia University Press.
Miller, Marian A. L. 1995. The Third World in Global Environmental Politics. Boul-
der, CO: Lynne Reiner.
Mills, Mike. 1988. “But Some Say ‘Too Little Too Late’: Ratification of Ozone
Pact Recommended.” Congressional Quarterly Weekly Report, February
20, 370.
Mitchell, Ronald, and Patricia Keilbach. 2001. “Situation Structure and Institu-
tional Design: Reciprocity, Coercion, and Exchange.” International
Organization 55, no. 4: 891–917.
Morrisette, Peter M., Joel Darmstadter, Andrew Patinga, and Michael Toman.
1991. “Prospects for a Global Greenhouse Gas Accord: Lessons from
Other Agreements.” Global Environmental Change 1, no. 3 (June):
209–23.
Parson, Edward. 1993. “Protecting the Ozone Layer.” In Institutions for the Earth:
Sources of Effective International Environmental Protection, ed. Peter Haas,
Robert Keohane, and Marc Levy, 27–74. Cambridge, MA: MIT Press.
Pierson, Paul. 2000. “Increasing Returns, Path Dependence, and the Study of
Politics.” American Political Science Review 94, no. 2: 251–68.
Rajan, Mukund Govind. 1997. Global Environmental Politics: India and the North-
South Politics of Global Environmental Issues. Delhi: Oxford University
Press.
Risse, Thomas, Stephen Ropp, and Kathryn Sikkink, eds. 1999. The Power of
Human Rights: International Norms and Domestic Change. Cambridge:
Cambridge University Press.
Rowlands, Ian. 1995. The Politics of Global Atmospheric Change. New York: Man-
chester University Press.
Sims, Holly. 1996. “The Unsheltering Sky: China, India and the Montreal Pro-
tocol.” Policy Studies Journal 24, no. 2:201–14.
Tolba, Mostafa. 1998. Global Environmental Diplomacy: Negotiating Environmental
Agreements for the World, 1973–1992. Cambridge, MA: MIT Press.
UNEP. 1986. Draft Report of the Ad Hoc Working Group on the Work of its First Session,
Geneva, 1–5 December 1986. UNEP/WG.151/1.4/draft_rep_of_the_
first_sess.
———. 1987a. Report of the Ad Hoc Working Group on the Work of its Second Session, Vi-
enna, 23–27 February 1987. UNEP/WG.167/2/Report_of_wg_on_its_
2nd_session.
———. 1987b. Report of the Ad Hoc Working Group on the Work of Its Third
Session, Geneva, April 1987. UNEP/WG.172/2/Report_of_wg_on_its_
3rd_session.
———. 1987c. Montreal List of Participants, 4 September 1987. UNEP/cpp_list_of_par
ticipants_Montreal.
———. 1989. Report of the Parties to the Montreal Protocol on the Work of Their First
Meeting, May 6, 1989. UNEP/Ozl.Pro.1/5.
———. 1990. Report of the Second Meeting of the Parties to the Montreal Protocol on Sub-
stances That Deplete the Ozone Layer, June 29, 1990. UNEP/OzL.Pro.2/3.
US House of Representatives. 1989. Ozone Layer Depletion, H.R. Comm. Print
1989.
Weisskopf, Michael. 1990a. “Administration Defends Resistance to Plan for
Helping Third World Cut CFCs.” Washington Post, May 10, A21.
———. 1990b. “US Drops Opposition to CFC Phaseout Fund; Business, Foreign
Leaders Had Urged Reversal.” Washington Post, June 16, A1.
Young, Oran. 1994. International Governance: Protecting the Environment in a State-
less Society. Ithaca, NY: Cornell University Press.
———. 1997. Global Governance: Drawing Insights from the Environmental Experience.
Cambridge, MA: MIT Press.
———. 1999. Governance in World Affairs. Ithaca, NY: Cornell University Press.
———. 2002. The Institutional Dimensions of Environmental Change: Fit, Interplay, and
Scale. Cambridge, MA: MIT Press.
CHAPTER 6
Agent-Based Models in the

Study of Ethnic Norms and Violence
Ravi Bhavnani
During the nightmarish April, May, and June of 1994 following the assassina-
tion of President Juvénal Habyarimana, some five hundred thousand to eight
hundred thousand Tutsi were killed by civilian militias and the Hutu-dominated
army in a matter of weeks. Moderate Hutu who failed to participate in the vio-
lence were, time and time again, instructed to kill their Tutsi neighbors or face
death at the hands of Hutu militias. When confronted with the choice of killing
ethnic others or being killed by members of their own ethnic group, individuals
with no prior disposition to engage in ethnic violence were turned into efficient
killing machines. Few refused to participate in the killing; the result that “neigh-
bors hacked neighbors to death in their homes, and colleagues hacked colleagues
to death in their workplaces . . . doctors killed patients, and school teachers
killed their pupils” (Gourevitch 1998, 115).1 The end result was that among the
Hutu, killing Tutsi became the norm, and similar behavioral norms have moti-
vated mass participation in, or complicity with, group violence in settings as di-
verse as Cambodia, Guatemala, Northern Ireland, and the former Yugoslavia.
While Rwanda’s culture has been described as one of fear and conformity,
this explanation does not to do justice to the level of participation—anywhere be-
tween two hundred thousand and five hundred thousand Hutu participated in
the genocide (Des Forges 1999; Mamdani 2001)—or the vehemence with which
Tutsi were massacred, leading events in Rwanda to be described as the “fastest
killing spree of the 20th Century” (Power 2001, 84). As Gourevitch (1998, 96)
notes, “Every Rwandan I spoke with seemed to have a favorite, unanswerable
question. For Nkongoli, it was how so many Tutsi had allowed themselves to be
121
122 RAVI BHAVNANI
killed. For Francois Xavier Nkurunziza, a Kigali lawyer, whose father was Hutu
and whose mother and wife were Tutsi, the question was how so many Hutu has
allowed themselves to kill.” The scale of violence was simply unprecedented, and
stands in marked contrast to violence in 1963 in which roughly thirteen thou-
sand lives were lost. Thus, “To believe that ordinary Rwandans killed, in their
hundreds and thousands, and perhaps more, because of a congenital transhis-
torical condition—‘a culture of fear’ or of ‘deep conformity’—would require
stretching one’s sense of credibility” (Mamdani 2001, 200). In Rwanda, Hutu re-
sistance to the killing was evident at both the individual and community level
(Des Forges 1999). Rather than being driven by fear of Tutsi, it was fear of fellow
Hutu that drove the reluctant to participate in the genocide.
Likewise, there is little doubt that structural factors that pertain to the econ-
omy, state capacity or penetration, or international aid flows—to name but a few—
have important implications for the nature and onset of violence. Yet the
conventional preference for tracking structural factors—which either tend to remain
constant or are replicated to some degree in most episodes of conflict—is overstated.
For one thing, prior levels of violence are inadequate predictors of future levels of
violence. Cities, regions, states, and countries are not inherently peaceful or prone
to interethnic violence. Rather, the scale and duration of violence inevitably vary
over time and across social contexts, as exemplified by the relatively localized and
contained episode of violence in Rwanda in 1963.2 Also explanations that empha-
size the role of a particular factor or triggering event—such as the assassination of
President Habyarimana—point to the correlation between the magnitude of the cat-
alyst and the scale of violence, but need to clarify why violence can erupt in the
absence of such a catalyst, or why similar catalysts can lead to different outcomes.
It follows that an adequate explanation for mass participation by reluctant
Hutu in the Rwandan genocide must address the associated issues of why they
participated in the killing, how they were persuaded to participate, and the effect
of widespread participation on the scale and duration of violence. In contrast
to explanations that point to a culture of conformity or highlight the importance
of structural factors, my contention is that mass participation by reluctant Hutu
in violence directed at Tutsi can be explained by the emergence of a violence-
promoting norm among the Hutu community at large. I argue that a complexity
theory with its simulation by agent-based modeling lends itself well to the study
of ethnic norms—behavioral norms defined in ethnic terms that effectively per-
suade members of an ethnic group to participate in violence against nominal
rivals. An agent-based model (ABM)—defined in terms of entities and dynamics
at the microlevel—can be used to explore why such behavioral norms emerge in
only some conflicts, prevail in some ethnic groups but not in others, and why
these norms can either promote interethnic violence or cooperation.
AGENT-BASED MODELS IN THE STUDY OF ETHNIC NORMS 123
In the next sections of this chapter, I briefly describe the heterogeneity of

Hutu beliefs and behavioral dispositions, the manner in which intragroup inter-
action increased behavioral conformity, and the subsequent emergence of a vio-
lence-promoting norm during the genocide in 1994. I then provide a brief intro-
duction to ABMs, discuss my preference for an exploratory as opposed to a consol-
idative modeling approach (Bankes 1994; Casti 1997), and describe how ABMs
may be used to study processes of norm formation and change within ethnic
groups. I conclude the chapter with a comparison of agent-based, game-theoretic,
and equation-based approaches.
RWANDA 1994: HETEROGENEITY,

INTERACTION, AND ADAPTATION
Rwanda is a densely populated, heterogeneous society, characterized by relatively

frequent interactions (and intermarriage) between members of the two major eth-
nic groups. As a result, one may reasonably assume that individuals varied in their
level of extremism and thus in the extent to which they harbored antipathy for
nominal rivals or believed they posed a threat.3 Given the variation in levels of
Hutu extremism, complicity with the state’s genocidal agenda was initially low
(Des Forges 1999). This changed remarkably over the course of the genocide de-
spite high expectations of behavioral conformity among the Hutu—expectations
that were clearly and repeatedly broadcast by the extremist regime and interhamwe.
The end result was that large numbers of Hutu killed Tutsi who were ac-
quaintances, colleagues, or, in extreme cases, relatives although individuals adapted
to behavioral expectations in different ways (Des Forges 1999). For instance, reluc-
tance to participate was manifest in inconsistent behavior:
The most ambivalent stories of the genocide I heard from survivors

were about Hutu who saved a friend or colleague in one place, only to go
and join the killing in another. . . . Could they have killed under duress—
knowing that if they refused or even appeared reluctant, they would surely
be killed—and saved a life when the opportunity presented itself ? (Mam-
dani 2001, 221)
In other instances, behavioral adaptation was more complete:
Everyone was called to hunt the enemy. . . . But let’s say someone is reluc-
tant. Say that guy comes with a stick. They tell him, “No, get a masu.” So,
OK, he does, and he runs along with the rest, but he doesn’t kill. They say,
“Hey, he might denounce us later. He must kill. Everyone must help to kill
124 RAVI BHAVNANI
at least one person.” So this person who is not a killer is made to do it.
And the next day it’s become a game for him. You don’t need to keep
pushing him. (Gourevitch 1998, 24)
Behavioral expectations were initially formed at the local (neighborhood,

community, village or town) level, and updated as roving bands of interhamwe
motivated communities that were less than zealous in their willingness to mas-
sacre the local Tutsi population. These expectations were underscored in no
small measure by Radio Milles Collines, RTLM, and Radio Rwanda broadcasts
of the notorious ten commandments which included calls for Hutu unity in the
face of the common Tutsi enemy, calls for all Hutu to stop being merciful and
undertake umuganda—a reference to customary work symbolizing the killing of
Tutsi, and the message that all Hutu should either kill Tutsi or be killed (Des
Forges 1999; Prunier 1995).
In the undeniably complex pattern of interaction (Hutu-Hutu, Hutu-
Tutsi) during the genocide, one key element that stands out is the random na-
ture of encounters. When a population mixes randomly, extremists eventually
have the opportunity to interact with moderates, and observe and punish their
behavior. Moderates, in turn, quickly update their behavior to conform to group
practice, and in so doing reinforce this behavior. For instance, no Hutu (moder-
ate or otherwise) could be sure of whom they may encounter or trust. Many
Hutu were turned in by relatives or neighbors, who bore grudges against them,
sought favors, or coveted their property. As a result, escaping became more dif-
ficult for members of the targeted group—moderate Hutu and all Tutsi. If a per-
son was let go by one party, he or she would be caught and killed by another.
The threat of sanctions therefore extended the violence over a wider range of tar-
gets. Whereas some Hutu may have willingly redressed their grievances against
particular Tutsi, under these conditions they were pressed to be indiscriminate.
While this brief discussion does not do justice to the intricacies or com-
plexities of the 1994 Rwandan genocide, it does underscore the importance of
studying violence from the bottom up—to specify behavior at the level of the
individual; to capture the heterogeneity and adaptation of individual prefer-
ences; and to assess the importance of group networks that determine who in-
teracts with, observes, and punishes whom and how often.
Alternatively stated, the causes of conflict constitute partial explanations at
best, and are inextricably linked to the process by which conflict unfolds. It is the
process an ethnic group goes through in the transition from conflict to violence
that determines whether an episode progresses beyond the common occurrence
of low-level violence to more isolated instances of high-scale violence. An expla-
nation that affords due attention to process therefore needs to be specified at the
level of the individual, to be capable of explaining phenomena at the aggregate
or group level, and to incorporate feedback mechanisms from the aggregate back
to the individual level.
ETHNIC NORMS
As baffling as the scale of violence in Rwanda is the number of those purported

to have participated in the killing. How, then, does one go about explaining the
participation of hundreds of thousands of Hutu? Why were similar levels of par-
ticipation not evidenced in previous episodes of Hutu-Tutsi violence? The ex-
planation I advance is rooted in the concept of an ethnic norm comprised of two
basic components: agreed-upon behavior and punishment for deviation from
this behavior (Axelrod 1986). In the context of interethnic rivalry, a norm thus
defined either compels or dissuades members of a group from engaging in vio-
lence against ethnic rivals. By implication, the strength of a norm dictates how
typical this behavior is of the group as a whole.
Individual members of an ethnic group are motivated to participate in or
oppose violence against nominal ethnic rivals for a variety of reasons. Individu-
als may follow their own convictions or beliefs, seek revenge for prior acts of vi-
olence, follow or go with the crowd, or simply derive entertainment value from
participation. Where individual motives are absent or insignificant, cultural
models—tacit knowledge structures or schemas that are both widely shared by
members of a social group and induce participation by a large number of indi-
viduals—may be used by “ethnic entrepreneurs” as motivating templates for in-
terethnic violence or cooperation.4 These schemas may emphasize obedience (an
inherent rationale that orders are to be followed, with an allowance, albeit small,
for disobeying orders that are contrary to any logic); detachment (cutting off one’s
feelings for an “enemy”); paranoia (fear of being killed by an enemy if the enemy
is not killed first); revenge (“tit-for-tat” behavior); patronage or work (a duty with a
quid pro quo of material rewards); honor (saving face); or religious duty (such as a
jihad or the building of a Hindu Rashtra). Each of these schemas may transform
or undermine traditional constraints on violence or (in the case of cooperation)
introduce constraints where none previously existed.5
Yet while group consciousness is necessary to ensure collective action, it
often proves insufficient when the individual costs of participation are high—
a point that is underscored by Mamdani, who notes that “fear could silence oppo-
sition, but it could not generate enthusiasm (for killing) . . . as they grew in scope,
the massacres targeted anyone, peasant or professional, who refused to join in the
126 RAVI BHAVNANI
melee” (2001, 219). Norms, on the other hand, clearly prescribe appropriate or ex-
pected behavior, and leave less room for individual choice. Given that the exis-
tence of an ethnic norm is a matter of degree—norms may be strong, weak, or
simply absent—the severity of punishment directly affects the strength of the norm.
In addition to punishments, the emergence of ethnic norms depends critically
upon the composition of an ethnic group (the number of those who support or op-
pose a particular course of action and their influence within the group) as well as
the speed with which behavioral expectations are transmitted to coethnics (how
frequently individuals observe, interact with, and enforce each other’s behavior).
It follows that while punishment, group composition, and social structure
all influence norm formation and change, it is difficult to attribute the emer-
gence of a norm to any specific factor, since norms are an emergent property of
social systems and their existence depends upon complex patterns of interaction,
influence, and internalization among individuals.6 To capture the emergent
properties of norm creation and change, I turn to complexity theory and its sim-
ulation by ABM.
AGENT-BASED MODELS
Generally, ABM are comprised of one or more types of agents, as well as a non-
agent environment in which the agents are embedded. The profile or “state” of
an agent can include various characteristics and preferences, as well as particu-
lar social connections (i.e., identities, memberships, networks) and a memory of
recent interactions and events.7 In addition to individual characteristics, agents
are defined by their decision-making heuristics and capabilities to act in response
to inputs from other agents and from the environment. Agents may also possess
adaptive mechanisms (learning or evolutionary) that lead them to change their
heuristics based on their own experience. Each agent’s behavior affects other
agents as well as the nonagent environment, resulting in behavioral change at
the group or system level.
The nonagent environment can encompass any variables external to the
agents that are relevant to behavior, ranging from physical features such as geog-
raphy or topography to things comprising states of the world like political, eco-
nomic and social conditions. An environment, therefore, is specified in terms of
various entities or dimensions, each with an associated “state.” The environ-
mental entities in a model usually have their own dynamics, describing how they
change over time independent of agent behavior. These variations can reflect
natural progressions (or regressions) according to logical rules and also involve
uncertainty or noise. In addition, they could represent the effects of shocks or
“triggers” such as sudden economic collapse, the mobilization of ethnic rivals, or
a military invasion. Besides following its own dynamic rules, the environment
may also adapt in response to agent behavior. The existence of interlaced feed-
back relationships—agent-agent and agent-environment interaction—leads to the
nonlinear, path-dependent dynamics that are characteristic of complex systems.
The model’s dynamics are studied by implementing the agents and the en-
vironment as a computer program. One then runs the program to simulate the
behavior of the agents and the dynamics of the environment. When an ABM is
simulated on a computer, agent behavior is generated as agents determine which
other agents to interact with, what to do when they interact, and how to interact
with the environment. The output from model simulations consists of both the
microlevel behavior of agents and changes in the environment, as well as the
emergent macrolevel structures, relationships, and dynamics that result from
the aggregation of this microlevel activity. In principle, the simulation can be run
hundreds or thousands of times—with various tracking measures or outcome vari-
ables summarized across runs—to study the variations in and sensitivity of results.
ABMs are well suited for studying dynamic processes—such as emergence
and spread of ethnic norms—that are sensitive to both historical contingencies
and situational factors. For instance, an exploratory model can serve as an ex-
perimental device to examine how members of an ethnic group might behave
under a variety of assumptions, while stopping short of offering precise and
detailed forecasts of how they will act given a particular set of circumstances.8
Consolidative modeling, on the other hand, usually involves the develop-
ment of “model” systems that represent “real-world” systems with easily measur-
able physical characteristics and components. These models often require
exhaustive inputs, which are then processed with computer programs that can
run to millions of lines of code. Ideally, this large amount of data can be trans-
formed into a useful or manageable form, but often the outputs are quite de-
tailed as well. Although most often applied in settings where each component
has clear physical properties, examples of a consolidative approach include com-
plex multiagent models that use agent architectures based on “naturalistic deci-
sion-making.” The consolidative approach, however, is largely inappropriate for
studying violent conflict, which is characterized by significant information
uncertainties and practical barriers to experimental validation.
MODELING THE EMERGENCE OF ETHNIC NORMS
Constructing an ABM to explore the dynamics of norm formation and change

within an ethnic group is part art and part science—the exact proportions of
which can vary greatly. The specification of agent attributes, decision rules, adap-
tive mechanisms, and the agent-interaction topology all require decisions on the
128 RAVI BHAVNANI
part of the modeler to include essential components and mechanisms that cap-
ture the problem at hand, while leaving out much that is important (as well as
much that is peripheral).9 In this section, I provide a high-level description of a
simple ABM to study the emergence of ethnic norms, and discuss how such a
model could be applied to analyze the conditions under which mass participa-
tion in ethnic violence occurs.
At the most basic level, the model may include agents who vary in terms of
their out-group extremism (their disposition to engage in or oppose violence against
nominal ethnic rivals) and their level of in-group tolerance (their propensity to pun-
ish coethnics for failing to adhere to their own, or externally defined, behavioral
standards). As a result, one may have agents who both engage in (or refrain from)
violence and punish those who fail to engage in (to refrain from) violence, agents
who simply act upon their own preferences (to engage in or refrain from vio-
lence) without punishing others, “hypocrites” who punish others for their failure
to conform to behavioral expectations but personally shirk, as well as agents who
remain neutral in the face of individual or group pressure.
In addition, agents may vary in their levels of in-group extremism (the
strength of punishments administered to coethnics). As noted earlier, intragroup
punishments can vary in severity from killing coethnics, threats to individuals,
family members, and relatives, and public humiliation to the destruction of per-
sonal property and the loss of one’s status within the group. The precise mix of
punishment (and reward), variation in the strength of these inducements, and
distribution of agents willing to apply these measures may all be designed to cap-
ture the specific mechanisms used by groups to induce collective compliance
(Oliver 1984; Kandori 1992; Posner and Rasmusen 1999). Finally, agents may
have different degrees of in-group influence, that is, what effectively distinguishes
ethnic entrepreneurs or leaders from other agents.
Such a model also makes it possible to embed agents in a set of social re-
lationships to determine interaction patterns within the group. Most ethnic
groups face collective action problems and these problems are likely to be more
pronounced when the costs of participation or compliance are high. As a result,
punishment often is used to bring individual behavior into conformity with
group practice and who talks to, observes, interacts with, or ultimately sanctions
whom is of critical importance. As a result, one is principally interested in how
and how often “like-minded” individuals observe and sanction the behavior of
individuals with contrasting or opposing views (Granovetter 1976; Tilly 1978;
Marwell, Oliver, and Prahl 1988; McAdam and Paulsen 1993; Opp and Gern
1993; Watts and Strogatz 1998).10 Networks that connect members of an ethnic
group to one another therefore become instrumental in determining the success
of efforts designed to achieve collective compliance and a variety of network

types—ranging from random to small world networks in which “ethnic entrepre-
neurs” have a disproportionate level of influence on the group—can be specified.
One also may “grow” such networks endogenously, based on factors such as spa-
tial proximity, past interaction, or similar (or different) preferences.11
It follows that individual agents may adapt their behavior in a variety of
ways depending on the nature of the encounter, the perceived punishment for
deviating from group behavior, the information environment, and the structure
of social relations or networks that characterize the group. For instance, agents
may update their behavior to conform to the expectations of other agents they
encounter, or to adhere more closely to perceptions of local behavior, or to be-
havioral expectations broadcast to the group as a whole. One way to capture be-
havioral change at the level of the individual is to specify a set of update rules
that determine the degree of behavior modification based upon factors such as
the individual agent’s current behavior, strength of punishment administered,
behavior or type of the punishing agent, and prevailing local or global behavior.
The agent-based model outlined above generally is implemented as a com-
puter program in which agents interact with and influence each other, resulting
in more or less conformity within the group. Such a model provides the user
with a great deal of flexibility, making it possible to alter the composition of an
ethnic group to represent domination by extremists or pacifists, or by highly con-
tentious or largely apathetic groups. Alternatively, one may choose to specify
agent characteristics independently, assume some correlation between out-group
extremism and in-group tolerance, as well as in-group tolerance and in-group ex-
tremism, or introduce a high degree of correlation across all agent characteristics.
One may also alter individual update rules, the interaction topology, and the
number of agents with a disproportionate level of influence on group members.
Spatial relationships can bias which agents are more likely to interact with other
agents. By explicitly specifying the structure of a social network—designed to cap-
ture patterns of interaction at the level of local communities, neighborhoods, or
larger spatial configurations—it becomes possible to assess the extent to which
network structure hinders or promotes behavioral conformity. Finally, by speci-
fying measures of aggregate behavior to capture the emergence of ethnic norms,
the model outlined here may be used to generate hypotheses linking group com-
position, network structure, and punishment regimes to norm formation and
change. One may therefore seek to determine whether norms are equally likely
to form in groups with similar aggregate preferences; under what conditions so-
cial structure affects the emergence and maintenance of norms; or whether
norms can emerge in the absence of punishment.
130 RAVI BHAVNANI
MODELING CHOICES
Because other formal approaches—namely, equation-based and game-theoretic

models—are available to study processes of norm formation and change, I con-
clude this chapter with a rationale for the use of ABM.
ABM vs. Equation-Based Models
An equation-based model (EBM) consists of a set of equations that usually ex-

press relations among observables. The evaluation of these equations produces
the evolution of the observables over time. These equations may be algebraic, or
they may capture variability over time (ordinary differential equations as used in
systems dynamics), or over time and space (partial differential equations).
For several reasons, ABMs are better suited to modeling complex adaptive
systems than EBMs (Parunak et al. 1998). First, ABMs are distinct in that they
are constructed in a “bottom-up” manner—specified at the level of individual
agents and their interactions with each other and the environment. By contrast,
EBMs generally focus on macrolevel entities and relationships, because equa-
tions at this level are easier to handle analytically and aggregate variables are
among the few observables that are consistently available. Thus, ABMs are capa-
ble of providing insight into the behavior of individual agents, whereas EBMs
are disposed to treat their actions as a “black box.”
Second, with ABMs one can accommodate myriad differences in agent
and environmental characteristics, as well as processes of change and adaptation
within the same model. Most EBMs instead employ a mean-field approach to de-
scribing trajectories and variances, once again for reasons of analytical tractabil-
ity. The focus on expected trajectories can be misleading, precisely because the
heterogeneity and adaptivity of agents lead to sensitive, path-dependent dynam-
ics that are not adequately captured by the mean trajectory or even by a simple
distribution over such trajectories.
Third, with ABMs it is relatively easy to embed agents in both physical and
social spaces in the same model. For example, agents can move in a two-dimen-
sional spatial topology with the specific structure of this topology designed to re-
flect the social context being modeled and bias agent interaction—who interacts
with whom and how often. This stands in marked contrast to EBMs, where
space may be built into partial differential equations, but where concerns about
tractability make it difficult explicitly to include social topologies and how they
bias interactions between agents.
Thus, ABMs make it possible to represent heterogeneous agents, each ex-
hibiting nonlinear rules of behavior and adaptive processes of various kinds,
while interacting with a variety of other agents selected as a result of spatial and
social interaction topologies. This flexibility is perhaps the fundamental reason
why ABMs are capable of replicating phenomena commonly exhibited by com-
plex social systems.
ABM vs. Game-theoretic Models
Likewise, an ABM is well suited as a methodology for studying emergent

processes such as norm formation and change relative to a game-theoretic model
(GTM). GTMs share the capacity to specify interactions in terms of individual
agents with particular sets of preferences and to evaluate their responses to dif-
ferent conditions, yielding certain nonlinear results. Given the nature of the
social problem, however, ABMs offer some distinct advantages.
First, GTMs typically assume that systems go to equilibria as limiting
states, and they do not focus on processes that unfold over time. GTMs therefore
lend themselves less readily to studying dynamic processes—such as norm forma-
tion and change—that emerge over time and are sensitive to both historical con-
tingencies and situational factors. Scholars have attempted to capture the
richness of interaction of real-world problems by increasing the number of play-
ers, permitting nonsimultaneous play, introducing the option to “exit” and
adding noise, altering the payoff matrix, and working on finite repetition (Axel-
rod and Dion 1988). While these refinements have led to a number of interest-
ing insights, they still remain equilibrium-centered. Moreover, the solutions
generally fall within the class of structure-induced equilibria, where the stability
of outcomes is determined by institutional arrangements such as the rules of ju-
risdiction and amendment control and thus by the structure of the game itself.
Second, GTMs do not lend themselves well to studying agent and envi-
ronmental heterogeneity. It is feasible to specify larger numbers of agents with
varying characteristics, but only with multiplayer games that quickly become in-
tractable. In fact, most applications of GTMs employ two or, at the most, three
types of agents, partly out of the overriding concern for being able to derive equi-
librium solutions. Meanwhile, environmental differences can only be obtained
by specifying many separate models, each with their own particular structure.
ABM, by contrast, can capture both forms of heterogeneity within the same
model, enable greater scope for variation, and implement dynamic changes—so
that these model components are also heterogeneous over time.
Third, in GTMs the characteristics of players are typically determined ex-
ogenously, and player “types” typically tend to be fixed—for example, “weak” or
“strong” states in deterrence games. It follows that it is difficult for the players in
GTMs to change their defining characteristics over time, unlike the agents in
132 RAVI BHAVNANI
ABMs, who can alter their preferences and even their traits in response to new en-
vironmental conditions and via other processes of adaptation. Thus, ABMs enjoy
an advantage in this instance to GTMs, because they address the dynamic nature of
the social problem, accommodate requisite heterogeneity of agents and environ-
ments, and build in the ability of agents to change in response to the conditions,
including group norms, they encounter and the experiences they undergo.
CONCLUSION
In a seminal article, Gould (1999) takes issue with the assumption that group
interests either implicitly lead to group action or explicitly stem from group con-
flict and result in group violence. Ethnic groups—more often than not—are con-
fronted with the problem of retaining the commitment of moderate members as
interethnic rivalry progresses from conflict to violence, and extreme members as ri-
valry progresses from conflict to cooperation. To explain how groups resolve these
dilemmas, the approach I advocate in this chapter explores the emergence of a be-
havioral norm defined in ethnic terms—a macrolevel outcome—by focusing on mi-
crolevel dynamics. ABMs, in particular, lend themselves well to the study of
emergent phenomena such as norm formation and change within a group. By re-
peating simulations and observing trajectories of participation, one can learn about
outcomes associated with different initial conditions as well as about processes of
norm formation and change, which supports the notion that where the system
ends is only as important as how it gets there, if not less important. ABMs also af-
ford an intuitively satisfying representation of real-world political situations. Most
models we construct “in our heads” involve individuals interacting with each other
and some environment. This accessibility of ABMs facilitates construction of
“what if”–type experiments that are critical for policy analysis, and yields output
that may readily be translated back into policy recommendations and practice.
NOTES
Acknowledgments: I am grateful to Robert Axelrod, David Backer, Pradeep Chhib-

ber, Neil Harrison, Ken Kollman, Scott Page, and Rick Riolo for their comments
on previous drafts. I would also like to acknowledge generous support from the
Center for the Study of Complex Systems at the University of Michigan for this
work. All faults remain my own.
1. One way to understand this phenomenon is to assume that internal
models are dominated by the survival instinct when individuals are faced with
the threat of death.
2. Bhavnani and Backer (2000) specify a computational model to explain

variation in the scale and duration of ethnic violence using data from Rwanda
and Burundi.
3. Strauss (2004) argues that the motivation to participate in the violence
is likely to have been heterogeneous, and that several theories are probably right
though he finds strong support for the argument that intra-Hutu coercion is
more likely to explain participation by “less violent” individuals, whereas fear or
anger are more likely to explain participation by “more violent” individuals.
4. Cultural models can concern things other than the ethnic norm sur-
rounding issues of violence, as certain rationales of conduct can be invoked in
support of actions other than aggressiveness or moderation. Obedience, detach-
ment, patronage, and honor all have meanings independent of issues of vio-
lence, as does arguably a jihad, rashtra, or the equivalent. Paranoia and revenge
are harder to separate. For approaches that use cultural models in explaining eth-
nic conflict, see Das 1996, Engineer 1995, and Kakar 1996.
5. Hinton (1998) illustrates how the Khmer Rouge used indoctrination to
reinforce a cultural model of detachment—“cutting off one’s feelings or heart”—to-
ward an enemy who moments earlier might have been a friend or family member.
6. In its starkest form, punishment involves the killing of coethnics who
refuse to adapt their behavior. Less-severe punishments include the destruction
of personal property; threats to the individual or his or her family; public hu-
miliation; loss of status, honor, or reputation within the group; ostracism from
the group; or bodily harm. Where punishments are subtle and executed dis-
creetly, “consciousness-raising” meetings, speeches, pronouncements, songs, slo-
gans, or chants may be used to call for a specific action or set of actions to be
taken against ethnic rivals. Such pronouncements are often couched in terms of
a communal, religious, or national duty, are accompanied by calls for ethnic sol-
idarity or unity, and invoke a “moral” obligation on the doer to perform the
stated task or assignment. Pronouncements may also carry thinly veiled threats
directed at more moderate (or extreme) members of the ethnic group, equating
transgression with sympathy for or even identification with ethnic rivals—thereby
making transgressors fair game for the very behavior they disdain. In many in-
stances, enforcement costs may be high but positive for individuals who obtain
some payoff from the enforcement of a norm. To simplify matters, one may as-
sume that the strength of the punishment captures the enforcement cost borne
by the punisher. Where enforcement costs are high, punishments are more likely
to be weak. Low enforcement costs, on the other hand, are more likely to give
rise to strong punishments.
7. Ancient history often provides important symbols and myths through
which to interpret current events. For instance, the Balkan conflict was, at least in
134 RAVI BHAVNANI
part, fueled by traditional identities of the “other” that were grounded in the fif-
teenth century. Such symbolic histories are a part of learning—and, thus, a source
of behavior—as much as other ideas and beliefs and personal daily experience.
8. Barring clear instances of ethnic norms—as in Rwanda, where pun-
ishments were widespread and in evidence, given the desired aim of mobilizing
the entire Hutu population to kill the entire Tutsi population—the fact that such
pernicious norms operate within ethnic groups exacerbates the difficulty of mea-
suring them. This is compounded by the fact that individuals are often reluctant
to divulge punishment for fear of further reprisal.
9. My discussion of ABM to study ethnic norms is not intended to be
mechanistic and or to minimize the gravity of interethnic violence. Given that
my unit of analysis is the individual, and that my primary concern is to under-
stand and model individual participation in violence against ethnic rivals, the
model is specified mainly in terms of individual characteristics, heuristics, and
behavior. My framework therefore stands in marked contrast to aggregate stud-
ies of war in international relations, where individual motivation and participa-
tion are, more often than not, filtered out of the analysis and more attention is
devoted to aggregate outcomes (conflict, deterrence, resolution).
10. Social movement theorists do regard networks as important for re-
cruiting participants for protest or rebellion. Despite their prominence in this lit-
erature, social networks have received limited attention in the context of ethnic
violence. For instance, Brass (1997) notes that all riot-prone towns do have—to a
greater or lesser degree—informal organizational networks that serve to mobilize
members. He does not, however, distinguish between different types of networks.
Likewise, Varshney (2002) bases his argument on the existence of interethnic net-
works that promote civic engagement and reduce conflict, but does not specify
the structure of these networks—how these networks may differ across contexts.
11. This also speaks to the difference between bonding (intragroup) and
bridging (intergroup) social capital.
REFERENCES
Axelrod, R. 1986. “An Evolutionary Approach to Norms,” American Political Sci-

ence Review 80, no. 4:1095–111.
Axelrod, R., and D. Dion. 1988. “The Further Evolution of Cooperation.” Sci-
ence 242:1385–90.
Bankes, S. 1994. “Exploratory Modeling for Policy Analysis.” Operations Research
41, no. 3:435–49.
Bhavnani, R., and D. Backer. 2000. “Localized Ethnic Conflict and Genocide in
Rwanda and Burundi.” Journal of Conflict Resolution 44, no. 3:283–307.
Brass, P. 1997. Theft of an Idol: Text and Context in the Representation of Collective
Violence. Princeton, NJ: Princeton University Press.
Casti, J. 1997. Would-Be Worlds: How Simulation Is Changing the Frontiers of Science.
Hoboken, NJ: John Wiley & Sons.
Das, V. 1996. Mirrors of Violence. London: Oxford University Press.
Des Forges, A. 1999. Leave None to Tell the Story: Genocide in Rwanda. New York:
Human Rights Watch.
Elster, J. 1989. “Social Norms and Economic Theory.” Journal of Economic Per-
spectives 3, no.4:99–117.
Engineer, A. 1995. Lifting the Veil: Communal Violence and Communal Harmony in
Contemporary India. Hyderabad: Sangam Books.
Gould, R. 1999. “Collective Violence and Group Solidarity: Evidence from a
Feuding Society.” American Sociological Review 64, no. 3:356–80.
Gourevitch, P. 1998. We Want to Inform You That Tomorrow We Will Be Killed with
Our Families: Stories from Rwanda. New York: Farrar, Straus and Giroux.
Granovetter, M. 1976. “Network Sampling: Some First Steps.” American Journal
of Sociology 81, no. 6:1287–1303.
Hinton, A. 1998. “‘Why Did You Kill?’ The Cambodian Genocide and the Dark
Side of Face and Honor.” Journal of Asian Studies 57, no. 1:93–122.
Kakar, S. 1996. The Colors of Violence: Cultural Identities, Religion and Conflict.
Chicago: University of Chicago Press.
Kandori, M. 1992. “Social Norms and Community Enforcement,” Review of Eco-
nomic Studies 59, no. 1:63–80.
Mamdani, M. 2001. When Victims Become Killers: Colonialism, Nativism, and the
Genocide in Rwanda. Princeton, NJ: Princeton University Press.
Marwell, G., P. Oliver, and R. Prahl. 1988. “Social Networks and Collective
Action: A Theory of Critical Mass. III.” American Journal of Sociology 94,
no. 3:502–34.
McAdam, D., and R. Paulsen. 1993. “Specifying the Ties Between Social Ties
and Activism.” American Journal of Sociology 99, no. 3: 640–67.
Oliver, P. 1984. “Rewards and Punishments as Selective Incentives: An Apex
Game.” Journal of Conflict Resolution 28, no. 1 (March): 123–48.
136 RAVI BHAVNANI
Opp, K. 1982. “The Evolutionary Emergence of Norms.” British Journal of Social

Psychology 21 (2): 139–49.
Opp, K., and C. Gern. 1993. “Dissident Groups, Personal Networks, and Spon-
taneous Cooperation: The East German Revolution of 1989.” Ameri-
can Sociological Review 58, no. 5:659–80.
Parunak, V., R. Savit, and R. Riolo. 1998. “Agent-Based Modeling vs. Equation-
Based Modeling: A Case Study and Users Guide.” Proceedings of Work-
shop on Multi-Agent systems and Agent-Based Simulation. MABS’98.
Posner, R., and E. Rasmusen. 1999. “Creating and Enforcing Norms, with Spe-
cial Reference to Sanctions.” International Review of Law and Economics
19, no. 3:369–82.
Power, S. 2001. “Bystanders to Genocide: Why the United States Let the Rwan-
dan Tragedy Happen.” Atlantic Monthly September:84–108.
Prunier, G. 1995. The Rwandan Crisis: History of a Genocide. New York: Columbia
University Press.
Sherif, M. 1936. The Psychology of Social Norms. New York: Harper.
Strauss, S. 2004. “The Order of Genocide: Race, Power, and War in Rwanda.”
Doctoral diss., Department of Political Science, University of Califor-
nia, Berkeley.
Tilly, C. 1978. From Mobilization to Revolution. Reading, MA: Addison-Wesley.
Ullmann-Margalit, E. 1977. The Emergence of Norms. Oxford: Oxford University
Press.
Varshney, A. 2002. Ethnic Conflict and Civic Life: Hindus and Muslims in India.
New Haven: Yale University Press.
Veerhagen, H. 2001. “Simulation of the Learning of Norms.” Social Science Com-
puter Review 19, no. 3:296–306.
Watts, D., and S. Strogatz. 1998. “Collective Dynamics of Small World Net-
works.” Nature 393:440–42.
CHAPTER 7
Alternative Uses of Simulation
Robert Axelrod
Let us begin with a definition of simulation. “Simulation means driving a model

of a system with suitable inputs and observing the corresponding outputs.” (Brat-
ley, Fox, and Schrage 1987, ix).
While this definition is useful, it does not suggest the diverse purposes to
which simulation can be put. These purposes include: prediction, performance,
training, entertainment, education, proof, and discovery.
Prediction. Simulation is able to take complicated inputs, process them
by taking hypothesized mechanisms into account, and then generate their
consequences as predictions. For example, if the goal is to predict interest
rates in the economy three months into the future, simulation can be the best
available technique.
Performance. Simulation can also be used to perform certain tasks. This is
typically the domain of artificial intelligence. Tasks to be performed include
medical diagnosis, speech recognition, and function optimization. To the extent
that the artificial intelligence techniques mimic the way humans deal with these
same tasks, the artificial intelligence method can be thought of as simulation of
human perception, decision-making, or social interaction. To the extent that the
artificial intelligence techniques exploit the special strengths of digital comput-
ers, simulations of task environments can also help design new techniques.
Training. Many of the earliest and most successful simulation systems were
designed to train people by providing a reasonably accurate and dynamic inter-
active representation of a given environment. An important example of the use
of simulation for training is flight simulators for pilots.
137
138 ROBERT AXELROD
Entertainment. From training, it is only a small step to entertainment.

Flight simulations on personal computers are fun. So are simulations of com-
pletely imaginary worlds.
Education. From training and entertainment it is only another small step
to the use of simulation for education. A good example is the computer game
SimCity. SimCity is an interactive simulation allowing the user to experiment with
a hypothetical city by changing many variables, such as tax rates and zoning pol-
icy. For educational purposes, a simulation need not be rich enough to suggest
a complete real or imaginary world. The main use of simulation in education is
to allow the users to learn relationships and principles for themselves.
Proof. Simulation can be used to provide an existence proof. For example,
Conway’s Game of Life (Poundstone 1985) demonstrates that extremely com-
plex behavior can result from very simple rules.
Discovery. As a scientific methodology, simulation’s value lies principally
in prediction, proof, and discovery. Using simulation for prediction can help
validate or improve the model upon which the simulation is based. Prediction
is the use that most people think of when they consider simulation as a scien-
tific technique. But the use of simulation for the discovery of new relationships
and principles is at least as important as proof or prediction. In the social sci-
ences, in particular, even highly complicated simulation models can rarely
prove completely accurate. Physicists have accurate simulations of the motion
of electrons and planets, but social scientists are not as successful in accurately
simulating the movement of workers or armies. Nevertheless, social scientists
have been quite successful in using simulation to discover important relation-
ships and principles from very simple models. Indeed, as discussed below, the
simpler the model, the easier it may be to discover and understand the subtle
effects of its hypothesized mechanisms.
Schelling’s (1974, 1978) simulation of residential tipping provides a good
example of a simple model that provides an important insight into a general
process. The model assumes that a family will move only if more than one-third
of its immediate neighbors are of a different type (e.g., race or ethnicity). The re-
sult is that very segregated neighborhoods form, even though everyone is initially
placed at random and everyone is somewhat tolerant.
To appreciate the value of simulation as a research methodology, it pays to
think of it as a new way of conducting scientific research. Simulation as a way of
doing science can be contrasted with the two standard methods of induction and
deduction. Induction is the discovery of patterns in empirical data.1 For exam-
ple, in the social sciences induction is widely used in the analysis of opinion sur-
veys and the macroeconomic data. Deduction, on the other hand, involves
specifying a set of axioms and proving consequences that can be derived from
ALTERNATIVE USES OF SIMULATION 139
those assumptions. The discovery of equilibrium results in game theory using

rational choice axioms is a good example of deduction.
Simulation is a third way of doing science. Like deduction, it starts with a
set of explicit assumptions. But unlike deduction, it does not prove theorems. In-
stead, a simulation generates data that can be analyzed inductively. Unlike typi-
cal induction, however, the simulated data comes from a rigorously specified set
of rules rather than direct measurement of the real world. While induction can
be used to find patterns in data, and deduction can be used to find conse-
quences of assumptions, simulation modeling can be used as an aid to intuition.
Simulation is a way of doing thought experiments. While the assumptions
may be simple, the consequences may not be at all obvious. The large-scale ef-
fects of locally interacting agents are called “emergent properties” of the system.
Emergent properties are often surprising, because it can be hard to anticipate the
full consequences of even simple forms of interaction.2
There are some models, however, in which emergent properties can be for-
mally deduced. Good examples include the neoclassical economic models in
which rational agents operating under powerful assumptions about the avail-
ability of information and the capability to optimize can achieve an efficient re-
allocation of resources among themselves through costless trading. But when the
agents use adaptive rather than optimizing strategies, deducing the consequences
is often impossible; simulation becomes necessary.
Throughout the social sciences today, the dominant form of modeling is
based upon the rational choice paradigm. Game theory, in particular, is typically
based upon the assumption of rational choice. In my view, the reason for the
dominance of the rational choice approach is not that scholars think it is realis-
tic. Nor is game theory used solely because it offers good advice to a decision-
maker, since its unrealistic assumptions undermine much of its value as a basis
for advice. The real advantage of the rational choice assumption is that it often
allows deduction.
The main alternative to the assumption of rational choice is some form of
adaptive behavior. The adaptation may be at the individual level through learn-
ing, or it may be at the population level through differential survival and repro-
duction of the more successful individuals. Either way, the consequences of
adaptive processes are often very hard to deduce when there are many interact-
ing agents following rules that have nonlinear effects. Thus, simulation is often
the only viable way to study populations of agents who are adaptive rather than
fully rational. While people may try to be rational, they can rarely meet the re-
quirement of information or foresight that rational models impose (Simon
1955; March 1978). One of the main advantages of simulation is that it allows
the analysis of adaptive as well as rational agents.
140 ROBERT AXELROD
An important type of simulation in the social sciences is “agent-based

modeling.” This type of simulation is characterized by the existence of many
agents who interact with each other with little or no central direction. The emer-
gent properties of an agent-based model are then the result of “bottom-up”
processes, rather than “top-down” direction.
Although agent-based modeling employs simulation, it does not necessarily
aim to provide an accurate representation of a particular empirical application. In-
stead, the goal of agent-based modeling is to enrich our understanding of funda-
mental processes that may appear in a variety of applications. This requires
adhering to the KISS principle, which stands for the army slogan “Keep it simple,
stupid.” The KISS principle is vital because of the character of the research com-
munity. Both the researcher and the audience have limited cognitive ability. When
a surprising result occurs, it is very helpful to be confident that one can understand
everything that went into the model. Simplicity is also helpful in giving other re-
searchers a realistic chance of extending one’s model in new directions. The point
is that while the topic being investigated may be complicated, the assumptions un-
derlying the agent-based model should be simple. The complexity of agent-based
modeling should be in the simulated results, not in the assumptions of the model.
As pointed out earlier, there are other uses of computer simulation in
which the faithful reproduction of a particular setting is important. A simulation
of the economy aimed at predicting interest rates three months into the future
needs to be as accurate as possible. For this purpose, the assumptions that go
into the model may need to be quite complicated. Likewise, if a simulation is
used to train the crew of a supertanker, or to develop tactics for a new fighter air-
craft, accuracy is important and simplicity of the model is not. But if the goal is
to deepen our understanding of some fundamental process, then simplicity of
the assumptions is important and realistic representation of all the details of a
particular setting is not.
NOTES
This chapter is excerpted from Robert Axelrod, “Advancing the Art of Simula-
tion in the Social Sciences,” in Simulating Social Phenomena, ed. Rosario Conte,
Rainer Hegselmann, and Pietro Terna (Berlin: Springer, 1997), 21–40, and is
used with permission.
1. Induction as a search for patterns in data should not be confused with
mathematical induction, which is a technique for proving theorems.
2. Some complexity theorists consider surprise to be part of the definition
of emergence, but this raises the question: surprising to whom?
ALTERNATIVE USES OF SIMULATION 141
REFERENCES
Bratley, P., B. Fox, and L. Schrage. 1987. A Guide to Simulation. 2nd ed. New
York: Springer-Verlag.
Cyert, R., and J. G. March. 1963. A Behavioral Theory of the Firm. Englewood
Cliffs, NJ: Prentice-Hall.
March, J. G. 1978. “Bounded Rationality, Ambiguity and the Engineering of
Choice.” Bell Journal of Economics 9:587–608.
———. 1991. “Exploration and Exploitation in Organizational Learning.” Organi-
zational Science 2:71–87.
Poundstone, W. 1985. The Recursive Universe. Chicago: Contemporary Books.
Schelling, T. 1974. “On the Ecology of Micromotives.” In The Corporate Society,
ed. Robert Morris. New York: Wiley, 19–64 (see especially 43–54).
———. 1978. Micromotives and Macrobehavior. New York: W. W. Norton. (See
especially 137–55.)
Simon, H. A. 1955. “A Behavioral Model of Rational Choice.” Quarterly Journal
of Economics 69:99–118.
CHAPTER 8
Signifying Nothing?
What Complex Systems Theory
Can and Cannot Tell Us about Global Politics
David C. Earnest
James N. Rosenau
Life’s but a walking shadow, a poor player

That struts and frets his hour on the stage
And then is heard no more. It is a tale
Told by an idiot, full of sound and fury,
Signifying nothing.
—The Tragedy of Macbeth, act 5, scene 5
So laments Shakespeare’s tragic protagonist at the news of his wife’s death.

While one may forgive the Scottish king for his pessimistic metaphor, “life” for
most of us connotes roseate meanings: dynamism, growth, learning, evolution,
and adaptation. So perhaps it is no surprise that the complexity sciences—explic-
itly concerned with these properties of a variety of systems, from physical to so-
cial—not only invoke the metaphor of life but also have postulated the idea of
“artificial” life (Langton et al. 1991; Langton 1994, 1995; Kauffman 1995; Lang-
ton and Shimohara 1997). It is unremarkable, furthermore, that social scientists
observe in human societies the dynamism, adaptability, and unpredictability of
organisms and ecological systems. Corning (2002), for one, argues that human
societies are “superorganisms” and global politics are becoming a “super-super-
organism.” Johnson’s “myth of the ant queen” (2001) explicitly postulates that
human cities organize themselves much like colonies of insects do: without the
centralized authority of an “ant queen.” Smith and Stevens (1997) reduce social
organization to the attachment behavior regulated by the brain’s cyclic produc-
tion of neurotransmitters known as “opioids.” More provocatively, Corning
notes that the biologist Edward O. Wilson argues “the humanities and social sci-
ences shrink to specialized branches of biology” (Wilson 1975, 547, quoted in
143
144 DAVID C. EARNEST AND JAMES N. ROSENAU
Corning 2000, 103). Indeed, there is a growing field of “biopolitics” (see Somit
and Peterson 1997). Rather than a walking shadow in our analyses, then, life and
its connotations of dynamism are central to an important line of contemporary
thought about social systems.
This includes the many IR theorists today who accept complexity and non-
linearity as a metaphor for the inordinate intricacy of global and international
politics. The proliferation and influence of supra- and subnational actors, sur-
prising cascading events like the Eastern European revolutions of 1989 or the
current crisis of multilateralism, the transformative effects of global information
technologies; the seemingly chronic inability of existing theories to provide reli-
able predictions—all these facts understandably make many students of politics
and societies sympathetic to theoretical approaches that posit instability, unpre-
dictability, and change in the international system. Complexity appears at first
glance to be precisely the paradigm we need to understand global politics today.
Furthermore, the simulation techniques and computer skills necessary for the
application of complex systems theories are within the grasp of international re-
lations scholars who have mastered more-conventional statistical or formal meth-
ods. Yet, by and large, international relations scholars who use complex systems
theories—not to mention complex systems theorists who study international pol-
itics—are few and far between. Clearly, international relations theory has been
slow to embrace complex systems for reasons other than the barriers to learning
its methods for investigating the intricacies of global politics. Why? Macbeth
might claim complex systems theories are tales told by an idiot, though we are
more optimistic about complexity’s prospects.
In this chapter we argue that those who study international relations have
failed to use complexity as a general theory of complex systems (“complex systems
theory”) because, while complexity is a meaningful metaphor, complex adaptive sys-
tems—at least as conventionally formulated by theorists like Holland (1992, 1995,
1998) and, in political science, Jervis (1997), Axelrod (1997), and Axelrod and
Cohen (1999)—differ in important ways from social and political systems. Al-
though they may behave in complicated and confusing ways, social systems have
structures of authority that may be inconsistent with the definition of complex
adaptive systems. These differences are more than mere definitional or typologi-
cal differences; we argue that in social systems, authority serves to minimize com-
plexity. One therefore cannot use complex systems theory to model even partly
centralized or hierarchical systems—precisely those types of systems that prolifer-
ate in the world of politics. We argue, furthermore, that by construction the sim-
ulation methods of complex systems theory cause the researcher to make
assumptions about those issues that are of most interest to international relations
scholars in particular and to political science in general: who the actors are and
SIGNIFYING NOTHING? 145
where authority resides. There is an underlying irony here. Although complex sys-
tems theory embraces contingent phenomena, it is silent on precisely those acci-
dents and path dependencies that are most important to international relations
theory. At a time in our discipline’s embrace of the contingency of social agency,
it is little surprise that few scholars are embracing theory whose methods treat as
exogenous the identities of political actors and the sources of authority.
From the perspective of international relations theory, the challenge of
complex systems theory is to model not merely dynamics but also the emergence
of actors’ identities and of political authority itself. Otherwise, IR scholars risk
modeling dynamic processes and systems that are theoretically uninteresting,
“sound and fury, signifying nothing”—what one might call “Macbeth’s objec-
tion.” Scholars who apply complex systems theory to questions of global politics
need to understand both these perils as well as the promise of its methods.
We base our criticisms on two premises. First, we take complex systems
theory at its word and assume it is indeed “theory.” For this reason we apply stan-
dards of positivist epistemology to their findings. We argue these standards are
appropriate, given the knowledge claims of Axelrod (1997) and Epstein and Ax-
tell (1996), among others, who argue their methods combine deductive and in-
ductive reasoning. To the degree that complex system theory is embedded in a
nonpositivist epistemology, as some argue, our criticisms may be inappropriate.
But to our knowledge, practitioners have engaged in little formal discussion of
either the epistemology of complex systems theory or the standards for knowl-
edge they set out for their work.
Our second premise relates to the first: to the degree complex systems the-
ory makes theory-like claims, it does so on the basis of its principal method,
known as agent-based modeling. Using these computer-based models, complex
systems theorists claim they have found nonobvious, generalizable, transmissible,
and replicable results (to the degree dynamism and indeterminacy are replicable
conditions). We argue that without the simulative methods of agent-based mod-
eling, complex systems theory has few if any methodological bases for staking its
claim as “theory” in the positivist sense. For this reason, the methodological short-
comings of agent-based modeling are, by extension, the inferential deficiencies of
complex systems theory. Without agent-based models, complex systems “theory”
reduces to a paradigm (“complexity”) rather than a theory—an indispensable ele-
ment of theory construction, to be sure, but a starting point in the process rather
than its culmination. We acknowledge that many complex systems theorists do
not share our premises. We believe, however, that the epistemology and ontol-
ogy of complex systems theory are poorly defined. We offer our criticisms not to
condemn the theory, but in the spirit of encouraging practitioners of complex sys-
tems theory to debate explicitly the foundation of their knowledge claims.
Before we elaborate these criticisms, we should note there is a galaxy of so-

cial and political issues in which authority and actors’ identities are not prob-
lematic. This is precisely the domain of theoretical inquiry in which the
simulation methods of complex systems theory hold the greatest promise. Econ-
omists long have recognized the similarity of markets and complex adaptive sys-
tems; the (hypothesized) diffuse nature of exchange allows them to simulate a
host of transactions and markets and to create interesting emergent phenomena
like poverty and the concentration of wealth. In politics, some types of informa-
tion cascades appear to emulate the informational dynamics of complex adaptive
systems. These information processes may include riots and protest movements
(Kuran 1991; Opp and Gern 1993; Lohmann 1994). Complex systems theory
also holds promise for the investigation of interest aggregation and voting be-
havior. Individuals form their political opinions from a host of information re-
sources that are as diffuse and decentralized as in a complex adaptive system. In
a range of questions about voting, markets, and information dynamics, then, au-
thority in fact is as dispersed as it is in a complex adaptive system. When as-
sumptions about agency are unproblematic, we can use simulated social systems
to investigate these questions. But for reasons we discuss below, we believe there
are inherent limits to the application of complex systems theory to a broader
range of questions about global politics.
GLOBAL POLITICS AS A COMPLEX ADAPTIVE SYSTEM
That global politics today is mystifying, intricate, and dynamic is beyond ques-
tion, and undoubtedly the reason that complexity appeals to IR scholars as a
metaphor. It is tempting to believe that these intricacies and dynamism are the
emergent properties of numerous, dispersed, and autonomous political
actors independently enacting simple local decision rules. Indeed, numerous IR
scholars, among whom we include ourselves, invoke complexity as a metaphor
(Rosenau 1990, 1997, 2003; Anderson 1996; Hughes 1997, 1999; Jervis 1997;
Earnest 2001a; Urry 2003; see also the essays in Alberts and Czerwinski 1997).
Rosenau’s turbulence model (1990, 1997) is but one example of a model that
broadly articulates global politics as a complex adaptive system. It explicitly
posits nonlinear relationships or cascades in politics; it articulates a world of nu-
merous “spheres of authority” or adaptive and decentralized political actors;
and it posits recursive relationships between political actors and their environ-
ments—what Rosenau calls “macro-micro linkages” (see also Smith 1997). Some
go even further and use complex systems theory or its antecedent, chaos theory,
as paradigms to understand patterns of conflict between states (Saperstein
1996; Axelrod 1997) and even the evolution of military organizations (Beau-
mont 1994). Clearly, students of IR and global politics have embraced the
paradigm of complexity.
Yet the paradigm of complexity holds greater sway than the theory does.
There are both methodological and, more importantly, epistemological reasons
for this. To understand why, we first discuss how complexity theorists use com-
puter models, with their attendant limitations, to model complexity in global
politics. We then explore the deeper ontological assumptions complexity theo-
rists must make about political actors and their motivations.
The Third Way
The overall behaviors of a social system, its dynamics and temporal trajectories—
what complexity researchers call its “emergent” properties—generally are what
most interest complexity theorists in the social sciences. Usually, social re-
searchers are most interested in emergent properties such as cooperation, trade,
warfare, disease transmission, and other social phenomena. By the definition of a
complex adaptive system, these emergent properties result from the local interac-
tions of numerous autonomous, independent agents pursuing local decision
rules. This massively parallel structure of the complex adaptive model usually begs
for the use of computing technologies that allow either parallel processing (that is,
computers with a microprocessor for each actor—a serious practical limitation) or
quasi parallel processing (software that iteratively processes instructions for each
actor on a single microprocessor before taking the next “step” in time).1 While it
is in principle possible for the researcher to simulate these systems manually—a
memorable example is Thomas Schelling’s (1978) use of pennies, dimes, and a
ruled piece of paper to study residential segregation—the volume of calculations
the researcher needs to undertake may be prohibitive. Complexity theorists thus
often rely on computers to conduct their social simulations.
Such computer-based models face a number of barriers to their acceptance
among IR theorists. Though our discipline’s cultural aversion to new methods
may be one of them, we choose to focus instead on the broader question of how
a researcher may use these computer simulation methods within a broader re-
search program. Two problems are immediately apparent. Computer simula-
tions of complex adaptive systems are, first of all, neither a deductive nor an
inductive method. Because simulations of complex adaptive systems typically do
not rely upon empirical data—though it seems to us that this need not be so—they
may be of little help in inducing patterns in the actual politics of the world
around us. Although complexity simulations seek in principle to discover pat-
terns, they do so through one of two different ways. First, researchers may ob-
serve an empirical phenomenon that they hypothesize is the emergent property
of a complex adaptive system. The researchers then make assumptions about ac-
tors, their decision-rules, and feedback from the environment and create and
run on the computer a model of the system to see if they can “grow” the phe-
nomenon or property (Epstein and Axtell 1996, 20). Epstein and Axtell’s “sug-
arscape” model (1996) typifies this approach to using a simulation for inductive
purposes. If successful, however, such a simulation relies upon the researcher’s
assumptions about actors, rules, and the environment. As we argue below, these
assumptions can often be highly problematic and potentially tautologous.
An alternative approach to induction is to base the model’s assumptions on
empirical data and then see what interesting emergent properties, if any, grow out
of the simulated system. Though it was not a simulation of a complex system per
se, Meadows et al.’s “World3” model of population growth and resource exhaus-
tion (1974) is a good example of this inductive use of empirical data as a basis for
an algorithm-based computer simulation. This approach still relies on the mod-
eler’s assumption of which variables are salient. As Miller (1998) shows in his test-
ing of the World3 model, assumptions about relevant causes can drive a model in
hidden, unexpected ways, giving rise not only to questions of model validity but
to inaccurate predictions as well. Given our apparent consensus that global poli-
tics are intricate, such an approach risks oversimplification, and in any case may
not yield any theoretically interesting emergent properties. When complexity the-
orists use inductive methods to inform their models, therefore, they face two crit-
icisms. Either their assumptions are empirically groundless and theoretically
underdetermined, or their simulations produce uninteresting dynamics that at
best have no referent in real world politics and at worst are indecipherable.2
Of course, deductive theorists have long argued that as a matter of episte-
mology, one should not reject a model or theorem on the basis of its axioms.
Rather, we should look at its explanatory and predictive value. A deductive theorist,
therefore, would have no objections to the problematic assumptions we think com-
plexity theorists make. Yet complex systems theory is not a deductive theory, for two
reasons. First, deductive methods seek explicitly to prove consequences that one
may logically derive from axioms. Complex systems theory by construction posits,
however, that it is difficult for the researcher to deduce consequences from his or
her initial assumptions. The nature of contingency in complex adaptive systems
means that numerous consequences are possible under a given set of assumptions,
and that the same axioms are likely to produce different, even divergent, outcomes.
To put it another way: the essence of path dependence is that while the researcher
may be able to deduce a set of possible outcomes, he or she cannot deduce “the” (or
even “the likely”) outcome(s) of a process. Second, the simulation methods of com-
plex systems theory are not deductive, because they do not prove theorems. Unlike
game theory or other deductive methods, complexity researchers cannot explicitly
test hypotheses about global politics. Complex systems theory, therefore, lacks both
the empirical appeal of induction and the disconfirmative value of deduction.
These observations about the complexity sciences are not novel, of course.
Practitioners of complex systems theory have defended their work against these
charges. Axelrod (1997) acknowledges that the simulation of complex adaptive
systems is neither inductive nor deductive. Rather, he calls this method a “third
way” of doing science, in which “simulated data . . . can be analyzed inductively”
(Axelrod 1997, 4). Epstein and Axtell (1996) similarly call the complexity para-
digm a “generative” science: “Artificial society modeling allows us to ‘grow’ social
structures in silico demonstrating that certain sets of microspecifications are suf-
ficient to generate the macrophenomena of interest” (20, emphases in original).
In this respect, both Axelrod and Epstein and Axtell argue complexity simula-
tions combine elements of inductive and deductive methods as an aid to the
researcher’s intuition.
Despite these advantages, the methods of complex systems theory alone
cannot prove or disprove hypotheses about global politics—although it is a theory
of process, it cannot be a theory of politics. To make sense of the intricacies of
contemporary global politics, then, researchers who use simulations of complex
adaptive systems must supplement these efforts with empirical investigations. Of
course, this requires scholars to embrace once again those methods—such as case
studies or statistical models like ordinary least squares—which we have derided
for their emphasis on stasis and linearity. Yet until a computer simulation can
disprove a hypothesis, complex adaptive systems are little more than thought
experiments on a computer—much ado about nothing.3
It is unclear, furthermore, that empirical tests of computer-simulated
processes can in fact test our hypotheses about actual dynamic systems. Although
Elliott and Kiel (1997) advocate such a complementary approach, this conjoint
simulative-empirical research design may hold greater promise for physical and bi-
ological systems than for social systems because of several distinctive features of
humans. For one, while humans may follow simple decision rules, they also may
not. Humans are both adaptive and habitual, capable of both learning and mis-
apprehension, and paradoxically irrational yet calculating in their interactions
with other human beings. We recognize that complex systems theorists argue they
can simulate learning and strategic behavior in their computer-based worlds; for
reasons we enumerate below, however, we remain yet to be convinced. A second,
and perhaps more important, feature of societies is the role of authority. Unlike
in physical complex adaptive systems, authority in human societies—and even
among social animals like apes or wolf packs—may be logically incompatible with
the definition of a complex adaptive system. A social system may have authority
present and be complex; it may be complex and adaptive; but it cannot have both
authority and adaptation through complex interactions of autonomous agents,

since centralized authority and individual autonomy, by definition, are mutually
exclusive. As we argue below, the conceptual conundrum of authority in complex
systems presents a fundamental challenge to the application of complex systems
theory to questions of world politics.
The Role of Expectations
The three witches prophesied to Macbeth his murderous actions. Human agents
typically rely, by contrast, on their own, less Delphic knowledge to inform their
behavior. A substantial literature suggests that human beings are not perfectly ra-
tional; we rely upon heuristics and other simplifications in our everyday decision-
making. The literature on the psychology of decision-making also suggests that
humans indulge in wishful thinking and other forms of motivated bias (see also
Levy 1997). In short, our decisions depend, at least in part, on who we are. At first
glance, these lines of thought seem to support a rejection of the assumption of
rationality and the adoption of “satisficing”-type decision rules that typify agents
in complex adaptive systems. We suspect, however, that this simplicity itself may
gloss over some important characteristics of expectations and interests in human
actors. The appeal of the rationality assumption is its simplicity, but the psychol-
ogy of decision-making suggests that human beings make decisions in messy ways
that are difficult to capture with simple assumptions about decision rules. While
complexity theorists can reasonably assume that biological or physical agents pur-
sue these simple decision rules, it is unclear that this assumption holds for human
agents, because humans’ expectations are contingent in part on their identities. If
so, then a simulated complex adaptive social system will make assumptions about
expectations and interests that are unlikely to capture this dependence.
Again, it is useful to understand the methods complexity researchers use to
simulate the adaptive and learning behavior of agents. Typically, complexity re-
searchers capture the adaptive behavior of agents through genetic algorithms (see
Holland 1995, 1998). These are software routines that each agent in the simula-
tion follows to learn, evolve, or adapt to his or her environment. Agents themselves
may adapt by changing their attributes or passing advantageous characteristics to
future generations, the agents’ decision rules may adapt through learning, or both.
Since students of global politics typically are interested in actors’ expectations and
interests rather than in their biological characteristics, we focus here on modeling
the evolution of decision rules of agents rather than their attributes. The modeler
typically assigns to the population of agents either randomly generated or theoret-
ically informed decision rules. Through iterative feedback from the simulated so-
cial environment, the agents adjust their decision rules through mutation, through
selection, or through reproduction. Fitter agents come to dominate the system,

while less fit agents may be “selected out” or become in some sense “peripheral”—
though this term belies the importance of these “lesser” agents in sustaining the
system’s dynamics. Through these processes, agents typically narrow the popula-
tion of decision rules; from a randomly generated set of rules, for example, the
researcher may find that the genetic algorithm produces a small population of “fit”
strategies.4 For example, Axelrod (1997, chap. 1) has shown how tit-for-tat emerges
as an optimal strategy in a complexity simulation of an iterated multiplayer
prisoner’s dilemma.
An important question is, then, whether or not genetic algorithms can em-
ulate the processes of how human beings make decisions. There are two reasons to
suspect that they cannot, both of which arise from how a political actor’s identity
can influence his or her expectations. First, the literature on interest groups tells us
the spatial organization of interests matter: a political actor’s identity as a passive or
active participant, for example, depends in part on how the interests of others are
spread throughout a system. Building on Olson’s seminal work (1965) on collec-
tive action problems, James Q. Wilson (1980) argues that the concentration or dif-
fusion of both the costs and of benefits from public policies will affect the type of
contestation in which individuals engage. When both costs and benefits are con-
centrated, interest group politics will result; when both are diffuse, majoritarian
politics will result; diffuse costs and concentrated benefits yield client politics; and
concentrated costs but diffuse benefits create entrepreneurial politics. In this re-
spect, an agent’s interests and, in turn, expectations will be shaped by the spatial or-
ganization of interests in the system as a whole. While complex systems theory in
principle embraces precisely this type of contingency in decision rules, the problem
arises with the a priori assumption of the spatial organization of costs and benefits.
A randomly generated population of interests, expectations, or decision rules
merely assumes that interests are diffuse and that politics, therefore, are either ma-
joritarian or entrepreneurial. Likewise, the modeler’s choice of fitness criterion
makes an assumption about the concentration or diffusion of costs and benefits
from the environment. If the modeler assigns specific interests or decision rules,
however, he or she places the model in either the world of client politics or inter-
est group politics. Wilson’s insight for complex systems theory is that the modeler’s
assumptions about interests—no matter what those assumptions may be—are not
theoretically neutral. The inferential problem is, therefore, determining to what de-
gree the emergent behavior of the social system results from the dynamics of the
system, and to what degree from the modeler’s assumptions. If multiple and di-
vergent outcomes may result from the same set of assumptions, as complex system
theory posits, it becomes exceedingly difficult to disentangle those results which
emerge endogenously from those that are true by construction. The modeler’s
assumptions, as much as the genetic algorithm, may end up shaping the path of
the evolution of agents’ decision rules. The danger of a tautology is obvious.
A second problem with genetic algorithms is a classic problem of most
models of human agency: they share an inability to capture the nuanced psy-
chology of human decision-makers. Humans do not follow simple decision rules;
we are subject to a panoply of psychological biases and errors. Many of these bi-
ases depend, furthermore, on who we are. As the literature on motivated bias
suggests (see Lodge, Tabor, and Galonsky 1999), our self-identification can affect
our perceptions and expectations: we interpret information as supporting our
desired outcome, and we incorporate discrepant or disconfirming information
as supporting our predispositions or earlier decisions. It is debatable, therefore,
whether human agents can accurately perceive environmental feedback in the
perfect way that the use of genetic algorithms suggests. If one accepts the con-
structivist critique that identities themselves are contingent (a notable “if” that
we address later), then interests and expectations are highly path dependent in
ways that genetic algorithms fail to capture. Indeed, Urry (2003) articulates an
extreme variant of this argument:
[T]here is no “structure” and no “agency”, no “macro” and no “micro” lev-

els, no “societies” and no “individuals”, and no “system world” and no
“life world”. This is because each such notion presumes that there are en-
tities with separate and distinct essences that are then brought into exter-
nal juxtaposition with its other. (122)
As Urry suggests, the very contingency of one’s political identity—whether or not

one has a stake in a given policy, whether or not one chooses to partake in po-
litical contestation—makes interests and in turn expectations accidental and path
dependent. In a simulated social system, the random assignment of decision
rules—indeed, any rule for the assignment of decision rules, expectations, inter-
ests, or identities—requires the researcher to make unfounded assumptions
about the structure of politics and its influence on decisions. Human psychology
and decision-making are so idiosyncratic as to make the ascription of simple
behavior rules highly problematic.
The Emergence of Authority
Politics is, our textbooks tell us, the authoritative allocation of values. The exer-
cise of authority is central to our understanding of global politics today, particu-
larly since so many researchers argue that authority has migrated away from the
institution of the nation-state. It has migrated upward to international and non-
governmental institutions and to global corporations; it has migrated downward
to local governments, civil society, terrorist cells, and others. Authority therefore
is problematic; most of us no longer assume the primacy of the nation-state. This
feature of global politics poses a double challenge to the application of complex
systems theory to international politics. First, we need a complex adaptive system
that shows how authority shifts from one authoritative actor to another and how
“layers” (or “spheres,” “nodes,” or “attractors”) of authority may result. Second,
and formidably more important, we need a complex adaptive system that shows
how authority emerges in the first place from the interactions among autono-
mous agents.
As we have already noted, the pattern of authority in a complex adaptive
system is one of its distinctive features: it has none. Authority is perfectly decen-
tralized; each agent decides and acts on the basis of internal rules that evolve in
response to environmental feedback. This is the logical antithesis of social au-
thority, in which a privileged agent makes allocative decisions for a group of
other actors. Unlike in a complex adaptive system, political authority often com-
pels individuals to act contrary to their internal rules: the beggar, no matter how
needy, will go to jail for stealing a loaf of bread. This raises the question of the
appropriateness of the complexity metaphor for the study of politics: are au-
thoritative systems logically incompatible with complex adaptation? Though they
did not consider the question in these terms, classical thinkers clearly thought
so. A complex adaptive social system, one that derives its dynamism and adapt-
ability from its precarious balance on the edge of chaos, is nasty, brutish, and
short according to Hobbes. The “natural” response to such systems, according to
classical thinkers and organizational theorists alike, is centralized authority: the
state, the firm, the hegemon, or the “leviathan.” Of course, the role of authority
in a complex system may be one of degree: some complex systems are character-
ized by little if any authority, while in others authority may represent the bound-
aries of the system or the “rules” within which autonomous agents enact their
rule models—much like economic agents pursue rule models (“get rich”) within
a system (“market”) in which property rights and the enforcement of contracts
are unproblematic. But if we take the definition of a complex system at its word,
in which decision-making is perfectly diffuse (or to extend our example, a market
in which agents can break contracts or steal from others), how can a social system
with its attendant structures of authority be a complex system? And if a complex
system can have “some” hierarchy, or alternatively some balance of centralized
and decentralized decision-making, what is the difference between a complex sys-
tem and other definitions or types of systems? What theoretical leverage does the
ontology of the complex adaptive system offer, particularly if its treatment of au-
thority is so elastic? The definition of a complex adaptive system thus seems in-
consistent with our conventional understanding of what an authoritative system
is. While this may offer certain conceptual advantages, complexity theorists have
yet to reconcile an authoritative social system with the decentralized authority of
a complex adaptive system.
This question is particularly troublesome, furthermore, for the application
of complex systems theory to problems of global politics. While the question of
cooperation under anarchy once preoccupied IR scholars, these days the disci-
pline of IR has a much more nuanced conception of authority in the interna-
tional system. Much of the study of global politics today is concerned with the
problematic relationships between sovereign states and supra- and subnational
authorities. Some of these supra- and subnational actors may not be forms of au-
thority as we have strictly defined it; they may not rely upon coercion, for exam-
ple, to maintain their influence over political actors. Both the literature on and
the popular perception of the “democratic deficit” suggest, however, that the co-
ercive power of supranational actors is not trivial. Such supranational actors may
in fact possess coercive powers typical of strictly defined authority. Global poli-
tics today is rife with examples of new forms of social authority, both of the per-
suasive and the coercive varieties. How can these patterns of authority result
from the complex interactions of autonomous actors? While it is tempting to
argue we can model these global and transnational processes as interactions
among states, for example, this only begs the question of why states—and not vot-
ers, NGOs, IGOs, or transnational elites—deserve ontological primacy. If the an-
swer is simply that it is easier to model authority as a unitary actor, then we have
merely committed the same methodological error as other methods: we choose
our models not because they are conceptually appropriate and theoretically use-
ful, but because they are easier to construct, implement, and understand.
The breakdown of authority is another of the central concerns of the study
of global politics that demonstrates the shortcomings of the concept of the com-
plex adaptive system. This process of breakdown arguably is itself the result of
the complex interactions of dispersed autonomous agents. For example Kuran
(1991), Opp and Gern (1993), and Lohmann (1994) each explain the Eastern
European revolutions of 1989 as the result of cascading information processes
among leaderless individuals. Riots and other forms of the erosion of authority
may be important phase transitions in complex adaptive social systems. Persua-
sive authority itself may derive, furthermore, from informal, decentralized social
structures rather than classic, Hobbesian centralized authority (Earnest 2001b):
“leaderless” groups may derive moral legitimacy precisely because citizens view
them as decentralized and spontaneous. Marion (1999) calls these informal
forms of cooperation “social solitons”; Rosenau (1990, 1997) calls them “spheres
of authority”; and Harrison (2001) terms them “nodes of order.” Indeed, one
area where complex systems theory holds considerable promise is the investiga-
tion of how authority cascades through political systems; how it shifts from one
authority “attractor” to another; and if, when, and where it may achieve a degree
of dynamic equipoise.
But these processes and questions require us to make some initial assump-
tions about authority in a complex adaptive system. For some questions of global
politics—such as the study of market transactions, information dynamics, and vot-
ing behavior—researchers can make some reasonable assumptions. But for other
questions these assumptions risk making any findings about authority highly
problematic. Before we can model shifts in authority, a complex adaptive social
system must have political authority in the first place. One possible way to do this
is to endogenize authority at the genesis of a complex adaptive system. We can
assume that a single agent makes allocative decisions for all other agents and can
enforce those decisions in the face of agents’ internal rule models. Setting aside
the question of whether or not this is truly a complex adaptive social system, this
approach cannot tell us about the sources of authority. If the emergence of au-
thority is, furthermore, path dependent (and there are good reasons to suspect it
is), this exogenous approach may assume away important evolutionary dynamics.
The alternative method—and in our minds more challenging but theoreti-
cally more fruitful—is to grow authority from the bottom up, as the emergent
property of a complex adaptive system. A number of complexity theorists already
have tackled this challenge. Axelrod’s (1997) tribute model shows that aggregate
collectivities may emerge from the behavior of decentralized agents in a complex
adaptive system. Likewise, Cederman (1997) seeks to endogenize the processes of
the constitution of states in the international system. Epstein and Axtell (1996)
show how markets may emerge and how actors will assume specialized roles as
creditors or debtors. Kollman, Miller, and Page (1997) simulate “instability” or
variations in the effectiveness of political institutions. While these approaches
are an important first step, they may not truly simulate the emergence of the au-
thoritative allocation of values for a population of agents. Axelrod’s tribute
model purports to demonstrate how collective action arises through coercion,
how power creates its own authority. Yet, the tribute model produces only quan-
titatively different actors; some states in the model develop more power than
others. But the emergent actors are not qualitatively different: they do not make
decisions for other agents in the system. Epstein and Axtell’s sugarscape model
is similarly devoid of agents that are qualitatively different. Though the economy
of the sugarscape creates debtors and creditors, both debtors and creditors are
functionally identical. Each follows local decision rules.
One might argue that creditors on the sugarscape, or strong states in the
tribute model, in fact are qualitatively different actors or authorities, since their
power or wealth deprives weaker agents of viable choices. Though these weaker
agents technically are not compelled in their enactment of their individual deci-
sion rules, the argument might go, the power of stronger actors presents them
with Hobson’s choices, and hence no choice at all. After all, any one of us may
choose not to pay our taxes and to face the consequences from the power of the
authority known as the IRS. Yet, this analogy shows how inappropriately the trib-
ute model or the sugarscape model treat the concept of an authoritative actor. Au-
thority is not merely the force to make allocative decisions for a collectivity. It is
also the force to define and shape the collectivity as well. After all, if we do not pay
our taxes, we go to jail. Social authority therefore includes the power to remove
any given actor from society itself; it therefore not only allocates values but affects
the aggregation of interests. This is where complex adaptive approaches to au-
thority and the emergence of new political actors have fallen short. Because deci-
sion-making is by definition perfectly diffuse in a complex adaptive system, a
complex adaptive system can neither simulate social authority nor describe how
social authority constitutes the very actors that are its subjects.
This argument anticipates, furthermore, how complex systems theory’s
approaches to authority fail to engage the constructivist critique of structuralism
in international relations theory. As Wendt (1987, 1994) and others have
argued, it is theoretically groundless for scholars to assume that any agents are
ontologically primitive in global politics. Spruyt (1994) shows that even the na-
tion-state is a contingent social construction. Ruggie (1986, 1993) similarly ar-
gues that differentiation among nation-states is a historically path-dependent
constitutive process. Whether they be states, voters, or organizations, political
agents themselves therefore are contingent and indeterminate. In politics,
whether domestic, international, or global, the rules of identity also may be issue-
specific—the political issue may determine who the “actors” are. In this respect,
the constitutive rules of identity—of who participates in the contestation of spe-
cific values—themselves may vary and adapt in a way that is not true of a biologi-
cal or physical complex adaptive system. As we noted above, social authority
itself can constitute political actors by adding them to society (through enfran-
chisement, liberation, or a host of other processes) or removing them (conquest,
imprisonment, and so on). Societies are not merely open systems, with political
agents entering and exiting. They are, rather, “protosystems” that include an in-
finite number of latent actors and dormant systems. These actors and systems
emerge, adapt, erode, shift, and dissolve with extraordinary speed. In politics,
therefore, there are no “agents” per se; rather, there are latent identities, attrib-
utes, or values that are context and spatiotemporally dependent, which other
actors may invoke to mobilize or remove political actors.
Lustick (2000) and Cederman (1997, particularly chap. 8) each have at-
tempted to endogenize this type of latency in political agents. Both approach the
question of agency by endowing the agents in their complex adaptive social sys-
tems with latent identity attributes. As their models run, contexts may emerge in
which the actors’ latent identities are activated. These are promising approaches
to the issue of the contingency of agency, but they are less developed on the ques-
tion of the contingency of authority. It remains to be seen whether complexity
theorists can endogenize fully these constitutive processes. In contrast to Lustick
and Cederman, most complexity simulations of politics treat these questions ax-
iomatically. By assigning rules of identity exogenously—that is, by assuming the ex-
istence of authority rather than endogenizing it as a problem—complex adaptive
approaches to world politics overlook those very concepts that are of greatest in-
terest to us as students of global politics: institutions and identities. Complexity
theorists make not just procedural assumptions, but also an important explana-
tory assumption: the very existence of political actors.5 A complex adaptive simu-
lation of global politics thus risks a cleverly disguised tautology: the emergence of
authority and agency results not from the adaptive, dynamic, nonlinear interac-
tions of the agents, but from the researcher’s own assumptions about authority
and agency at the model’s genesis. The simulative methods of complexity thus risk
obfuscating important assumptions made by complexity researchers about actors
and authority in global politics today. It is little wonder, therefore, that students
of global politics remain skeptical about complex systems theory’s methods.
THE SOUND AND THE FURY, SIGNIFYING NOTHING?
“Can machines think?” the mathematician Alan Turing (1950) asked more than
a half century ago, and he set out to create a measure by which artificial intelli-
gence researchers could assess their progress in creating cognizance. The result-
ing standard—the Turing test—specifies that scientists will have succeeded in
creating artificial intelligence when a human interrogator cannot distinguish be-
tween a human respondent and a computer respondent. One way to respond to
our criticisms of complex systems theory is, we believe, to ask a similar question
about “artificial authority”: can a machine command compliance? Toward this
end, the works of Axelrod (1997), Cederman (1997), Lustick (2000), and others
hold some promise in “growing” authority as the emergent property of a com-
plex adaptive system. Soon, complex systems theorists and students of global pol-
itics may face a question similar to Turing’s: how will we recognize authority in
a simulated complex adaptive system?
It is beyond the scope of this chapter to answer this question fully, but a few
guidelines may help develop a research program in artificial authority. Following
the example of the Turing test, we propose an experimental design that incorpo-
rates human subjects into the virtual world of the simulated social system. If and
when a complexity researcher succeeds in generating authority from the bottom up

in a complex adaptive social system, the researcher should rewrite the program so
that human subjects can participate as if they were agents within the system.
Human subjects receive and transmit the same information from and to their en-
vironment as the silicon subjects do, receive the same rewards or punishments, and
influence their neighbors and environment according to the same rules, but they
are free to follow their own decision rules. Under such conditions, when human
subjects comply with the system’s authority in the same manner as the silicon
agents do, then one may reasonably argue the system is authoritative.6 Such a pro-
gram may not only begin to address our criticisms but may, in fact, contribute to
our theoretical understanding about contingencies of agency and authority.
We believe this is a difficult, though not unattainable, standard, because
the world of politics is conceptually different than the world of physical and bio-
logical complex adaptive systems. In politics, actors—whether states, IGOs,
NGOs, terrorist cells, or voters—have expectations that are more than the simple
rule models posited for complex adaptive agents. Their decisions and expecta-
tions depend in turn on who they are and their position not only in political
space, but also in time and the context of values. We cannot even assume, fur-
thermore, that the agents are ontologically primitive. Of course, these contingen-
cies are—conceptually, at least—precisely the type of spatiotemporal path
dependencies that complexity researchers are interested in understanding. The
problem to us is not with the paradigm but with the epistemology of complexity.
The practical necessity of relying upon computers to simulate complex adaptive
social systems risks not merely oversimplification, but also the exogenous treat-
ment of interests and identities that are of most theoretical interest to students
of international politics today. The methods of complexity do not easily transcend
the disciplinary boundary between the social and biological or physical worlds. It
behooves us as students of international politics to demonstrate how complex sys-
tems theory can answer the difficult questions of authority and agency. We hope
such a research program on artificial authority will begin to address the ontologi-
cal conundrum of authority in a complex system and to articulate the degree to
which complex systems theory depends upon a nonpositivist epistemology.
The concept of a complex adaptive system appeals to us precisely because
it embraces the intricacies, nonlinearities, and unpredictability that we observe
daily in world politics. Complex systems theory offers a paradigm that explicitly
rejects the concepts of equilibrium and stasis that seem so inappropriate for our
understanding of international politics. It has, in short, considerable conceptual
appeal. But in our view, the complex adaptive simulations of world politics that
we know of are promising yet incomplete—as Macbeth might say, they are full of
sound and fury, but as yet signify very little. Until the methods of complex sys-
tems theory can create artificial authority from the ground up, therefore, we fear
its hour on our conceptual stage is drawing to a close.
NOTES
1. The best-known simulation software is, of course, the Santa Fe Institute’s

Swarm software, available for a number of different operating systems (http://www.
swarm.org). Other software packages have lower learning thresholds, particularly
NetLogo (developed at Northwestern University, http://ccl.northwestern.edu/net
logo/) and StarLogo (developed at MIT, http://education.mit.edu/starlogo/).
RePast (University of Chicago, http://repast.sourceforge.net/) and Ascape (the
Brookings Institution, http://www.brook.edu/dybdocroot/es/dynamics/
models/ascape/ReadMe.html) are two other simulation software packages. All are
freely available with numerous sample simulations.
2. For example, Taber (2001) recalls a conference at which two modelers
of fractals confessed they had “no earthly idea” how their model worked. See
Taber 2001, 24.
3. According to methodological falsificationism, the utility of a method
lies not only in its ability to disconfirm a hypothesis, but also in its failure to dis-
confirm hypotheses once it has demonstrated the ability to do so. This is the
essence of logical positivism: while we cannot prove hypotheses, we can derive
some confidence in our knowledge when we fail to disprove them. In this sense,
complex systems theory faces two challenges. First, it must establish its ability to
test and disconfirm hypotheses. For example, to establish its bona fides we may
use it to disconfirm hypotheses that we have already falsified with other meth-
ods. Second, only after it has established its ability to disconfirm can complexity
theory be used to probe and test new hypotheses. This second challenge is ar-
guably more difficult to achieve. Otherwise, we cannot know whether or not the
failure to disconfirm a hypothesis arises from the robustness of the hypothesis or
from the methodological shortcomings of complex systems theory.
4. Miller (1998) argues that genetic algorithms may themselves be useful
in probing the weaknesses of the specification of complexity models and identi-
fying the modeler’s key assumptions.
5. See Hoffman and Johnson 1997. Indeed, to many constructivists the
assumption that a political system exists is not procedural, the assumption that
actors have interests is not procedural, and the assumption that these interests
are related to—or unrelated to—the actors’ position in space and time is not pro-
cedural. All of these “assumptions” that agent-based modeling requires are pre-
cisely those questions of greatest theoretical interest to an important school of
thought about global politics.
6. It is admittedly difficult to operationalize measures that test whether

or not human subjects comply with a model’s authority in the same manner
as silicon subjects. Since complex systems theory posits that agents’ behaviors
are correlated with each other, we cannot assume the independence of the be-
havior of silicon agents when human agents participate in the system. In
other words, we cannot assume that silicon agents would necessarily have
complied with authority in the absence of human agents, or alternatively that
they would not have.
REFFERENCES
Alberts, David S., and Thomas J. Czerwinski, eds. 1997. Complexity, Global
Politics, and National Security. Washington, DC: National Defense
University.
Anderson, Peter J. 1996. Global Politics of Power, Justice and Death: An Introduction
to International Relations. London: Routledge.
Axelrod, Robert. 1997. The Complexity of Cooperation: Agent-Based Models of Com-
petition and Collaboration. Princeton, NJ: Princeton University Press.
Axelrod, Robert, and Michael D. Cohen. 1999. Harnessing Complexity: Organiza-
tional Implications of a Scientific Frontier. New York: Free Press.
Beaumont, Roger. 1994. War, Chaos, and History. Westport, CT: Praeger.
Cederman, Lars-Erik. 1997. Emergent Actors in World Politics: How States and Na-
tions Develop and Dissolve. Princeton, NJ: Princeton University Press.
Corning, Peter A. 2000. “The Sociobiology of Democracy: Is Authoritarianism
in Our Genes?” Politics and Life Sciences 19 (March): 103–8.
———. 2002. “Synergy and the Evolution of ‘Superorganism,’ Past, Present, and
Future.” Paper presented at the annual meeting of the Association for
Politics and the Life Sciences. Montreal, August 11–14.
Earnest, David C. 2001a. “The Complexity of Globalization: Meeting the Chal-
lenges of Structuration Theory.” Paper presented at the Forty-second
Annual Convention of the International Studies Association. Chicago,
February 20–24.
———. 2001b. “Will No One Rid Me of This Meddlesome State? Social Inequal-
ity and the New Social Contract.” Paper presented at the Forty-second
Annual Convention of the International Studies Association. Chicago,
February 20–24.
Elliott, Euel, and L. Douglas Kiel. 1997. “Nonlinear Dynamics, Complexity, and
Public Policy.” In Chaos, Complexity and Sociology: Myths, Models, and The-
ories, ed. Raymond A. Eve, Sara Horsfall, and Mary E. Lee, 64–78.
Thousand Oaks, CA: Sage.
Epstein, Joshua M., and Robert Axtell. 1996. Growing Artificial Societies: Social Sci-
ence from the Ground Up. Washington, DC: Brookings Institution.
Harrison, Neil E. 2001. “The Value of a Complexity Metaphor for International
Political Economy,” Prepared for the 42nd Annual Convention of the
International Studies Association in Chicago, February 20–24.
Hoffman, Matthew J., and David Johnson. 1997. “Change and Process in a
Complex World: Using Complexity Theory to Understand World
Politics.” Paper presented at the International Studies Association-
Midwest meeting. Cleveland, October 3–5.
Holland, John H. 1992. “Complex Adaptive Systems.” Daedalus 121, no. 1
(Winter): 17–30.
———. 1995. Hidden Order: How Adaptation Builds Complexity. Reading, MA:
Addison-Wesley.
———. 1998. Emergence: From Chaos to Order. Cambridge, MA: Perseus.
Hughes, Barry. 1997. Continuity and Change in World Politics: Competing Perspec-
tives. Upper Saddle River, NJ: Prentice-Hall.
———. 1999. International Futures: Choices in the Face of Uncertainty. Boulder, CO:
Westview.
Jervis, Robert. 1997. System Effects: Complexity in Political and Social Life. Prince-
Johnson, Steven. 2001. Emergence: The Connected Lives of Ants, Brains, Cities and
Software. New York: Scribner.
Kauffman, Stuart. 1995. At Home in the Universe: The Search for the Laws of Self-
Organization and Complexity. New York: Oxford University Press.
Kollman, Ken, John H. Miller, and Scott E. Page. 1997. “Political Institutions
and Sorting in a Tiebout Model.” American Economic Review 87, no. 5
(December): 977–92.
Kuran, Timur. 1991. “Now Out of Never: The Element of Surprise in the East
European Revolutions of 1989.” World Politics 44 (October): 7–48.
Langton, Christopher G., ed. 1994. Artificial Life III: Proceedings of the Workshop
on Artificial Life, held June 1992 in Santa Fe, New Mexico. Reading, MA:
Addison-Wesley.
———. 1995. Artificial Life: An Overview. Cambridge, MA: MIT Press.

Langton, Christopher G., and Katsunori Shimohara, eds. 1997. Artificial Life V:
Proceedings of the Fifth International Workshop on the Synthesis and Simula-
tion of Living Systems. Cambridge, MA: MIT Press.
Langton, Christopher G., et al., eds. 1991. Artificial Life II: Proceedings of the Work-
shop on Artificial Life Held February, 1990 in Santa Fe, New Mexico. Red-
wood City, CA: Addison-Wesley.
Levy, Jack S. 1997. “Prospect Theory, Rational Choice, and International Rela-
tions.” International Studies Quarterly 41, no. 1 (March): 87–112.
Lodge, Milton, Charles Taber, and Aron Chase Galonsky. 1999. “The Political
Consequences of Motivated Reasoning: Partisan Bias in Information
Processing.” Paper presented at the Annual Meeting of the American
Political Science Association. Atlanta, September 2–5.
Lohmann, Susanne. 1994. “The Dynamics of Information Cascades: The Mon-
day Demonstrations in Leipzig, East Germany, 1989–1991.” World Pol-
itics 47, no. 1 (October): 42–101.
Lustick, Ian. 2000. “Agent-Based Modeling of Collective Identity: Testing Con-
structivist Theory.” Journal of Artificial Societies and Social Simulation 3,
no. 1 (January). Available at http://www.soc.surrey.ac.uk/JASSS/3/1/
1.html.
Marion, Russ. 1999. The Edge of Organization: Chaos and Complexity Theories of For-
mal Social Systems. Thousand Oaks, CA: Sage.
Meadows, Dennis L., et al. 1974. Dynamics of Growth in a Finite World. Cam-
bridge, MA: Wright-Allen Press.
Miller, John H. 1998. “Active Nonlinear Tests (ANTs) of Complex Simulation
Models.” Management Science 44, no. 6 (June): 820–30.
Olson, Mancur. 1965. The Logic of Collective Action: Public Goods and the Theories
of Groups. Cambridge, MA: Harvard University Press.
Opp, Karl-Dieter, and Christiane Gern. 1993. “Dissident Groups, Personal Net-
works and Spontaneous Cooperation: The East German Revolution of
1989.” American Sociological Review 58, no. 5 (October): 659–80.
Rosenau, James N. 1990. Turbulence in World Politics: A Theory of Change and Con-
tinuity. Princeton, NJ: Princeton University Press.
———. 1997. Along the Domestic-Foreign Frontier: Exploring Governance in a Turbulent
World. Cambridge: Cambridge University Press.
———. 2003. Distant Proximities. Princeton, NJ: Princeton University Press.
Ruggie, John Gerard. 1986. “Continuity and Transformation in the World

Polity: Toward a Neorealist Synthesis.” In Neorealism and Its Critics, ed.
Robert O. Keohane, 131–57. New York: Columbia University Press.
———. 1993. “Territoriality and Beyond: Problematizing Modernity in Interna-
tional Relations.” International Organization 47, no. 1 (Winter): 139–74.
Saperstein, Alvin M. 1996. “The Prediction of Unpredictability: Applications of
the New Paradigm of Chaos in Dynamical Systems to the Old Problem
of the Stability of a System of Hostile Nations.” In Chaos Theory in the
Social Sciences: Foundations and Applications, ed. L. Douglas Kiel and Euel
Elliott, 139–63. Ann Arbor: University of Michigan Press.
Schelling, Thomas C. 1978. Micromotives and Macrobehavior. New York:
W. W. Norton.
Smith, Thomas S. 1997. “Nonlinear Dynamics and the Micro-Macro Bridge.” In
Chaos, Complexity and Sociology: Myths, Models, and Theories, ed. Raymond
A. Eve, Sara Horsfall, and Mary E. Lee, 52–63. Thousand Oaks, CA: Sage.
Smith, Thomas S., and Gregory T. Stevens. 1997. “Biological Foundations of
Social Interaction: Computational Explorations of Nonlinear Dynam-
ics in Arousal-Modulation.” In Chaos, Complexity and Sociology: Myths,
Models, and Theories, ed. Raymond A. Eve, Sara Horsfall, and Mary E.
Lee, 197–214. Thousand Oaks, CA: Sage.
Somit, Albert, and Steven A. Peterson. 1997. Darwinism, Dominance, and Democ-
racy: The Biological Bases of Authoritarianism. Westport, CT: Praeger.
Spruyt, Hendrik. 1994. The Sovereign State and Its Competitors: An Analysis of Sys-
tem Change. Princeton, NJ: Princeton University Press.
Taber, Charles S. 2001. “Of Spells, Potions, and Computational Social Science.”
Political Methodologist 10, no. 1 (Fall): 23–26.
Turing, Alan M. 1950. “Computing Machinery and Intelligence.” Mind 59,
no. 236 (October): 433–60.
Urry, John. 2003. Global Complexity. Cambridge: Polity.
Wendt, Alexander E. 1987. “The Agent-Structure Problem in International Re-
lations Theory.” International Organization 41, no. 3 (Summer): 335–70.
———. 1994. “Collective Identity Formation and the International State.” Ameri-
can Political Science Review 88, no. 2 (June): 384–96.
Wilson, Edward O. 1975. Sociobiology: The New Synthesis. Cambridge, MA: Har-
vard University Press.
Wilson, James Q. 1980. The Politics of Regulation. New York: Basic Books.
CHAPTER 9
When Worlds Collide

Reflections on the Credible Uses
of Agent-Based Models in
International and Global Studies
Desmond Saunders-Newton
It is quite possible that Agent-Based Models (ABMs)—or, more precisely stated,

agency-level computational models—will reshape the practice and use of social
science/inquiry. When contrasted with methodological approaches that solely
focus on the aggregate levels of analysis, it is possible to envision the realization
of an algorithmic social science and the more explicit inclusion of social science
theory and insights into the praxis of policy analysis and applied interna-
tional/global studies. Moreover, as alluded to in the title, ABMs allow for a
metaphorical collision of worlds, or rather an ability to combine, compare, and
decompose theoretical views of the world not easily accomplished in the past.
But before exploring these inquiry possibilities, it is prudent to bound these pos-
sibilities by considering the theoretical challenge to the credible use of computa-
tional models of agency.
The preceding chapter by Earnest and Rosenau questions whether there
can be a complex systems theory of political systems. They raise two objections.
The first is that complex systems theory “lacks both the empirical appeal of in-
duction and the disconfirmative value of deduction.” Therefore, they argue, it is
not really a theory. Their second objection is that complex systems theory cannot
capture and model the essence of political systems: the central role and impor-
tance of authority. For issues in which authority is less pervasive, like those in
some of the earlier chapters in this volume, the complexity paradigm offers new
thinking about old problems and suggests new hypotheses and explanatory
165
166 DESMOND SAUNDERS-NEWTON
methods. But Earnest and Rosenau argue that in those issue-areas in which au-
thority is more influential, complex systems theory cannot describe the impor-
tance of authority and ABMs cannot simulate its effects.
In this chapter, I do not directly challenge either of these criticisms of the
usefulness of complex systems theory in explaining or predicting political systems.
Earnest and Rosenau’s objections have merit in that they raise concerns about our
ability to rigorously create knowledge about our world, as well as the appropriate
substantive focus of the disciplines that constitute the field of study known as in-
ternational relations or world politics. On the other hand, their concerns about
our ability to reason from models, either via deduction, retroduction,1 abduction,2
or induction, can also be raised when we consider the “status quo” methodologi-
cal approaches used by scholars in this field. In my opinion, exchanges related to
methodological adequacy will likely fall into the category of reasonable scholars may
differ. Instead, I argue that the epistemological problems associated with ABMs are
worthy of note but that they are overstated; I show that ABMs can improve policy-
making even if both of these criticisms are true. Several US government agencies, in-
clusive of the Defense Advanced Research Projects Agency (DARPA) and the
Center for Technology and National Security Policy, are already exploring the de-
sign and employment of systems using multiple ABMs to generate policy options
and model potential costs and benefits of different choices.
Even before theorists have worked out the ontological and epistemological
problems associated with using complexity to explain political systems, we see the
use of complex systems thinking—and the agent-based models that it supports—in an
attempt to improve the choices of political leaders and reduce the risks of action in
international politics. Such emergent applications not only reflect a desire to bring
social science knowledge to bear on “problems of the day,” but also are consistent
with the innovation generated by the tools of this intellectual approach. As noted
by Frederick Turner, the new science (chaos and complexity) has “placed within our
grasp a set of very powerful tools—concepts to think with. We can use them well or
badly, but they are free of many of the limitations of our traditional [methodologi-
cal] armory” (Turner 1997, xii.) With this in mind, this chapter explores a scheme
for improving our ability to use these new perspectives and methods. This section
will be followed by a more explicit explication on some of the anticipated uses of
ABM methods in the realm of international security praxis and inquiry.
EFFECTIVE COMPUTATIONAL INQUIRY
The introductory chapter defined agent-based modeling as the simulation of

complexity in the social sciences. Such a methodology offers the prospect
WHEN WORLDS COLLIDE 167
of representing the interactions of agents comprising a heterogeneous system

of autonomous actors. This capability affords an opportunity to explore theo-
ries of complex international and global social phenomenon. I would like to
deepen this definition by asserting that agent-based modeling—or, more
broadly considered, computational methods that emphasize agency-level phe-
nomena—reflect a maturing transdiscipline that allows analysts and inquirers
not only to consider increasingly complex phenomenology in a rigorous fash-
ion, but also to pursue such inquiry in a more interdisciplinary fashion.
Agency-based modeling reflects a methodological approach that allows for
considering phenomena at levels of aggregation, or resolution, more granular
than that of the Westphalian nation-state. Several methodologies afford the ana-
lyst a means of studying the forms and dynamics of a social system, and they vary
by distinctive methodological styles and model ontologies (disciplines). Regardless
of pedigree, what these models have in common are an ability to explicitly con-
sider, and relate, agents, interactions, and environment. At least three distinct ap-
proaches potentially explain ecologies of agents and their interactions: multiagent
simulation, computational social network analysis, and sociophysics. A potential
typography for envisioning agency-based methodology is shown in figure 9.1.
Figure 9.1. A Typology of Agency-Based Methods

While these methodological approaches allow us to represent the behavior

of agents in the form of either individuals or institutions, they differ in terms of
the mechanisms that allow for—or give rise to—interactions between the agents. As
a result, computational ontologies, or rather the “what is” that is instantiated as
algorithm, will vary across methodological approaches, and this gives rise to epis-
temological implications: the ability to “know” will vary. Even though “what you
can know” may well differ between agency-based methodological approaches, the
ability to consider intuitively resonant levels of resolution is a true benefit of the
approaches. Put simply, this modeling approach makes sense to important sets of
constituents beyond the analytic community—for example, modeling neophytes with
decision-making authority and analysts trained in traditional modeling traditions and
shaped by hard-earned expertise.
Of course, no methodological approach comes cheap or free of costs. All
tools used by individuals involved in inquiry come with constraints such as the-
oretical assumptions or computational artifacts or a lack of ontological isomor-
phism.3 Thus, a prime concern as we consider the use of such a methodology is
its credibility (Dewar et al. 1996). Closely related to this notion of credibility is
the willingness of practitioners such as decision-makers or professional analysts
to accept the value-added data and information generated by a methodology. Put
another way, does the model result in a convincing or resonant narrative of
analysis or inquiry? To that end, I address the general issue of computational
epistemology and methodology as it relates to the use of computational models
of agency. Further, I consider how such concerns are more specifically related to
the examination of complex international and global social phenomenon.
COMPUTATIONAL EPISTEMOLOGY
AND METHODOLOGY
As a modeling formalism, ABM is an appropriate computational ontology for rep-

resenting much of the knowledge and data that arise from social actors operating
and interacting in an international and global context. As Bankes, Lempert, and
Popper suggest, one reason for the growing popularity of agent-based modeling is
its flexible representation of reality: “[I]t is an appropriate ontology for represent-
ing much of the knowledge and data that is available about social actors and so-
cial systems” (377). The following paraphrases suggest how Bankes, Lempert, and
Popper (2002, 377) characterize the efficacy of ABMs in the social sciences:
• The agents in an ABM can be based upon the wide breadth of

actors that arise in natural and artificial systems—for example,
individuals, groups, or institutions.
• The decision algorithms that these agents use can be based

on knowledge and data available regarding the decision be-
havior, perceived or actual, of the associated individuals,
groups, or institutions.
• What we know about the relationships between such agents
can easily fit into the agent-based mechanism.
• Environmental processes and effects that are “not inherently
agent based in character” can readily be reflected in hybrid com-
putational models with significant agent-based components.
In summary, and in comparison with competing modeling frameworks,

such as numerical models based on systems of differential equations or symbolic
logic or linear models of behavior, the agent-based approach provides much
greater facility for capturing the information that is available. Moreover, execut-
ing the resulting simulations can be used to infer the dynamic implications of
the combination of knowledge and assumption that is incorporated in the
model (Bankes, Lempert, and Popper 2002, 377).
While this modeling approach to representing human complex systems
can be very effective, representation and verisimilitude are only a part of our epis-
temological concerns. More precisely, what insights generated from the use of an
agent-based model are valid relative to how we frame phenomena that occur in
the world? Without doubt, this same question should be considered when em-
ploying other methodological approaches; the answer to this question when fo-
cused on non-ABM methodological approaches partly explains the many serious
attempts to incorporate ABM into social science/inquiry practice.
Whatever the shortcomings of non-ABM methods, it is still necessary to
consider the primary principle of epistemology: how do we credibly learn
“things” about the world by using ABMs? Furthermore, given the nature of ABM
as a computational method, how do we deepen our knowledge by performing
computations? While philosophical in their nature, these questions point to the
necessity of identifying and developing criteria for performing research based on
computational modeling.
To that end, I posit the following model-centric version of these questions
as a springboard for considering the issue of computational epistemology: What
must characterize a model in order for it to be useful in answering a scientific or
policy question? Focusing on this variant of the epistemological question, we are
drawn to consider the nature or use of models of social phenomena. I suggest
that there are at least two important uses of models: prediction and exploration.
With respect to predictive models, we immediately are faced with chal-
lenges that have harried users of analytic methods for studying social processes
since the first half of the 1900s. These challenges include a lack of veridicality
and associated difficulty with identifying the “correct” representation of the na-
ture and trajectory of human complex systems, as well as the correct frame for in-
cluding important physical science principles such as path dependency and
model parsimony. These characteristics are admittedly something of a caricature
of how social scientists practice their craft. For example, the inability to ascertain
the “truth” of how the world operates is not a challenge only faced by those en-
gaged in social inquiry. There continue to be questions across the disciplines of
physical and biological sciences about mechanisms that define the world’s oper-
ations—for example, the numerous frameworks for reconciling the quantum
world with the macroworld. Physics as a discipline is not viewed as any less rig-
orous for these debates. However, for a number of historical reasons, social sci-
ence praxis has been viewed as being less precise and rigorous than colleagues
involved in the physical sciences (Flyvbjerg 2001; Wallerstein 1996). As asserted
by Flyvbjerg, part of this perception is likely driven by using the wrong metric for
considering inquiry efficacy and quality.
Some of these issues, however, are partially addressed by technological ad-
vances. Problems once viewed as intractable, at least analytically, are less so as a
result of advances in computational methods and reductions in the cost of com-
putation. In addition, the relative importance of path dependency—or, more cor-
rectly, the ability to predict the one systemic trajectory—diminishes as we come to
value less the finding of the optimal responses to social problems. By focusing on
finding robust yet acceptable solutions that are valid across many possible or
plausible future outcomes, the need for perfect prediction is lessened.
This revision in perspective is important, given that a single model capable
of allowing for trusted assertions about the future states of a complex social sys-
tem would reflect a “mirror world” (Gelernter 1991). Such a model—that is, an
isomorphic algorithmic artifact—would be a model sufficiently correct that we
could peer into it and then learn about the world in which we exist. To that end,
a model is predictive if its output is comparable to outcomes in the real world
or actual system of interest within some well-characterized error process. Assum-
ing a model accomplishes this goal, and consequently supersedes the aforemen-
tioned challenges, epistemological issues are resolved. Much as a cured patient
does not question the successful treatment, if a model can be shown to accu-
rately represent the world and predict outcomes, epistemological quibbles
become irrelevant.
Many problems of interest to students of complex international and global
social systems are not amenable to predictive modeling. Moreover, this challenge
is not specific to ABM methodology. The inability to predict, however, does not
and should not suggest that the ABM approach is not useful. Given the afore-
mentioned epistemological concerns and the failure of other methods to predict
social outcomes, the inability of ABMs to predict social outcomes should not be
a primary measure of their efficacy. If, as presumed, complex systems are inher-
ently unpredictable, criteria other than the quality of point predictions must be
used to measure the quality of theory.
A very different and equally credible use of a model is in the context of a
computational experiment that supports the use of models in “exploration.” In
such an approach, computations in support of inquiry or generation of “surprise”
are viewed as an experiment. In such an instance, a computer-based model or sim-
ulation serves as a platform for performing computational experiments in which
one can map the inputs for a specific case to the outputs that measure the associ-
ated systemic behavior. In such a case, one can use a computational model or sim-
ulation to perform experiments whose outcomes are useful in constructing
credible arguments. In this inquiry framework, a model is not considered a mirror
of the world, but serves as laboratory equipment. Moreover, as noted by Bankes, a
good model is not necessarily the one that is an isomorph of the actual system, but
is rather one that can be used to perform crucial experiments that are useful in the context
of an argument or problem (Bankes, Lempert, and Popper 2002, 379). In fact, there is
no reason to believe that such a model need be realistic at all.
Modeling based on computational experimentation has been called “ex-
ploratory modeling.” It is differentiated from predictive modeling by not at-
tempting to limit the explicit uncertainty that arises from not having the one
“correct” model. Since predictive modeling arises from the praxis of theoreti-
cal science, it is biased toward deductive reasoning and measures research qual-
ity using the criteria of validity. The rigor of experimental science is based on
abductive and inductive logic, and is defined in terms of falsifiability and re-
producibility (Popper 1979). The epistemology of experimental science has
been considered in great detail, and by analogously casting computational social
science problems in terms of experimental science, many epistemological con-
cerns are mitigated. Thus, the methodology for using computational ap-
proaches such as ABM embraces a transition from using just one single,
“correct” model of the world to the use of an ensemble of alternative models.
These alternative models can be differentiated along dimensions of theory,
method, specification, and scenario.
Theoretical differentiation suggests comparing model results that are at-
tributable to the disciplinary frameworks underlying the models. Methodologi-
cal differentiation allows us to contrast, utilize, and synthesize modeling
techniques—for example, agent-based and systems dynamics simulations. The
Based upon Ed Waltz, Knowledge Management inthe Intelligence Enterprise (Boston Artech house, 2003),
p. 177, as modified by Desmond Saunders–Newton.
Figure 9.2. Non-Peircian Reasoning Processes
foci of these models are different (one is at the level of the agent and the other at
the macrosystem level), but each provides useful insights into understanding so-
cial behaviors in isolation, collaboratively and integratively (Saunders-Newton
and Graddy 2001). With respect to specification, it is easy to envision how
changes in model specification, which can be viewed as a separate model, can
give rise to very different results. As for scenario differentiation, this speaks to
coupling a given model with certain expectations about the world in which it
operates. For the policy community, the scenario describes how a represented
system may behave as a result of changes in policy or the social environment.
Regardless of the basis upon which these alternative models are generated,
the ensemble of models will likely contain more information than one single
model; by conducting large numbers of modeling experiments, it is possible to
derive insights through the exploration of the properties of an ensemble of
alternative models that no single model computer model could reveal. This
approach lies at the heart of the exploratory modeling approach (Bankes 1993,
1996; Dewar et al. 1996; Saunders-Newton and Scott 2001).
Further, by inferring across this ensemble as opposed to seeking insights
from one model, we can induce invariant properties and abduct robust responses
to these model outcomes. It is worth further emphasizing that these benefits are
made possible by the availability and use of computational experiments. More-
over, the use of induction and abduction in a collaborative fashion reflects an in-
triguing interaction between human and machine “intelligences.”
As illustrated by figure 9.2, the reasoning process reflects an interesting vari-
ation on the efforts of Charles Peirce’s consideration of reasoning.In this instance,
the graphic illustrates the various interrelations between reasoning approaches. The
integration of these approaches is very dependent on the ability to exploit compu-
tational approaches—for example, experimentation to expansively consider new hy-
potheses, models and responses. Further, one can use figure 9.2 to differentiate
between methods typically associated with standard hypo-deductive practice from
those that make use of exploratory modeling to reflect reasoning approaches such
as retroduction, abduction, and induction. As well as this ability to more effectively
use alternative reasoning or inferencing approaches, the relationship between
human “wetware” and algorithmic “software” can be structured to leverage human
ability to adeptly make inferences from complex patterns. Thus, we move further
along in the process of making use of algorithmic approaches to aid in decision-
making and inquiry.
TOWARD AN ALGORITHMIC SOCIAL SCIENCE
The previous section considered how traditional critiques of computational

models of social systems are not necessarily a true delimiting factor in the use of
ABM. In anticipation of an algorithmic social science with ABM as a prominent
methodological approach, a number of technologist, scientists, and policy ana-
lysts have began to explore how ensembles of ABMs can even now be used to
generate insights into complex social and political systems, and how this ap-
proach is being integrated into the policy community.
Before briefly describing an effort relevant to international and global
studies, it is important to make explicit the motivation for such efforts and a
rationale for using computational models as a key methodological pillar. With
respect to motivation, it is not difficult to assert that the issues of interest in
the realm of international security can be categorized as not only complicated,
but also often complex. While for many these terms are often synonymous,
Mark Lazaroff (BAE Systems Advanced Information Technologies, Intelligence Innovation Division)
and David Snowden (IBM).
Figure 9.3. Complicated and Complex Problems
they have very distinct meanings for research methodologists. As illustrated in

figure 9.3, a complicated problem can often be viewed as one with identifiable
casual relationships and known uncertainties. However, a complex problem is
one much more consistent with emergent patterns.
It is not difficult to further assert that the current international regime is
easy to represent as complex, and often viewed by practitioners and laypersons as
near-chaotic. Unfortunately, if the problem is defined as chaotic, or near-chaotic,
its supposed incomprehensible nature will often result in the unwillingness to
use rigorous, structured approaches to consider the problem. If such approaches
are used to consider a “wicked” problem, the necessary simplifications often
make the model and its associated results of questionable use to decision-makers.
The objective then is to frame near-chaotic problems so that they are classified as
complex, and to use a methodology that is appropriate to the problem domain.
Complexity science moves the problem from the chaotic to the complex domain,
while ABMs provide a means to assess phenomena in such regimes.4
An algorithmic social science—an applied and theoretical transdiscipline

based on the use computational models in the form of ABMs—not only affords a
more appropriate consideration of a “complex” problem domain, but also provides
a bridge between the two most contentious epistemological factions that reside in
the disciplines that study human processes: naturalistic inquiry and positivist or re-
ductionist science. Bridging these two “factions” is important as a means of exploit-
ing the insights derived about individuals and their artifacts, such as institutions
and technologies, in order to deal with ever-present challenges such as violent con-
flict and complex humanitarian disasters. Computational models, as defined by
Taber and Timbone (1996, 3), are a way to render theory in which a model is
loosely defined as a representation of theory about real-world phenomena that
serves as a bridge between theory and data. More to the point, models serve as a lan-
guage for expressing theory. Thus, a computational (algorithmic) social science is
the algorithmic instantiation of social science theory. As shown in figure 9.4, the
language of algorithmic social science (computational symbolic processing) falls be-
tween natural language and mathematical formalism. As a general statement, it is
able to provide insights into the deep narrative and rich detail associated with nat-
ural language, and maintain the rigor associated with mathematical dialects. An-
other way of envisioning such a language is as analytic narrative.
The ability to both make tough security problems less chaotic and provide a
methodology more supportive of effective analytical narrative5 is a driving motiva-
tion for a number of Department of Defense near-term and longer-term research
and development efforts using ABMs. While the efforts are fairly numerous, a few
Figure 9.4. The Language of Algorithmic Social Science

Figure 9.5. System-Level Architecture of the Pre-Conflict Management

Tools Program
directly resonate with a substantial amount of the literature in international and

global studies. These include efforts in the realm of preconflict and postconflict op-
erations—that is, actions necessary to avoid violent conflict or more effectively deal
with the consequences of using military force. One such effort is a program being
managed out of the National Defense University’s Center for Technology and Na-
tional Security Policy called the Pre-Conflict Management Tools (PCMT) Program.
This effort began in the spring of 2003 and was primarily concerned with develop-
ing a process and associated technologies for both anticipating “potential” violent
conflict, and identifying actions that could preclude or reduce the likelihood of
such an event. It is further thought that information and knowledge of relevance
to the avoidance of conflict can also be used to reduce conflict intensity and dura-
tion if such an event cannot be forestalled, and also enhance the likelihood of post-
conflict success. The general conceptual format for this effort is shown in figure 9.5.
The PCMT approach was an attempt to efficaciously couple a self-sustaining
analytic database with models of social vulnerability and authoritative networks
and process so as to support activities occurring in a collaborative decision-making
and analysis environment. While the database—which is concerned with the near-
real-time collection, categorization, and visualization of local- and regional-level

data, as well as traditional nation-state-level data—is of interest, of greater relevance
to this chapter is the aspect of the program concerned with identifying conditions
that could precipitate violent conflict and with better understanding the actors and
institutions capable of avoiding or ameliorating conflict. As originally envisioned,
the PCMT effort was to be composed of multiple models that allowed for the rep-
resentation of individual and institutional behavior and interactions at various
levels of resolution—for example, ABMs of local-, regional-, and national-level phe-
nomena and system-level models.
In light of budgetary constraints, the effort developed and integrated a lim-
ited number of models of social structure and dynamics in order to demonstrate
an ability to use multiple models in a coordinated and structured fashion. To
that end, the initial PCMT effort made use of three models of social vulnerabil-
ity and authoritative network dynamics. In terms of modeling classes, the mod-
els include an econometric model of development and civil war, a qualitative
computational model of internal conflict and state failure, and a model of au-
thoritative network dynamics. The econometric model is based on the work of
the World Bank’s Paul Collier, who focused on the relationship between civil
war and development policy (Collier et al. 2003), and the qualitative model is a
computational instantiation of the Fund for Peace’s Conflict Assessment System
Tool (CAST). The network dynamics models of authority were developed by the
Institute for Physical Sciences.
These models comprise a collection of disparate modeling approaches, in-
cluding a single ABM as of the summer of 2005, and are integrated by the use
of computer-assisted reasoning methods.6 It is important to note that ABMs are
of importance in thinking about the issues of societal formation and fragility, as
well as the networks of authority capable of redressing or addressing conditions
that give rise to social fragility. In addition, ABMs have proven quite helpful in
the collaboration effort, because they provide an effective starting point for con-
verting data into compelling information and knowledge for decision-makers
spanning various US government agencies and important stakeholders from
other societal sectors, such as coalition partners, nongovernmental organiza-
tions, and multinational corporations.
To craft information and knowledge so to increase its use in these deci-
sion-making environments, it is necessary to consider an appropriate means of
translating model output into decision-enabling information. The PCMT re-
searchers employ computer-assisted reasoning methods in conjunction with sim-
ulation models such as ABMs to create large ensembles of plausible future
scenarios. This particular stratagem supports a robust adaptive planning (RAP)
approach to reasoning under conditions of complexity and “deep uncertainty”
that typically defeat analytic approaches (Lempert, Popper, and Bankes 2002).7
It is anticipated that the RAP approach being used by the PCMT effort will pro-
vide greater insight into the vulnerabilities of societies and policies often under-
taken to address these vulnerabilities. ABMs are an important component of this
effort, particularly for the aspect of the PCMT architecture that is concerned
with identifying “proponents and opponents to peace,” and are viewed as credi-
ble ways of “knowing” in their own right. It is fully expected that once the PCMT
program moves from the technology demonstration stage into operational use, a
majority of all future models used in the social-vulnerability aspect of the pro-
gram will be ABMs.8
CLOSING THOUGHTS
In closing, the use of ABMs are viewed in a positive fashion by many persons inter-
ested in making better use of social science insights in future international security
operations. I would be remiss if I did not note that the pervasive engineering culture
may make more difficult the acceptance of social science disciplines, and their re-
lated models, because they cannot have the predictive accuracy of models of fluid
dynamics or classical dynamics. However, it is evident that proposed solutions gen-
erated by these highly accurate models are not particularly effective in addressing
the problems of greatest concern. For example, more accurate weapons do not
change the conditions that give rise to terrorist or transnational criminal networks.
Interestingly enough, the instantiation of many of the hard-earned insights
of social processes in algorithmic models is aiding in their acceptance among
many in this praxis community. The greater challenge may be less about ABMs
than about their use. Retraining analysts from formulating problems in a fashion
amenable to a point prediction or optimal solution to framing problems in terms
conducive to the generation of a robust solution or to computational experi-
mentation reflects a cultural shift that will require time.
However, an effort such as the PCMT program suggests that such a shift
may be less than a generation away. In fact, a number of efforts are currently ex-
ploring the use of ABMs to better understand how actions propagate forward into
time as consequences and externalities. Such a capability underpins operational
concepts such as “effects-based operations.” A number of organizations such as
the Defense Modeling and Simulation Organization9 and the Defense Advanced
Research Projects Agency are finding that ABMs are effective mediums for ad-
dressing the current analytic shortcomings in the area of effects-based operations
(Saunders-Newton and Frank, 2002) as well as for considering notions such as
long-term strategic assessment (Lempert, Popper, and Bankes 2003).
One interesting approach under consideration would allow us to address

an assertion often attributed to Herbert Simon: “[T]he soft sciences are actually
the really hard sciences.” Implicit in this assertion is our general inability to make
generalizable statements based on a sample size of one. Conversely, if we had a
larger sample of planets to observe, we could likely infer generalizable insights that
would aid us in thinking about actions on our own world. Unfortunately, the lack
of pervasive faster-than-light technology and a severe shortage of observable in-
habited planets make induction from a pool of planetary evidence difficult. How-
ever, societies or cultures created in-silica are now becoming increasingly possible.
By exploiting interesting advances in the realm of computational anthropology
(Gessler 2002), as well as insights garnered from the observation of persistent mas-
sively multiplayer online gaming communities, it possible to envision the creation
of artificial cultures as a means of better understanding the plausible trajectory of
cultures or societies being impacted by formulated policies and strategies, or
rather the possible consequences of actions as they propagate into time. A possi-
ble effort under consideration is the creation of persistent artificial cultures whose
emergent institutions and artifacts can be viewed over thousands of computer
years to assess social and cultural stability under various stressors. By instantiating,
growing, and maintaining large ensembles of these societies and cultures, it be-
comes possible to assess the impact of strategies, policies, and actions across ex-
tremely rich/deep computational models of different groups. Such a technical
approach would be quite consistent with the earlier discussion of experimental
design and computational epistemology, and would likely be impossible to realize
without the use of agency-based models of culture.
The challenges for developing and using such modeling artifacts are daunt-
ing, because they require thinking deeply about mappings between real-world
and virtual-world ontologies, as well as means of considering the epistemological
rigor and methodological sufficiency. These challenges, however, are not overly
different from the considerations we should make now when we attempt to use
analytic tools based largely on Newtonian physics to understand social processes.
Such considerations are the price we must pay in order to understand the
answers we receive to the questions we ask.
NOTES
1. While the concept of retroduction as originally defined by Charles

Sanders Peirce has multiple—seemingly contradictory—meanings, I suggest that it
can best be understood as the process of conjecturing a new hypothesis beyond
a current frame of discernment coupled with a search for evidence to affirm the
new hypothesis. It should be noted that some suggest that Peirce’s ultimate
intent was to relate deduction, abduction, and induction under the general con-
ceptual frame of retroduction (Chiasson 2001).
2. Abduction, or inference to the best explanation, is defined as a form of in-
ference that attempts to identify the most appropriate or plausible hypothesis for
explaining a given collection of data or body of evidence. This type of inference
process is constrained by the quality of each hypothesis—individually and relative
to one another—as well as the extensiveness of the search across “explanation or
hypothesis” space and the net importance of drawing a conclusion (Josephson
and Josephson 1996).
3. This refers to the less-than-perfect mapping between natural or real-
world ontology and the ontological principles underlying the instantiation of a
artificial society or culture in a computational environment.
4. In actuality, methods beyond ABMs will allow complexity per-
spectives to address problems in the chaotic domain—for instance, nonlinear
dynamics.
5. “Analytic narrative” is a phrase meant to suggest a rigorous, yet com-
pelling, means of sharing complex information or knowledge supportive of the
decision-making process.
6. Future variants of the PCMT effort will move toward including addi-
tional ABMs into the modeling suite.
7. As defined by Bayesian decision theorists, deep uncertainty is the con-
dition where the decision-maker does not know, or multiple decision-makers
cannot agree on, the system model, the prior probabilities for the uncertain
parameters of the system model, and/or the value function.
8. Other recent efforts that have explored the use of ABMs include the
Marine Corps Combat Development Center’s Project Albert and the Advanced
Research and Development Activities’ Non-Linear Human Dynamics Program.
9. In July 2003, DMSO managed a workshop exploring the ability
to model and simulate personality and culture. The expectation is to use in-
sights from this workshop as a means to improve the explanatory capabilities
and level of realism of the next generation of models and simulations
(Workshop 2003).
REFERENCES
Bankes, S. 1993. “Exploratory Modeling for Policy Analysis.” Operations Research

41, no. 3 (May–June): 435–49.
Bankes, S., and J. Gillogly. 1994. “Exploratory Modeling: Search through Spaces
of Computational Experiments.” In Proceedings of Third Annual Confer-
ence on Evolutionary Programming, ed. A. V. Sebald and L. J. Fogel,
353–60. River Edge, NJ: World Scientific.
Bankes, Steven, Robert Lempert, and Steven Popper. 2002. “Making Computa-
tional Social Science Effective: Epistemology, Methodology and Tech-
nology.” Social Science Computer Review 20, no. 4 (Winter): 377–88.
Chiasson, Phyllis. 2001. Peirce’s Pragmatism: The Design for Thinking. Amsterdam:
Rodopi.
Collier, P., V. L. Elliott, Havard Hegre, Anke Hoeffler, Marta Reyaal Querol,
and Nicholas Sambanis. 2003. “Breaking the Conflict Trap: Civil War
and Development Policy.” World Bank Policy Research Report No.
26121.
Dewar, James, S. Bankes, J. Hodges, T. Lucas, D. Saunders-Newton, and P. Vye.
1996. “Credible Uses of the Distributed Interactive Simulation (DIS)
Environment.” RAND Publications (MR-607-A).
Eve, Raymond A., Sara Horsfall, and Mary L. Lee. 1997. Chaos, Complexity, and
Sociology: Myths, Models, and Theories. Thousand Oaks, CA: Sage.
Gelernter, David. 1991. Mirror Worlds: Or the Day Software Put the Universe in a
Shoebox, How It Will Happen and What It Will Mean. Oxford: Oxford
University Press.
Flyvbjerg, B. 2001. Making Social Science Matter: Why Social Inquiry Fails and How It
Can Succeed Again. Cambridge: Cambridge University Press.
Gessler, Nicholas. 2002. “Computer Models of Cultural Evolution.” In Evolution
in the Computer Age. Proceedings of the Center for the Study of Evolu-
tion and the Origin of Life. Sudbury, MA: Jones and Bartlett Publishers.
Josephson, John R., and Susan G. Josephson, eds. 1996. Abductive Inference:
Computation, Philosophy, Technology. Cambridge: Cambridge Univer-
sity Press.
Lempert, Robert, S. Popper, and S. Bankes. 2002 . “Confronting Surprise.”
Social Science Computer Review 20, no. 4 (Winter): 420–40.
———. 2003. Shaping the Next One Hundred Years: New Methods for Quantitative,
Long-Term Policy Analysis. Santa Monica, CA: RAND Publications.
Peirce, C. S. 1931–58. The Collected Papers of Charles Sanders Peirce, ed.
C. Hartshorne and P. Weiss (Vols. 1–6) and A. Burks (Vols. 7–8).
Cambridge, MA: Harvard University Press.
Popper, Karl R. 1979. Objective Knowledge: An Evolutionary Approach. Oxford:

Oxford University Press.
Saunders-Newton, D., and H. Scott. 2001. “But the Computer Said! . . . A Typol-
ogy for Using Computational Modeling Methods in Public Sector
Decision-Making.” Social Science Research Review 19, no. 1 (Spring):
47–65.
Saunders-Newton, D., and Aaron Frank. 2002. “Effects-Based Operations:
Building the Analytic Tools.” Defense Horizon, no. 19. Center for Tech-
nology and National Security Policy Publication. October.
Saunders-Newton, D., and E. Graddy. 2001. “‘Way to Better Way’: Simulation
Paradigms, Decision Inferences, and Public Sector Enterprise Manage-
ment.” Proceedings of the Eurosim Congress, Delft, Netherlands.
Taber, Charles S., and Richard J. Timpone. 1996. Computational Modeling. Santa
Monica, CA: RAND Publications.
Turner, Fredrick. 1997. “Foreword: Chaos and Social Science.” In Chaos, Com-
plexity and Sociology: Myths, Models and Theories, ed. Raymond A. Eve,
Sara Horsfall, and Mary L. Lee. Thousand Oaks, CA: Sage.
Wallerstein, Immanuel. 1996. Open the Social Sciences: Report of the Gulbenkian
Commission on the Restructuring of the Social Sciences. Palo Alto, CA: Stan-
ford University Press.
“Workshop on Cultural and Personality Factors in Military Gaming.” 2003.
Sponsored by the Defense Modeling and Simulations Office. Alexan-
dria, VA, July 2003.
C H A P T E R 10
Complex Systems and the

Practice of World Politics
Neil E. Harrison
The study and practice of world politics has for too long been distorted by ratio-
nal choice. This conveniently simple model has misled generations of scholars
and policy-makers (Smith 2004). Like a cancer it changes minds and institutions
until its simpleminded rationality seems utterly human: “Taking a preference for
the maximization of self-interest or even utility as a given begets both a cognitive
and a political reality in which individuals and political leaders alike come to
view such behavior as normatively acceptable and as the standard by which gov-
ernment should operate. . . . Rational choice preserves the status quo. . . . Thus,
public policy as is becomes the public policy interest as it ought to be” (Petracca
1991, emphasis in the original).
This book has proposed a better way of understanding world politics.
Chapter 1 described the complexity paradigm built on an understanding of the
characteristics of complex systems and shows how ideas from complexity can be
adapted to world politics. Chapter 2 compared general and complex systems tax-
onomies and, thereby, further elaborated the complexity concepts and ideas that
may be used to construct complex systems theories of issue-areas in world politics.
Chapters 3 through 6 illustrated complexity and its benefits by applying com-
plexity concepts and sketching complex systems theories for specific issue-areas.
Chapters 7, 8, and 9 debated the epistemology and methods of complex systems.
In this chapter, I first show how complex systems concepts can improve
how we think about and understand world politics. In the next section, I con-
sider Earnest and Rosenau’s epistemological critique of complex systems theory
in chapter 8. In the third section, I show how complexity could reform policy in
world politics. In a short coda, I summarize the many benefits of the complexity
183
184 NEIL E. HARRISON
paradigm and the theories it can spawn. Throughout this chapter I indicate sev-
eral paths for further development and application of the complexity paradigm.
A BETTER WAY TO UNDERSTAND WORLD POLITICS
This book has outlined the concepts of a complex systems taxonomy for world
politics and offered four cases that demonstrate their application. Complex sys-
tems concepts can improve current theory, generate novel insights, and allow ex-
ploration of new possibilities.
Improving Current Theory
Complex systems concepts can extend or elaborate current knowledge, a critical

measure of a new taxonomy (see chapter 2). For example, Hoffmann shows how a
complex systems theory can provide the microfoundations to international nego-
tiations and agreements. Constructivism argues that interests and identity are con-
structed through interaction between states. Structure does not determine agent
choices, because “agents and structures are produced or reproduced by what actors
do” (Wendt 1994, 390). Thus, agents have a degree of freedom that introduces po-
tential for dynamic system change. But as states are treated as units (Wendt 1994,
385), constructivism cannot explain how states exercise this freedom of choice.
Thus, there is no explanation of the microfoundation of macroprocesses (and so
of the sources of change in the international system).
Conceptualizing the state as a complex system that is an agent in the
international system (a meta-agent), Hoffmann fills this gap in constructivist the-
ory. He treats international negotiations as the coevolution of adaptive states,
which fixes attention on the internal processes by which states exercise their
freedom in choosing their identity and interests and thereby influence interna-
tional processes and other states’ choices. Constructivists describe identity as
“grounded in the theories which actors hold about themselves and one another
and which constitute the structure of the social world” (Wendt 1992, 397). This
language is close to complexity concepts of internal models. Constructivism sees
identity as formed and changed through socialization in the international sys-
tem, but complexity expands behavior to include both the emergent domestic
processes and international coevolution, much as suggested, though more stati-
cally, in Putnam’s two-level games (1988).
Bhavnani’s analysis of the Rwanda genocide also uses complexity concepts
as an adjunct to conventional theories. He does not deny the usual explanations
of the causes of the violence. The ethnic hatreds, government propaganda, and
COMPLEX SYSTEMS AND THE PRACTICE OF WORLD POLITICS 185
the death of the Hutu president are well recognized. He uses complexity concepts
to answer a puzzle that conventional theories cannot touch: the initial conditions
of the Rwanda conflict, however reasonable, cannot explain the rapidity and mag-
nitude of the killing. However, the dynamic evolution of the killing can be cap-
tured by complex systems concepts that fill out the narrative of simpler models.
As most conventional theory is at the level of the international system,
there are many opportunities to add microfoundations with complexity concepts
(which do not preclude greater simplifications at higher levels of analysis). Issue-
area theories also can be adapted. For example, complex systems could explain
how epistemic communities actually influence states’ policies in international
environmental issues.1
Generating Novel Insights
In a dynamic world, a constant stream of new ideas and hypotheses is essential

to understanding. Paradigms change because explanatory failures of the old par-
adigm have accumulated, a new paradigm is available that explains more than
the old, and the majority of scholars in the field recognize these two conditions
(Kuhn 1970). Not only have rational choice paradigms failed to explain much of
the institutionalism of the modern international system but they also have mis-
led policy-makers (Smith 2004).
Complex systems concepts can generate radically novel hypotheses. Walt
Clemens hypothesizes that the Protestant practice of debating holy texts institu-
tionalized activities essential to effective democracy and open markets: literacy
and open and civilized debate on policy matters. Reminiscent of Weber’s (1958)
explanation of capitalism, this is an unexpected but very plausible hypothesis
generated using complexity concepts. A bifurcation in the cultural paths of some
peoples seems to have made them more resilient under empire and more suc-
cessful at exploiting the opportunities presented by its demise. Informal institu-
tions developed from religious conviction generate politically effective behaviors
that manage ethnic differences and support economic development. A complex
systems theory of development led Clemens to ask different questions and
search for data in new places.
While correlation is not causation, Clemens’s argument is novel and cer-
tainly plausible, and deserves further investigation. Because it shows similarities
with some theories of economic development (e.g., De Soto 2000), it may have
application elsewhere. Clemens offers some empirical evidence in support of his
hypothesis and outlines how to more fully test it. Whether or not the hypothesis
is disproved, it illustrates complexity’s possibilities for innovative thinking about
old and new problems.
Exploration
In addition to consolidative or predictive modeling that represents reality with

extensive data sets,2 complexity can generate exploratory models to test ideas
about the relations of agents, systems, and environment. Abstract mathemati-
cal models are used to work through the logic of relationships among physical
variables, and Einstein’s thought experiments helped to develop relativity the-
ory. Saunders-Newton argues that exploratory simulations through an ensem-
ble of models can generate more information, more flexibly, than a single
model can. By reducing uncertainty and generating novel insights into social
behavior and multiple action options, exploratory modeling can improve pol-
icy. Using metaphor and argument rather than models, Sandole shows that
contending views can be reconciled and policies crafted to prevent ethnic con-
flict or mitigate its effects.
THE NATURE OF POLITICS AND

EPISTEMOLOGY AND METHODS
Earnest and Rosenau’s critique of complex systems theory is rich and detailed.
There is not space here to respond fully to every matter they discuss. So, in this
section I assess the reasonableness of the two premises on which they found their
arguments. First, they note that there is no epistemology of complex systems in
world politics. Without one, they argue, there can be no theories of complex sys-
tems. Second, they criticize the potential for isomorphism between model and
reality in world politics. Because politics is about authority that, by definition,
limits (constrains) self-organization, they question whether world politics in
reality is a complex system as commonly understood.
Epistemology
Earnest and Rosenau’s critique is based in a positivist epistemology. While

positivism is the dominant world politics methodology, it is problematic in so-
cial science and rejects simulation as neither logical nor empirical and, thus, as
without meaning.
There is agreement on what epistemology is but not on how to pursue it.
Epistemology is the effort to distinguish true knowledge from false and helps to
determine if one theory is better than alternate theories. Knowledge is conven-
tionally defined as something like “justified true belief.”3 Positivism asserts that
there are two sources of knowledge: deductive logic and empirical evidence. The
former is a priori true; the latter must be verifiable by experience. The meaning
of a statement is equivalent to a definition of the empirical conditions under

which it is true or false. Thus, a statement is meaningful only if it can be ob-
served or can, in theory, be observed (Ayer 1936).4 The best empirical verifica-
tion of a proposition is correct prediction of observable events.
Positivism is problematic in the social sciences because of problems of
observation and of correspondence between abstract concepts and real objects.
The problem of identifying valid empirical indicators for theoretical concepts is
broadly recognized but overlooked in deference to the simplicity of positivism.
Arthur (1994) similarly rejects positivist/rationalist theory in economics as a
beautiful model that bears little relationship to the reality of human behavior.
Humans are less deductive and more inductive than positivism posits.
Evolutionary epistemology is one explanation of knowledge accumulation
that should be considered for complex systems in social science. Blind variations
(almost guesses) in knowledge are selected through biological, psychological, and
social processes and retained according to their contribution to individual or
group survivability (Campbell 1960).5 In this view hypothesis generating simula-
tions and exploratory modeling are as legitimate sources of knowledge variations
as any other. Once selected, knowledge variations become part of the selection
mechanism for further variations, in effect mimicking selection by reality. Orga-
nized in a nested hierarchy of selectors, the knowledge system becomes ever
more intelligent and adaptive. The principle of downward causation—“all
processes at the lower level of a hierarchy are restrained by and act in conformity
to the laws of the higher level” (Campbell 1974)—completes a complexity model
of knowledge formation. The whole is partly constrained by the behavior of the
parts (emergence), but the parts are partially constrained by the whole (down-
ward causation).
Finally, an epistemology can be designed to specifically accommodate sim-
ulations generating “justified true belief.” For example, Marney and Tarbert
(2000) discuss an epistemology in which simulation in social science is a valuable
“third leg” supplementing and complementary to theorizing and empirical test-
ing. The purpose of science, they argue, is to map theoretical constructs onto ob-
servable reality (much like positivism). But mapping in the social sciences is
shaky at best, because (unlike the physical sciences) social science reality does not
fit into binary logic categories. Physical objects are either members or not mem-
bers of a category. An electron is nothing but an electron; it is never a molecule.
But human subjects can “belong to one or more of a number of referent groups
or be in any one or more of a number of psychological states” (para. 5.16). Thus,
in the social sciences the “cross-mapping and redundancy” that simulation can
provide are an especially valuable contribution to knowledge accumulation
(a point Saunders-Newton also makes).
Authority in Complex Systems
Taking authority as the critical variable in world politics, as do Earnest and Rose-
nau, is a value-laden move. Easton’s (1981) definition of politics as the authori-
tative allocation of valued things is not the sole conventional definition of
politics nor necessarily the most appropriate for our purpose. Politics has also
been defined as the formation and rivalry of groups (Schmitt 1976) and, more
generally, as the generation of the structures and norms that govern human col-
lectivities (for example, Arendt 1958). Following this latter view, politics is the
process by which the institutions governing collective life are organized. Under-
standing politics this way, it is eminently reasonable to model it as a complex sys-
tem. Some agents in the system always have more influence over the form and
function of institutions than other agents do. But authority is not the whole
measure of politics and, except for subscribers to the realist view, not the sole
object of research in world politics.
For three reasons, the existence of authority is not fatal to complex sys-
tems theories of world politics. First, authority operates through formal and in-
formal institutions. Informal institutions, like cultural practices, are shared
meanings and emerge from agent interactions mediated through prior states of
such institutions.
Second, social systems are not binary—for example, either authoritarian or
not. It is always a matter of degree. I suggest that all societies are complex and
can be modeled with complexity concepts, but some have more central control,
and thus less complexity, than others.
Third, for authority to be fatal to complexity it must be centralized. As
Earnest and Rosenau acknowledge, globalization is diffusing authority from the
state to other organizations (see Strange 1996). While individual agent decision-
making may be as or more limited than before, it is because of less centralized au-
thority. Freedom House reports that in 1900 no state was an electoral democracy
with universal suffrage. By 2000, 120 of 192 countries were rated as electoral
democracies. Coupled with economic liberalization, proliferating democracy di-
minishes centralized authority and increases self-organization and complexity.
The decentralization of authority from the state to a large number of diverse pri-
vate and public organizations competing economically and politically itself cre-
ates complexity, and the influence of authority in the modern world political
system can be better captured through complexity concepts than through a sim-
ple model more relevant in a past era of state dominance.6
Even within highly centralized authority systems, agents always have
choices. They can choose to follow orders or they can refuse, accepting pun-
ishment. Often in social situations there is a third way: not to follow or refuse
but to “game” the system, working around the rules to personal advantage.7
For all its control and punishment, the Soviet state did not absolutely control
the behavior of its people, and popular resistance became informally institu-
tionalized. This human ability to adapt to environmental conditions reduces
the importance of formal authority and makes system response to authority
less predictable.
Earnest and Rosenau rightly argue that authority should be treated
endogenously as a property of the system. This permits investigation of how
authority emerges, evolves, and dissolves and how it influences formal and
informal institutions. However authority is defined and handled within complex
systems theories, it is an advance in world politics that the effect of authority is
even considered problematic. In rational choice theories, the role of authority
is clear but clearly wrong.
Complexity Methods
Evolutionary epistemology rejects positivism’s deductive creation of a theoretical

world from self-evident truths (axioms) in favor of models that represent reality.
In this and related epistemologies, theory, model, and phenomena are “inde-
pendent entities,” and science comprises analytical and ontological activities that
relate theory to phenomena with models—and theories become sets of models
(Henrickson and McKelvey 2002).
Models of complex systems may be mathematical or computational. Com-
mon mathematical techniques for analyzing stochastic dynamic systems in several
disciplines (including economics) can be adapted for complex systems. For example,
some complex systems can be described by Markov chains in which the distribution
of future events is independent of the history of the system. At any time, the future
state of the system is a probabilistic function of the present, without regard to the
past. Under these conditions, modified Monte Carlo methods for estimating sam-
ple distributions can reduce computational complexity. So, the relevant statistical
techniques are generally referred to as “Markov-chain Monte Carlo.” Richards
(2000) reviews mathematical methods for modeling nonlinear political systems.
Cioffi-Revilla (1998) uses several mathematical methods to develop a for-
mal theory of politics and uncertainty. Although he does not specifically refer
to complexity or complex systems and he does not theorize from individual
agents, he discusses nonlinear systems, how macrobehavior comes from
microevents, and the influence of macrocontext on microlevel phenomena.
His methods go beyond ordinary least-squares regression analysis to include
Boolean logic, nonlinear and maximum likelihood estimation, and survival and
event history analysis.
Every paradigm comes with its own language; for some scholars the lin-
guistic gymnastics of postmodernism is as challenging as formal theory and
mathematical notation are for others. The modeling challenge of complexity re-
ally is no different from the difficulties of researching across language and cul-
ture barriers, a skill quite common in comparative politics and IR.
As with any paradigm, the choice of method depends on the research
puzzle being investigated. Some students of world politics will build custom
programs or learn advanced mathematics but others may use several propri-
etary or open source software simulation packages (Agent Sheets, Swarm, etc.).
Still others may wish only to specify units and interaction rules and “out-
source” model construction to computer programmers, applied mathemati-
cians, or statisticians.
REFORMING POLICY
The most dangerous aspect of positivist, rational choice theories is that they fos-
ter the belief that the causes of problems and the consequences of alternate re-
sponses can be known with a high probability. The most important service of a
complexity paradigm might be to free policy-makers from this overweening
hubris and inculcate a sense of uncertainty. The greatest contributor to the suc-
cess of the Kennedy administration’s nuanced response to the Cuban Missile
Crisis was the failure of the overly simplistic decision-making for the Bay of Pigs
(Janis 1982; Allison 1971). Sandole argues that Realpolitik policy can never end
ethnic conflict, because it feeds off the biological belonging that bifurcates the
social world into “us” and “them” and also drives ethnocentrism and ethnic con-
flict. Thus, success in the War on Terror demands accepting a more inclusive
and nuanced interpretation of events than offered by simple Realpolitik.
To intervene effectively in complex systems requires, first, that policymak-
ers recognize the inherent uncertainties in their understanding of both the sys-
tem and the effects of our interventions therein. Second, policy must seek out
points of leverage that may be well hidden.
Complex systems in world politics demand policy caution. Brian Arthur
suggests that when intervening in complex systems “you want to keep as many op-
tions open as possible. You go for viability, something that’s workable, rather
than what’s “optimal”. . . because optimization isn’t well defined anymore. What
you’re trying to do is maximize robustness, or survivability, in the face of an ill-
defined future. And that, in turn, puts a premium on becoming aware of nonlin-
ear relationships and causal pathways as best we can. You observe the world very,
very carefully, and you don’t expect circumstances to last” (quoted in Waldrop
1992, 331–34). The best that policy-makers can hope for are policies that are
excellent rather than optimal, and they need to be prepared to change.8 While
consistent pursuit of ideological purity may enhance reelection, pragmatic policy
is more realistic.
Cautious intervention demands a broad conceptualization of possibilities.
In Sandole’s analysis of ethnic conflict, neither Realpolitik nor Idealpolitik can
stand alone. A TIT FOR TAT strategy reflects complexity by combining the need for
order and security with “niceness” and “foregiveness.” Order and security alone
cannot allay underlying tensions and change hearts and minds. But building
Idealpolitik’s positive peace usually requires Realpolitik’s negative peace of militar-
ily imposed stability. In complexity terms, a social group’s behavior comes from
the interaction of its internal model with external reality. Realpolitik advocates
regulating the conflicting groups’ environment to increase selection pressure for
appropriate changes in their internal models. But internal models can also be
influenced by education, as Idealpolitik advocates. Combining education and
selection increases the rapidity of learning and norm change.
Because complex systems are counterintuitive, good policy requires think-
ing broadly about problems and finding leverage points for intervention. Mead-
ows (1997) suggests nine possible leverage points. The least useful—namely,
changing parameters in the system—is “diddling with details, arranging the deck
chairs on the Titanic. . . . If the system is chronically stagnant, parameter changes
rarely kick-start it. If it’s wildly variable, they usually don’t stabilize it.” But para-
meters are where we put “probably ninety-five percent of our attention.” This is
simple policy.
Meadows’s four most effective intervention points are the most relevant to
the present discussion. First, changing the rules of the system changes behavior.
Positive and negative incentives work. As scholars of world politics have long re-
alized, malfunction of systems can often be traced to the rules and “who has
power over them.” Second, self-organization drives economic processes, techno-
logical innovation (Nelson and Winter 1982), and other social changes. As mu-
tations drive evolution, increasing social diversity increases self-organization and
emergence: “Let a thousand flowers bloom and ANYTHING could happen” (Mead-
ows 1997). This is a powerful and dangerous point of intervention that govern-
ments concerned with control and predictability rarely consider. A third and
bigger point of leverage is the goals in the system. In political terms, this is
Kennedy’s inaugural address (“ask not what government can do . . .”) and Rea-
gan’s call to get government off the backs of the people. Changing the goals of
the system works through the internal models of system agents.
Finally, the most important leverage is the “paradigm or mindset out of
which the system arises.” Cultural norms work on each individual’s internal
model and through informal institutions select behaviors that are socially ap-
propriate. Industrial societies will continue to blithely consume nonreplaceable
resources as long as their paradigm is anthropocentric mastery of “nature” (Har-
rison 2000). In world politics, the increasing acceptance of the rational choice
paradigm in policy circles makes scholars complicit in the policies that form the
reality of world politics (Smith 2004). In Gramscian language, a social paradigm
or “mind-set” is supported by a hegemony of economic and political goals and in-
tellectual and moral discourse (Hoffman 1984). Redirecting political trajectories
requires change in both structures and ideas; in the language of complexity, it
means reinforcing changes in both institutions and internal models. Gramsci’s
approach was incremental, but ideational change may be better accepted when it
is so sudden and substantial as to be an epiphany.
In complex systems, problems are unclear, solutions have uncertain effects,
and points of leverage are never simple. But complex systems can be pushed and
prodded, and changed; yet, caution is required and instruments are imprecise. Be-
cause institutions and social systems are influenced by human perceptions of the
world and how it works, dethroning the rational choice paradigm is the best way
for scholars to positively influence world politics. But policy under complexity
opens many other avenues of research, and the benefits are likely to be great.
TOWARD THEORY
Earnest and Rosenau argue that there cannot be a general “complex systems the-
ory” in world politics. There are no acceptable standards by which we can know
knowledge of complex political systems, and ABM simulations, the method-
ological basis for a complex systems theory, exogenize authority, the essence of
political systems. In this chapter, I have argued that epistemologies for complex
systems theories can be found and that authority in modern industrial democ-
racies does not defeat complexity and its simulation. But I do accept their gen-
eral position, though for a different reason. Some of the central tenets of
complexity—emergence and the importance of context and initial conditions—do
make a generally applicable and ahistorical complex systems theory of world pol-
itics logically impossible.
Complexity is more than a metaphor—the taxonomy outlined in chapters
1 and 2 forms a basis for theory-building—but less than an encompassing theory.
Yet, complex system theories of individual issue-areas in world politics are possi-
ble. Differences between issue-areas in agents’ values and beliefs and in informal
institutions (norms) and formal ones (organizations, etc.) may prevent an all-
encompassing complex systems theory. But theories—sets of premises, assump-
tions, concepts, and statements of their relations from which testable predictions
can be made—that endogenize these issue-area characteristics are not only possi-
ble but preferable. Humans have different expectations in regard to environ-
mental issues—more ethical and inclusive ones (Harrison 2000)—than in regard
to economic issues, where they are more individualist, and they pursue their
goals differently in each issue-area. Similarly, the institutions that mediate
agents’ interactions are unique to each issue-area. At the international level, the
World Trade Organization and the United Nations Environmental Programme
are structured and operate very differently.
The complexity paradigm offers a novel perspective on world politics at all
levels that will generate new theories and models of issue-areas. It also encour-
ages innovative methods for understanding political reality and advising policy-
makers. This book has defined complexity concepts and ideas and introduced
many potential theoretical challenges, and throughout this chapter I have noted
many directions for theoretical development of theories of complex systems and
their empirical testing. This paradigm can increase our understanding of the
complexity of world politics and reduce the probability of surprising events.
NOTES
1. Risse-Kappen (1994, 187) argues that the theory of epistemic commu-

nities has failed to specify “the conditions under which specific ideas are selected
and influence policies while others fall by the wayside.”
2. Historical experimentation discussed in chapter 2 would be of this type
if the right data could be found.
3. It was conventional to define knowledge so until Gettier (1963)
showed that this is an incomplete definition without some additional condi-
tions. For our present purposes, the conventional or short-form definition of
knowledge is adequate.
4. For example, until the proof by Eddington in 1919, Einstein’s conclu-
sion that gravity bends light was only a proposition, but the conditions under
which it could be verifiable were clear.
5. For example, Popper’s view of science as accumulating knowledge by the
selection mechanism of refutation operating on randomly evolving conjectures.
6. The extent to which states have lost authority or other organs have
gained authority is much debated (e.g., Strange 1996).
7. The general population breaks official rules regularly and with relative im-
punity. The IRS audits about 1 percent of all returns, most cars on the interstates
drive up to ten miles an hour over the speed limit (seemingly accepted as a “safe”
margin by drivers and police), and perhaps 10 percent of GDP is unmeasured and
belongs to the “black economy.”
8. This is “satisficing” (Simon 1955). One mechanism for finding an ex-
cellent (but not the very best) solution is “simulated annealing.” See Kauffman
1995, 248–52.
REFERENCES
Allison, Graham T. 1971. The Essence of Decision. Boston: Little, Brown.

Arendt, Hannah. 1958. The Human Condition. Chicago: University of Chicago
Press.
Arthur, W. Brian. 1994. “Complexity in Economic Theory.” American Economic
Review 84, no. 2:406–11.
Ayer, A. J. 1936. Language, Truth, and Logic. New York: Oxford University Press.
Campbell, Donald T. 1960. “Blind Variation and Selective Retention in Cre-
ative Thought as in Other Knowledge Processes.” Psychological Review
67, no. 6:380–400.
———. 1974. “Evolutionary Epistemology.” In The Philosophy of Karl Popper, ed.
Paul Arthur Schilpp, 413–643. The Library of Living Philosophers, 14.
LaSalle, IL: Open Court.
Cioffi-Revilla, Claudio. 1998. Politics and Uncertainty: Theory, Models and Applica-
tions. Cambridge: Cambridge University Press.
De Soto, Hernando. 2000. The Mystery of Capital: Why Capitalism Triumphs in the
West and Fails Everywhere Else. New York: Basic Books.
Easton, David. 1981. The Political System: An Inquiry Into the State of Political
Science. Chicago: University of Chicago Press.
Gettier, Edmund L. 1963. “Is Justified True Belief Knowledge?” Analysis
23:121–23.
Harrison, Neil E. 2000. Constructing Sustainable Development. Albany: State Uni-
versity of New York Press.
Henrickson, Leslie, and Bill McKelvey. 2002. “Foundations of ‘New’ Social Sci-
ence: Institutional Legitimacy from Philosophy, Complexity Science,
Postmodernism, and Agent-Based Models.” Proceedings of the National
Academy of Sciences of the United States of America 99, Suppl. no. 3 (May
14): 7288–95 (accessed at www.pnas.org on May 15, 2005).
Hoffman, John. 1984. The Gramscian Challenge: Coercion and Consent in Marxist
Political Theory. New York: Basil Blackwell.
Janis, Irving. 1982. Groupthink. Boston: Houghton Mifflin.
Kauffman, Stuart. 1995. At Home in the Universe: The Search for the Laws of Self-
Organization and Complexity. New York: Oxford University Press.
Kuhn, Thomas. 1970. The Structure of Scientific Revolutions. 2nd ed. Chicago: Uni-
versity of Chicago Press.
Marney, J. F., and F. E. Tarbert. 2000. “Why do Simulation? Towards a Work-
ing Epistemology for Practitioners of the Dark Arts.” Journal of Artificial
Societies and Social Simulation 3, no. 4 (accessed at http://www.soc.
surrey.ac.uk/JASSS/3/4/4.html).
Meadows, Donella H. 1997. “Places to Intervene in a System.” Whole Earth, Win-
ter (accessed at http://www.wholearthmag.com/ArticleBin/109.html
on February 7, 2003).
Nelson, Richard R., and Sidney G. Winter. 1982. An Evolutionary Theory of Eco-
nomic Change. Cambridge, MA: Harvard University Press.
Petracca, Mark P. 1991. “The Rational Choice Approach to Politics: A Chal-
lenge to Democratic Theory.” Review of Politics 53 (Spring): 289–319.
Putnam, Robert D. 1988. “Diplomacy and Domestic Politics: The Logic of Two-
Level Games.” International Organization 42, no. 3 (Summer): 427–60.
Richards, Diana. 2000. Political Complexity: Nonlinear Models of Complexity. Ann
Risse-Kappen, Thomas. 1994. “Ideas Do Not Float Freely: Transnational Coali-
tions, Domestic Structures and the End of the Cold War.” International
Organization 48, no. 2 (Spring): 185–214.
Rosenau, James N., and Mary Durfee. 2000. Thinking Theory Thoroughly: Coher-
ent Approaches to an Incoherent World. Boulder, CO: Westview Press.
Schmitt, Carl. 1976. The Concept of the Political. New Brunswick, NJ: Rutgers
University Press.
Simon, Herbert A. 1955. “A Behavioral Model of Rational Choice.” Quarterly
Journal of Economics 69 (February): 99–118.
Smith, Steve. 2004. “Singing our World into Existence: International Relations
Theory and September 11.” International Studies Quarterly 48, no. 3
(September): 499–515.
Strange, Susan. 1996. The Retreat of the State: The Diffusion of Power in the World
Economy. Cambridge: Cambridge University Press.
Weber, Max. 1958. The Protestant Ethic and the Spirit of Capitalism. Trans. Talcott
Parsons. New York: Charles Scribner’s Sons.
Wendt, Alexander. 1992. “Anarchy Is What States Make of It: The Social Con-
struction of Power Politics.” International Organization 6, no. 2 (Spring):
391–425.
———. 1994. “Collective Identity Formation and the International State.” Ameri-
can Political Science Review 88:384–96.
Contributors
ROBERT AXELROD is Arthur W. Bromage Distinguished University Professor

of Political Science and Public Policy at the University of Michigan. He is best
known for his interdisciplinary work on the evolution of cooperation which has
been cited in more than five hundred books and four thousand articles. His cur-
rent research interests include complexity theory (especially agent-based model-
ing), and international security. His most recent books, The Complexity of
Cooperation: Agent-Based Models of Competition and Collaboration and (with Michael
D. Cohen) Harnessing Complexity: Organizational Implications of a Scientific Frontier,
explore the application of complexity to understanding politics and organiza-
tions. He has been elected president of the American Political Science Associa-
tion for 2006-07.
RAVI BHAVNANI, Ph.D. in political science (University of Michigan-Ann

Arbor, 2003), is an Assistant Professor at Michigan State University. He received
his doctoral degree in comparative politics and methodology, with an emphasis
on agent-based modeling. His research focuses on the micro-foundations of mass
participation in ethnic violence, civil war, and popular rebellion in Sub-Saharan
Africa and South Asia.
WALTER C. CLEMENS, JR. is Professor of Political Science, Boston University,

and Associate, Harvard University Davis Center for Russian and Eurasian Studies.
He has written or edited fifteen books, including The Baltic Transformed: Complexity
Theory and European Security (2001) and Dynamics of International Relations: Conflict
and Mutual Gain in an Era of Global Interdependence (2d. ed., 2004). His current re-
search focuses on the ways that revolutions in literacy and free thought, initiated in
Europe circa 1500, have contributed to societal fitness as understood by complex
systems theory. Clemens studied in Vienna and Moscow but received an A.B. from
Notre Dame University, Magna Cum Laude, and Ph.D. from Columbia University.
He has taught at the University of California at Santa Barbara and at M.I.T. and lec-
tured widely in Eurasia, the Americas, and along the Pacific Rim.
197
198 CONTRIBUTORS
DAVID C. EARNEST is an Assistant Professor of Political Science and Interna-

tional Studies at Old Dominion University, Norfolk, Virginia, where he teaches
international political economy, international relations theory and political
methodology. His substantive research focuses on the political incorporation of
migrants in democratic societies, while his methodology interests are in the ap-
plication of agent-based models to problems of international politics. He is co-au-
thor of On the Cutting Edge of Globalization: An Inquiry into American Elites
(Rowman and Littlefield, 2005; with James N. Rosenau, Ole R. Holsti and Yale
H. Ferguson). Previously he held an appointment as a Fellow in Political-Military
Studies at the Center for Strategic and International Studies in Washington, DC.
NEIL E. HARRISON’s research interests include complex systems and interna-

tional environmental politics. His book Constructing Sustainable Development
(SUNY 2000) linked both research interests. He also has co-edited Science and
Politics in the International Environment (Rowman and Littlefield 2004) with Gary
Bryner and has published other articles and chapters on international environ-
mental politics. He received his doctorate from the Graduate School of Interna-
tional Studies at the University of Denver in 1994 and has taught in Colorado,
Wyoming, and Taiwan.
MATTHEW J. HOFFMANN is an Assistant Professor in the department of

Political Science and International Relations at the University of Delaware. His
research focuses on global environmental governance, complexity theory, and so-
cial constructivism. His recent publications include the book Ozone Depletion and
Climate Change: Constructing a Global Response from SUNY Press (2005) and a
coedited volume Contending Perspectives on Global Governance with Alice Ba (Rout-
ledge Press 2005).
JAMES N. ROSENAU is University Professor of International Affairs at The

George Washington University. A former president of the International Studies
Association, he has authored a number of books and articles. Among his recent
books three stand out retrospectively as a trilogy: Turbulence in World Politics: A
Theory of Change and Continuity (Princeton 1990), Along the Domestic-Foreign Fron-
tier: Exploring Governance in a Turbulent World (Cambridge 1997), and Distant Prox-
imities: Dynamics Beyond Globalization (Princeton 2003).
DENNIS J. D. SANDOLE received his Ph.D. in Politics from the University of

Strathcyde in Glasgow, Scotland in 1979. He is professor of conflict resolution
and international relations at the Institute for Conflict Analysis and Resolution
(ICAR) at George Mason University. A founder-member of ICAR, he worked
closely with conflict resolution pioneer John Burton in Britain and the US. His
main areas of interest include “identity” conflict/conflict resolution in the
CONTRIBUTORS 199
Balkans, Middle East, Caucasus, Central Asia and Southeast Asia; e.g., under-
standing and dealing with the complex etiology of what has come to be called in
the Western media “suicide bombing” and other acts of terrorism. His most re-
cent book is Capturing the Complexity of Conflict: Dealing with Violent Ethnic Con-
flicts of the Post-Cold War Era (Thomson Learning 1999).
DESMOND SAUNDERS-NEWTON is a member of the senior management

staff of BAE SYSTEMS Intelligence Innovation Division where he leads the
Social Computation and Complexity Directorate. He also is an Adjunct Associ-
ate Professor in the University of Southern California’s School of Policy, Plan-
ning & Development and a member of the External Advisory Board of George
Mason University’s Center for Social Complexity. He has held appointments
as a consulting scientific advisor at DARPA, a senior scientist and program di-
rector at the National Defense University’s Center for Technology & National
Security Policy, and served as the American Association for the Advancement
of Science S&T Advisor to the Deputy Under Secretary of Defense for Ad-
vanced Systems and Concepts. In addition to his work in the areas of interna-
tional security and science/technology policy, Dr. Saunders-Newton has over
a decade of experience in research methodology and social science research
(domestic and international) in a variety of research agencies. He holds degrees
from the Pardee RAND Graduate School (Ph.D., M.Phil), University of Michi-
gan (MPP), and Lawrence University (B.A.), and has pursued post-doctoral stud-
ies at RAND and the Santa Fe Institute.
J. DAVID SINGER earned his Ph.D. in 1956 from New York University and is
Professor Emeritus at the University of Michigan. He has taught at Vassar and in
Oslo, Geneva, Groningen, Taipei, and Mannheim. He founded the Correlates of
War Project and was its Director from l963 to 2003. He has published more than
twenty books including Nations at War: A Scientific Study of International Conflict
(Cambridge University Press 1998 with Daniel Geller) and Resort to Arms: Inter-
national and Civil War, 1816–1980 (Sage 1982, with Melvin Small), and over one
hundred articles including “The Level-of-Analysis Problem in International
Relations” published in World Politics in 1961.
SUNY series in Global Politics
James N. Rosenau, Editor
American Patriotism in a Global Society—Betty Jean Craige

The Political Discourse of Anarchy: A Disciplinary History of International
Relations—Brian C. Schmidt
Power and Ideas: North-South Politics of Intellectual Property and Antitrust—Susan K. Sell
From Pirates to Drug Lords: The Post–Cold War Caribbean Security Environment—
Michael C. Desch, Jorge I. Dominguez, and Andres Serbin (eds.)
Collective Conflict Management and Changing World Politics—Joseph Lepgold and
Thomas G. Weiss (eds.)
Zones of Peace in the Third World: South America and West Africa in Comparative
Perspective—Arie M. Kacowicz
Private Authority and International Affairs—A. Claire Cutler, Virginia Haufler, and
Tony Porter (eds.)
Harmonizing Europe: Nation-States within the Common Market—Francesco G. Duina
Economic Interdependence in Ukrainian-Russian Relations—Paul J. D’Anieri
Leapfrogging Development? The Political Economy of Telecommunications Restructuring—
J. P. Singh
States, Firms, and Power: Successful Sanctions in United States Foreign Policy—George
E. Shambaugh
Approaches to Global Governance Theory—Martin Hewson and Timothy J. Sin-
clair (eds.)
After Authority: War, Peace, and Global Politics in the Twenty-First Century—Ronnie
D. Lipschutz
Pondering Postinternationalism: A Paradigm for the Twenty-First Century?—Heidi
H. Hobbs (ed.)
Beyond Boundaries? Disciplines, Paradigms, and Theoretical Integration in International
Studies—Rudra Sil and Eileen M. Doherty (eds.)
International Relations—Still an American Social Science? Toward Diversity in Interna-
tional Thought—Robert M. A. Crawford and Darryl S. L. Jarvis (eds.)
Which Lessons Matter? American Foreign Policy Decision Making in the Middle East,
1979–1987—Christopher Hemmer (ed.)
201
202 SUNY SERIES IN GLOBAL POLITICS
Hierarchy Amidst Anarchy: Transaction Costs and Institutional Choice—Katja Weber

Counter-Hegemony and Foreign Policy: The Dialectics of Marginalized and Global Forces
in Jamaica—Randolph B. Persaud
Global Limits: Immanuel Kant, International Relations, and Critique of World Politics—
Mark F. N. Franke
Money and Power in Europe: The Political Economy of European Monetary Cooperation—
Matthias Kaelberer
Why Movements Matter: The West German Peace Movement and U. S. Arms Control
Policy—Steve Breyman
Agency and Ethics: The Politics of Military Intervention—Anthony F. Lang, Jr.
Life After the Soviet Union: The Newly Independent Republics of the Transcaucasus and
Central Asia—Nozar Alaolmolki
Information Technologies and Global Politics: The Changing Scope of Power and
Governance—James N. Rosenau and J. P. Singh (eds.)
Theories of International Cooperation and the Primacy of Anarchy: Explaining U. S.
International Monetary Policy-Making After Bretton Woods—Jennifer Sterling-Folker
Technology, Democracy, and Development: International Conflict and Cooperation in the
Information Age—Juliann Emmons Allison (ed.)
Systems of Violence: The Political Economy of War and Peace in Colombia—Nazih
Richani
The Arab-Israeli Conflict Transformed: Fifty Years of Interstate and Ethnic Crises—
Hemda Ben-Yehuda and Shmuel Sandler
Debating the Global Financial Architecture—Leslie Elliot Armijo
Political Space: Frontiers of Change and Governance in a Globalizing World—Yale Fer-
guson and R. J. Barry Jones (eds.)
Crisis Theory and World Order: Heideggerian Reflections—Norman K. Swazo
Political Identity and Social Change: The Remaking of the South African Social Order—
Jamie Frueh
Social Construction and the Logic of Money: Financial Predominance and International
Economic Leadership—J. Samuel Barkin
What Moves Man: The Realist Theory of International Relations and Its Judgment of
Human Nature—Annette Freyberg-Inan
Democratizing Global Politics: Discourse Norms, International Regimes, and Political
Community—Rodger A. Payne and Nayef H. Samhat
SUNY SERIES IN GLOBAL POLITICS 203
Landmines and Human Security: International Politics and War’s Hidden Legacy—
Richard A. Matthew, Bryan McDonald, and Kenneth R. Rutherford (eds.)
Collective Preventative Diplomacy: A Study of International Management—Barry
H. Steiner
International Relations Under Risk: Framing State Choice—Jeffrey D. Berejikian
Globalization and the Environment: Greening Global Political Economy—Gabriela Kütting
Sovereignty, Democracy, and Global Civil Society—Elisabeth Jay Friedman, Kathryn
Hochstetler, and Ann Marie Clark
United We Stand? Divide and Conquer Politics and the Logic of International Hostility—
Aaron Belkin
Imperialism and Nationalism in the Discipline of International Relations—David Long
and Brian C. Schmidt (eds.)
Globalization, Security, and the Nation State: Paradigms in Transition—Ersel Aydinli
and James N. Rosenau (eds.)
Identity and Institutions: Conflict Reduction in Divided Societies—Neal G. Jesse and
Kristen P. Williams
Globalizing Interests: Pressure Groups and Denationalization—Michael Zürn (ed., with
assistance from Gregor Walter)
International Regimes for the Final Frontier—M. J. Peterson
Ozone Depletion and Climate Change: Constructing A Global Response—Matthew
J. Hoffmann
States of Liberalization: Redefining the Public Sector in Integrated Europe—Mitchell
P. Smith
Mediating Globalization: Domestic Institutions and Industrial Policies in the United
States and Britain—Andrew P. Cortell
The Multi-Governance of Water: Four Case Studies—Matthias Finger, Ludivine
Tamiotti, and Jeremy Allouche, eds
Building Trust: Overcoming Suspicion in International Conflict—Aaron M. Hoffman
Global Capitalism, Democracy, and Civil-Military Relations in Colombia—Williams
Avilés
Index
3PF. See Three Pillar Framework assumptions, 168

4⫹2 framework, 55 contingent identities in, 152
defined, 126, 127
abduction, 166, 171, 173, 180 DOD research of, 175, 176, 180
ABM programs in DOD, 175, 176, 180 endogenizing authority in, 155
ABMs. See Agent-based models epistemological concerns, 171
actor sensitivity, 51 epistemological implications of, 168
adaptive behavior epistemology of, 169
compared to rational choice, 139 exploration by, 171
adaptive planning, robust, 177, 178 identity in, 150
agency, 27 incomplete, 158
and independent thinking, 88 implementation, 129
and structure, 7 information capture by, 169
contingency, 157 inquiry possibilities of, 165
agency-level computational models. method problems, 145
See agent-based models methods of, 151, 157
agent-based modeling modeler assumptions, 144
as transdiscipline, 167 modeler expectations, effect of, 150–52
agent-based models, 5, 27, 33, 38, problem of simple behavior rules, 152
113–14, 122–23, 126–27, 140, 167, typology, 167
173–74 users of, 168
advantages of, 177 uses of, 169
and aggregate studies of war, 134 agent-based systems, 82
and decision rules, 150 agents, 3, 114, 158, 168
and dynamic processes, 127 adaptation by, 98
and emergence, 132 adaptive, 99
and equation-based models, 130–31 adaptive mechanisms, 126
and game-theoretic models, 131–32 and social system, 98
and understanding processes, 140 basic needs of, 87
appropriate ontology, 168 behavior of, from internal model,
as aid to policy-making, 166–68 9–10, 102
as experiments, 171 defined, 3, 9, 98
205
206 INDEX
agents (continued) Bush, George W., 31, 50, 57

diversity of, 132 butterfly effect, 74
ecologies of, 167
in social relations, 128 causation, 11–13, 18, 27, 34–36, 64
modeling adaptive behavior of, 129 in open systems, 8
ahistorical theory, 192 in world politics, 13
algorithmic social science, 173–78 cause and effect. See causation
language of, 175 CAS. See complex adaptive systems
analysts Cederman, Lars-Erik, 13, 157
need for retraining, 178 CFCs. See chlorofluorocarbons
analytic narrative, 180 chaos, 166
ant queen, myth of, 143 “edge of,” 58, 63, 74, 89, 153
Antarctic ozone “hole,” 100 chaotic problems, 174
arms races, 51 Chechnya, 86
Arthur, W. Brian, 13, 98, 190 chlorofluorocarbons, 99–101
assumptions industry, 107
by modeler, effect of, 148 link to ozone depletion, 103
autarky, 39 technology, 106
authority clash of civilizations, 12, 47, 58
and coercion, 154 Club of Rome. See “Limits to Growth”
and complexity, 144, 146, 149 coethnic punishment, 133
breakdown of, and CAS, 154 coevolution, 74, 81, 184
central in politics, 152 in North and South, 108
diffused by globalization, 188 of internal models, 113
emergence of, 152–57 coevolutionary processes, 98, 110
in complex systems, 188–89 Cold War, 1, 29, 43
problem of endogenizing, 155 Commonwealth of Independent States, 81
problematic, 189 complex adaptation, 96
“Turing test” for, 157 complex adaptive systems (CAS), 73–90
authority cascades, 154 (Also see complex systems)
Axelrod, Robert M., 13, 52, 53, 54, 96, and constructivism, 156
144, 149, 154 and politics, 144
as thought experiments, 149
balance of power, 11, 56 breakdown of authority, 154
Baltic States, 80 concepts of, 74–75
behavioral adaptation, 123 conceptual tools of, 89
Bible contingency in, 148
date in vernacular, 85–86 fundamental insight of, 83–84
publication dates, 90 no authority, 153
biology world politics as, 146
and prejudice, 48 complex systems
effect on group identity, 48–49 (Also see complex adaptive systems)
biopolitics, 144 and conflict studies, 51–52
Bin Laden, Osama, 12 and rules, 5–6
blue eyes/brown eyes test, 46 characteristics of, 3
Bosnia, 52–57, 65 compared to simple systems, 2–5
bottom-up modeling, 140 different from general systems, 32–36
INDEX 207
leverage points in, 190–92 superior to orthodoxy, 112

nonlinearity in, 5 complexity methods, 189–90
predictability of, 4 complexity science, 11, 73, 174
complex systems theories, 165 defined, 2
and induction, deduction, 149 computational epistemology, 179
challenge of, for IR/world politics, 145 computational models, 133, 175
defined, 2 as methodological pillar, 174
from simple rules, 5 computational social network analysis, 167
limits to methods of, 149 conflict
not deductive, problems from, 148 and initial conditions, 63
potentially integrative, 38 as balancing mechanism, 55
complexity, 28 as process, 52, 53, 63
aggregate studies of war, compared, 134 conflict resolution, 50–51
and authority, 144, 146, 149, 188–89 and complexity, 52–57
and change, 58 conflict transformation, 63
and conflict resolution, 52–57 consolidative modeling, 127
and constructivism, 7, 9, 10 constructivism, 7, 9, 10, 29, 184
and decision-making, 3–4 assumptions of, 158
and equilibrium, 4–5, 52 contingency
and exploration, 186 in complex adaptive systems, 148
and living systems, 3 complex adaptation, 97
as generative science, 149 Correlates of War, 30, 34
as metaphor, 144 COW. See Correlates of War
causation in, 11–12, 34–36 credibility of methods, 168
compared to orthodox IR/world culture, importance of, 84–86
politics theories, 6–13
compared to complicated, 174 Dayton Peace Agreement, 52, 54, 55, 57
conceptual tools of, 87, 89 decentralization of authority, 188
conditions, small changes in initial, 52 decision algorithms, 169
defined, 2–6 decision rules, contingency in, 151
epistemology of, 186–87 decision-making
epistemology of, non-positivist, 158 and complexity, 3–4
from simple rules, 5 and simplification, 174
generates novel insights, 185 deduction, 148, 166, 180
improving orthodox theory, 184–85 deep narrative, 175
in social sciences, 143 deep uncertainty, 180
in world politics, 6–13 Defense Advanced Research Projects
measures of, 3–4 Agency, 178
new thinking from, 33, 166, 185 Defense Modeling and Simulation
paradigm or theory, 147 Organization, 178
parsimony in, 37 democracy, 76
patterns in, 147 and self-organization, 81
policies proposed by, 84 in form only, 86
policy implications of, 190–92 Department of Defense
problems of metaphor in politics, 153 research in ABMs, 175, 176, 180
proposed empirical testing, 87–89, dissipative structures, 4
111–14, 127–30 dissipative systems. See dissipative structures
208 INDEX
DOD. See Department of Defense exploratory modeling, 127, 173

domestic politics, 7, 9, 28 external, internal view, 38
dynamic systems, 10–11 extra-rational, 31
dynamics at microlevel, 122
feedback, 28–29, 30–31, 98
early warning system, 61 negative, positive, 4
“edge of chaos,” 58, 63, 74, 89, 153 positive, 29
Einstein, Albert, 193 feedback, iterative, 150
elites, political, 28 fitness, 11, 74–75, 80, 82
Elliott, Jane, 46–47 acquiring fitness, 84–86
emergence, 7, 32–34, 81–82 inclusive, 48, 64
and ABMs, 132 of countries, 75, 84–86
of authority, 152–57 of society, 73, 89
of universal participation rule, 109 fitness criterion
of ethnic norms in ABMs, 128 modeler’s choice of, 151
emergent properties, 147 fitness landscape, 75, 83, 86
of state, society, 74 flip. See Systemic change
emergent phenomena, 74 formal institutions, 188
empirical testing, of complexity, proposed, formal modeling, 113
87–89, 111–14, 127–30 Freedom House, 80, 188
EPA. See US Environmental Protection
Agency Gell-Mann, Murray, 3
epistemic communities, 185, 193 general systems, theories, 27
epistemology, 171, 179 compared to complex 32–36
evolutionary, 187 genetic algorithms, 150, 151, 158
for simulations, 187 globalization, 188
of complexity, 186 Gramsci, Antonio, 192
non-positivist, 158, 186–87
positivist, 186 HDI. See UN Human Development Index
equilibrium, unstable, 4–5, 52 Hobbes, Thomas, 153
ethnic consciousness, 88 Holland, John H., 5, 27, 98, 144
ethnic entrepreneurs, 125, 129 homeostasis, 5, 11
ethnic groups, 132 human behavior, 149
ethnic norms, 45–51, 125–26 and biology, 47–49
defined, 125 Human Development Index. See UN
emergence of, 126 Human Development Index
ethnic violence, 121–26 human nature, 6, 44, 47–48, 49
and bottom-up processes, 124 human social networks, 74
ethnocentrism, 43, 45–51
defined, 45 Idealpolitik, 47, 55, 57
in-group, 45 identity, 43, 158
sociobiology of, 48 issue specific, 156
sources of, 47–49 induction, 140, 148, 166, 171, 173
EU. See European Union inference-making, improving, 173
European Union, 61, 80, 82, 99, 103 informal institutions, 188
evolutionary epistemology, 187 in-group, 45, 128
experimentation, 34 initial conditions, small changes in, 52
INDEX 209
inquiry efficacy and quality, 170 Locke, John, 6

institutions, 11, 27, 35, 84, 188, 193 logical positivism, 158
interest groups, 151 London Amendment, 108
interests, 9, 44, 158 Luther, Martin, 84–86, 88
accidental, path dependent, 152
interhamwe, 123 macrobehavior and microevents, 132, 189
internal models, 9–10, 35, 47, 49, 98, mapping, 187
132, 191 Markov chains, 12
adaptation of, 108 and Monte Carlo methods, 189
and agent behavior, 102 mastery of nature, 192
change in South, on ozone depletion, mathematical models, of complexity, 189
104 Maturana, Humberto, 10
diversity in, 125 measures of theory quality, 171
operationalizing, 114 media, 28
problem of simple behavior rules, 152 mental model. See internal model
US change in, on ozone depletion, 106, meta-agents, 8–9, 27, 34–35, 114, 184
110 method, for research puzzle, 190
internal rule models. See internal models methodology
international organizations, 36 adequacy of, 166
international relations. Also see world falsification as, 158
politics simulative, 157
international relations theories, orthodox, styles of, 167
7, 12, 27, 32–36 methods of complexity, 33, 185, 189–90
failures of, 1–2 microevents and macrobehavior, 189
compared to complexity, 6–13 micro-macro linkages, 146
international system, 57 models. Also see Agent-based models
international theorists. See international defined, 170
relations theories differentiation between, 171
Islamic societies, and literacy, 86 ensemble of, 171
in social science, predictive accuracy
Jervis, Robert, 13 of, 178
jihad, 133 isomorphism of, 170
Jevons, W. Stanley, 7 mathematical, computational, 189
nature of, 28
Kaplan, Morton, 27 predictive, defined, 169–70
Kauffman, Stuart A., 89, 143 reasoning from, 166
Koran. See Qur’an specification of, 172
Kuhn, Thomas S., 57 use by analysts, 178
modeler assumptions, not theoretically
learning, by agents, 9, 35 neutral, 151
levels of analysis, 26, 27, 35, 165 modeling
leverage points in complex system, 190–92 exploratory, 173
“Limits to Growth,” 33 rules or attributes, 150
literacy Montreal Protocol, 29, 100, 101, 105,
evolution of, 87 110
importance of, 84–86 and US internal model, 106–8
without debate, 86 multi-agent simulation, 167
210 INDEX
national consciousness, 88 parsimony and complexity, 113

National Defense University, 176 path dependence, 12, 29–30, 148, 152, 170
national interests, 32 PCMT. See Pre-Conflict Management
nation-state. See state Tools Program
North Atlantic Treaty Organization, 54, peace
56, 61, 80 negative, 56, 63
Natural Resources Defense Council, 106 positive, 63
nature and nurture, 49 peace-keeping, 57
negative peace, 56 Peirce, Charles Sanders, 173, 179
neoclassical economics, 6 perception, of decision-makers, 35
neorealism, 10 points of leverage, in complex systems,
New European Peace and Security 190
Systeem (NEPSS), 60–63, 64, 65 policy effects, uncertainty of, 192
Newton’s universe, 6 politics, and complex adaptive systems,
nonlinear rules of behavior, 130 144
nonlinear systems, 189 Popper, Karl, 193
nonlinearity, 12 positive sum common security, 61
as metaphor, 144 positivism, 1, 158, 186–87
norm of universal participation, 107 dangers of, 190
norms, 30, 99 in social sciences, 187
emergence of, 126 positivist methods. See positivism
ethnic, 125–26 postmodernism, 190
of behavior, 121 post-Soviet Eurasia, 73, 75–77
North Atlantic Treaty Organization. and complexity, 81–83
See NATO democracy, market capitalism in, 77
NRDC. See Natural Resources Defense ethnic issues in, 80
Council zones of, 76
power, 44, 56
obedience, 133 Pre-Conflict Management Tools Program,
ontology, 26 176–78, 180
of agent-based models, 168 prejudice, and biology, 48
open systems, 4, 8, 28 preventive diplomacy, 61
order for free, 82 Prigogine, Ilya, 4, 29
organization, compared to structure, 10 Prisoner’s Dilemma, 53
Organization for Security and Coopera- problem solving, global, 63
tion in Europe, 54, 61, 62 problem, moving from chaotic to
orthodox theories, improving, 37–38, complex, 174
108–11, 184–85 problems, unclear, 192
Osama Bin Laden, 12 process tracing, 114
OSCE. See Organization for Security and processes, 13
Cooperation in Europe constitutive, 157
out-group, 45, 128 of decision-making, 31
ozone depletion regime, 95, 99–108, 110 simulation, insight into, 138
understanding of, with ABMs, 140
paradigm shift, 57 provocability, 53, 54, 56
parsimony, 36, 170 punctuated equilibrium, 75, 82–83
and complexity, 37 Putnam, Robert D., 10
INDEX 211
Qur’an, 86 North only, 106

Northern problem, 101
racial prejudice, learned, 46 of ozone depletion, 101
rational choice, 7, 9, 96, 109, 115, 139, operationalizing, 114
183, 185 South demands change, 107–8
compared to adaptive behavior, 139 universal participation, 101
rationality, 30–32 Russian Federation, 80
objective, 9, 32 Rwanda, 121–25
of social aggregations, 31 characteristics, 123
subjective, 9–10
rationality assumption, appeal of, 150 satisficing, 150, 194
realism, 11, 32 scenario differentiation, 172
complexity compares, 2 Schelling, Thomas, 147
hegemonic theory, 54 schema. See internal models
Realpolitik, 43–47, 55, 57 security, positive sum common, 61
and War on Terror, 50 selection, 56
reasoning selection mechanism, 193
computer-assisted methods, 177 self-organization, 2–6, 83
non-Peircian process, 172 and democracy, 81
recursive interaction, 38 and diversity, 191
refutation, as selection mechanism, 193 and free thought, debate, 86
regime and hierarchy, 84
emergence, 96 in world politics, 6–13
formation, for ozone depletion, 99–108 self-organized criticality, 75, 82
ozone, 110 sensitivity
participation, 97 of actors, 51
regime theory, 95, 102, 109 to initial conditions, 64
and complexity, 108–11 September 11, 2001, 1, 12, 43, 49, 50, 51,
challenges to, 96 64
changes from complexity, 108 Serbia, 52–57
system rules exogenous, 112 Serbs, 65
Republic of Srpska, 54 “shadow of the future,” 53, 55
research puzzle and choice of method, Simon, Herbert, 9, 179
190 simple systems, 2–5
retroduction, 166, 179 and complex compared, 2–5
Robbers Cave experiment, 46 automobile as, 2
robust adaptive planning, 177, 178 characteristics of, 3
rule model. See internal model defined, 2
rule of participation, 103 predictability of, 4
emergence of, 109 static, 4
rules, 5 simplification
contingency in decision, 151 and rigorous problem analysis, 174
of behavior, nonlinear, 130 by common knowledge, 17
rules, system, 97, 111, 191 simulated annealing, 194
dynamic compared to static, 102 simulation, 113
evolution of, 101 as experiment, 34
for participation, 109 as research methodology, 138, 139
212 INDEX
simulation (continued) general, 32

definition, 137 general and complex compared, 32–36
epistemology of, 187 goals of, 191
incomplete, 158 nested, 26–27
multi-agent, 167 paradigm of, 191
of CAS, testing theorems, 148 social, continuously evolve, 29
of complex systems, examples, 5 system rules, 111, 191
principal value of, 138 operationalizing, 114
uses of, 137, 138 systemic change, 29
with computers, 147 in Yugoslavia, 58
simulation software, 158 systemic trajectory, 170
simulative methods, 157
simulative-empirical research design, 149 taxonomy, 27, 32
Singer, J. David, 25 criteria of good, 36–37
social capital, 134 defined, 25
social networks, 134 taxonomy of complexity, 6
social organization, as attachment technology, effect on social science, 34
behavior, 143 terrorism, 58, 62, 64
social science theories, general systems, 27, 32–36
algorithmic, 173–78 theories, international relations. See
and positivist epistemology, 187 theories, world politics
generalizable statements in, 179 theories, world politics, 10, 12
problems of, 179 compared to complexity, 6–13
social science models, predictive accuracy failures of, 1–2
of, 178 orthodox, 7, 12, 27, 32–36
societal fitness, 73, 83, 89 theory quality, measuring, 171
sociophysics, 167 theory, ahistorical, 7, 192
solutions, robust compared to optimal, Three-Pillar Framework, 59–60
170 defined, 60
state, 17, 30–31, 35 time, arrow of, 4
as complex adaptive actors, 97 Tit-for-Tat, 53, 55, 125, 191
as emergent system, 7 top-down modeling, 140
as meta-agents, 8–9 Transparency International, 77
feedbacks within, 30–31
in orthodox theories, 9 Ukraine, 80
state of nature, 6 umuganda, 124
stratospheric ozone depletion. See ozone UN Development Programme, 88
depletion UN Human Development Index, 78–80,
structural factors, 122 83, 85–86, 87, 88, 90
structure, 11 countries ranked by 79–80
and agency, 7 uncertainties, inherent, 190
compared to organization, 10 uncertainty, 51
surprise, in world politics, 1 United Nations Environment Programme,
systems, 104, 193
and psychology, 33 United States, 99, 103
defined, 2, 33 internal model on ozone depletion,
emergent properties in, 33 102, 106, 110
INDEX 213
unstable equilibrium, 4–5, 52 World Bank, 177

U.S. See United States world politics
U.S. Environmental Protection Agency, as complex adaptive system, 146
106, 110 as processual flow, 13
as self-organizing complex system, 2,
Varela, Francisco, 10 6–13
veridicality, 170 orthodox and complex compared, 6–13
Vienna Convention, 99, 100, 101, 103, orthodox theories of, 7, 12, 27, 32–36
115 world systems theory, 10
World Trade Organization, 193
Waldrop, M. Mitchell, 5 worldview, 17
Waltz, Kenneth, 10
War on Terror, 77, 190 Yugoslavia, 60, 76
and Realpolitik methods, 50 internal conflict, 52–57
wicked problems, 174 Yugoslavian war, 53
POLITICAL SCIENCE
Complexity in World Politics

Concepts and Methods of a New Paradigm
Neil E. Harrison, editor
Despite one hundred years of theorizing, scholars and practitioners alike

are constantly surprised by international and global political events. The
collapse of communism in Europe, the 1997 Asian financial crisis, and 9/11
have demonstrated the inadequacy of current models that depict world
politics as a simple, mechanical system. Complexity in World Politics shows
how conventional theories oversimplify reality and illustrates how concepts
drawn from complexity science can be adapted to increase our understand-
ing of world politics and improve policy. In language free of jargon, the book’s
distinguished contributors explain and illustrate a complexity paradigm of
world politics and define its central concepts. They show how these concepts
can improve conventional models as well as generate new ideas, hypotheses,
and empirical approaches, and conclude by outlining an agenda of theoretical
development and empirical research to create and test complex systems
theories of issue-areas of world politics.
“This book is well written and easily accessible, with essays by some of
the major thinkers in the field of complexity science. It makes a number
of intellectual contributions and helps fill a gap in the existing literature.”
— Scott E. Page, coeditor of Computational Models in Political Economy
Neil E. Harrison is Founder and Executive Director of the Sustainable Develop-
ment Institute and the author of Constructing Sustainable Development, also
published by SUNY Press.
A volume in the SUNY series in Global Politics

James N. Rosenau, editor
State University of New York Press

www.sunypress.edu

Harrison Complexity in World Politics

Uploaded by

Copyright:

Available Formats

You might also like

Harrison Complexity in World Politics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Harrison Complexity in World Politics

Uploaded by

Copyright:

Available Formats

Complexity

James N. Rosenau, editor

Concepts and Methods of a New Paradigm

State University of New York Press

© 2006 State University of New York

All rights reserved

Printed in the United States of America

No part of this book may be used or reproduced in any manner whatsoever

For information, address State University of New York Press,

Production by Judith Block

Library of Congress Cataloging-in-Publication Data

Complexity in world politics : concepts and methods of a new paradigm / edited by

1 Thinking About the World We Make 1

2 Complexity Is More Than Systems Theory 25

3 Complexity and Conflict Resolution 43

4 Understanding and Coping with Ethnic Conflict 73

5 Beyond Regime Theory: Complex Adaptation 95

6 Agent-Based Models in the Study of Ethnic Norms 121

7 Alternative Uses of Simulation 137

8 Signifying Nothing? What Complex Systems Theory 143

9 When Worlds Collide: Reflections on the 165

10 Complex Systems and the Practice of World Politics 183

List of Titles, SUNY series in Global Politics 201

Thinking About the World We Make

agent interactions that we call “complexity”. The study of complexity in systems

FROM SIMPLE TO COMPLEX

A system is a portion of the universe within a defined boundary, outside of

TABLE 1.1. CHARACTERISTICS OF SIMPLE AND COMPLEX SYSTEMS

Few agents Many agents

computers to govern feedback cycles and responses to environmental changes.

The characteristics summarized in table 1.1 and described in this section

COMPLEX SYSTEMS IN WORLD POLITICS

Constructivism broadens “the array of ideational factors that affect inter-

The uncertainty of complex social systems calls into question conventional

theories presume that causation is proximate and proportionate. Like most of

PLAN OF THE BOOK

Drawing on complex adaptive systems theories (a version of complexity

phenomena in an interdisciplinary way. In addressing the epistemological issues

5. As a skeptic, Hume also argued that causation is a human construct.

Collins, R., R. Hanneman, and G. Mordt. 1995. “Discovering Theory Dynamics

C. Roe Goddard, John T. Passé-Smith, and John G. Conklin, 9–24.

Complexity Is More Than Systems Theory

Neil E. Harrison with J. David Singer

The previous chapter emphasized the differences between conventional theories

A taxonomy “has no descriptive, predictive or explanatory power, since it contains

faults of general systems theories and a well-argued explanation of the benefits of

Agents, Not Actions

Political systems are conventionally assumed to be closed and homeostatic: dis-

reduced the time available for consideration of alternatives by decision-makers,

Path Dependence and Randomness

Rationality assumptions are used as a convenient simplification in both orthodox

Related to the orthodox presumption that social aggregations are rational

DIFFERENCES BETWEEN THE GENERAL AND THE COMPLEX

The “natural” sciences—most notably physics, mechanics, and chemistry—are still

Cause and Effects

Complex-systems concepts encourage theory that covers multiple levels of analy-

the interplay among participants in decision-making groups within the executive.

Encompassing and Improving Orthodoxy