Hindawi

Complexity
Volume 2020, Article ID 6105872, 16 pages
https://doi.org/10.1155/2020/6105872

Review Article
An Introduction to Complex Systems Science and Its Applications

1,2
Alexander F. Siegenfeld and Yaneer Bar-Yam2
1
Department of Physics, Massachusetts Institute of Technology, Cambridge, MA, USA
2
New England Complex Systems Institute, Cambridge, MA, USA

Correspondence should be addressed to Alexander F. Siegenfeld; asiegenfeld@gmail.com

Received 11 February 2020; Accepted 22 May 2020; Published 27 July 2020

Academic Editor: Carlos Gershenson

Copyright © 2020 Alexander F. Siegenfeld and Yaneer Bar-Yam. This is an open access article distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the
original work is properly cited.
The standard assumptions that underlie many conceptual and quantitative frameworks do not hold for many complex physical,
biological, and social systems. Complex systems science clarifies when and why such assumptions fail and provides alternative
frameworks for understanding the properties of complex systems. This review introduces some of the basic principles of complex
systems science, including complexity profiles, the tradeoff between efficiency and adaptability, the necessity of matching the complexity
of systems to that of their environments, multiscale analysis, and evolutionary processes. Our focus is on the general properties of
systems as opposed to the modeling of specific dynamics; rather than provide a comprehensive review, we pedagogically describe a
conceptual and analytic approach for understanding and interacting with the complex systems of our world. This paper assumes only a
high school mathematical and scientific background so that it may be accessible to academics in all fields, decision-makers in industry,
government, and philanthropy, and anyone who is interested in systems and society.

1. Introduction complexity of systems to that of their environments. Section

3 considers the analysis of complex systems, attending to the
How can we scientifically approach the study of complex oft-neglected question of when standard assumptions do
systems—physical, biological, and social? Empirical studies, and—more importantly—do not apply. Section 4 discusses
while useful, are by themselves insufficient, since all ex- principles for effectively intervening in complex systems
periments require a theoretical framework in which they can given that their full descriptions are often beyond the limits
be interpreted. While many such frameworks exist for of human comprehension. Section 5 provides further
understanding particular components or aspects of systems, reading. Section 6 concludes the work.
the standard assumptions that underlie most quantitative
studies often do not hold for systems as a whole, resulting in
a mischaracterization of the causes and consequences of 2. Basic Principles of Complex Systems Science
large-scale behavior.
This paper provides an introduction to complex systems 2.1. Why Complex Systems Science? Complex systems
science, demonstrating a few of its applications and its science considers systems with many components. These
capacity to help us make more effective decisions in the systems could be physical, biological, or social. Given this
complex systems of our world. It focuses on some general diversity of systems, it may seem strange to study them all
properties of complex systems, rather than on the modeling under one framework. But while most scientific disciplines
of specific dynamics as in the subfields of dynamical systems, tend to focus on the components themselves, complex sys-
agent-based modeling and cellular automata, network sci- tems science focuses on how the components within a system
ence, and chaos theory. Section 2 introduces key concepts, are related to one another [1]. For instance, while most ac-
including complexity profiles, the tradeoff between efficiency ademic disciplines would group the systems in Figure 1 by
and adaptability, and the necessity of matching the column, complex systems science groups them by row.
2 Complexity

Examples of Behaviors

Gas Pond Crowds

life
Random

Cannonball Infections Armies

Coherent

Snowflake Humans Corporations

Correlated

Physical Biological Social

Figure 1: Each column contains three examples of systems consisting of the same components (from left to right: molecules, cells, and
people) but with different relations between them. Each row contains systems representing a certain kind of relationship between
components. For random systems, the behavior of each component is independent from the behavior of all other components. For coherent
systems, all components exhibit the same behavior; for example, the behavior (location, orientation, and velocity) of one part of the
cannonball completely determines the behavior of the other parts. Correlated systems lie between these two extremes, such that the
behaviors of the system’s components do depend on one another, but not so strongly that every component acts in the same way; for
example, the shape of one part of a snowflake is correlated with but does not completely determine the shape of the other parts. Implicit in
these descriptions is the necessity of specifying the set of behaviors under consideration, as discussed in Section 2.2. (Image source: [2]).

Systems may differ from each other not because of system components [3–7]. Other examples of self-organized
differences in their parts but because of differences in how behaviors include the spontaneous formation of conversa-
these parts depend on and affect one another. For example, tion groups at a party, the allocation of goods in a decen-
steam and ice are composed of identical water molecules but, tralized economy, the evolution of ecosystems, and the
due to differences in the interactions between the molecules, flocking of birds. Such large-scale behaviors and patterns
have very different properties. Conversely, all gases share cannot be determined by examining each system part in
many behaviors in common despite differences in their isolation. By instead considering general properties of sys-
constituent molecules. The same holds for solids and liquids. tems as wholes, complex systems science provides an in-
The behaviors that distinguish solids from liquids from gases terdisciplinary scientific framework that allows for the
are examples of emergence: they cannot be determined from discovery of new ideas, applications, and connections.
a system’s parts individually. Fluid turbulence, as one might A full description of all the small-scale details of even
observe in a flowing river, is an example of how the rela- relatively simple systems is impossible; therefore, sound
tionships between parts can give rise to emergent large-scale analyses must describe only those properties of systems that
behaviors and patterns that are self-organized, meaning that do not depend on all these details. That such properties exist
they arise not from some external or centralized control but is due to universality, a phenomenon that will be discussed in
rather autonomously from the interactions between the Section 3. Statistical physics provides an underlying insight
Complexity 3

that allows for the discovery of such properties: namely, that system, we think of the behaviors arising from the ordered
while attempting to characterize the behavior of a particular arrangement of molecules in a human, not the behaviors
state of a system (e.g., a gas) may be entirely intractable, arising from the maximally disordered arrangement of
characterizing the set of all possible states of the system may molecules in a gas. It therefore may be tempting to conclude
not only be tractable but may also provide us with a model of that complex systems are those with reduced disorder. But
the relevant information (e.g., the pressure, temperature, the systems with the least disorder are those in which all
density, and compressibility). In other words, taking a step components exhibit the same behavior (coherent systems in
back and considering the space of possible behaviors provides Figure 1), and such behavior is easy to describe and thus not
a powerful analytical lens that can be applied not only to intuitively complex.
physical systems but also to biological and social ones. To resolve this apparent paradox, we must consider that
the length of a system’s description depends on the level of
detail used to describe it. Thus, complexity depends on scale.
2.2. What Is Complexity? We define the complexity of a
On a microscopic scale, it really is more difficult to describe
behavior as equal to the length of its description. The length
the positions and velocities of all the molecules of the gas
of a description of a particular system’s behavior depends
than it is to do the same for all the molecules of the human.
on the number of possible behaviors that system could
But at the scale of human perception, the behaviors of a gas
exhibit [8]. For example, a light bulb that has two possible
are determined by its temperature and pressure, while the
states—either on or off—can be described by a single bit: 0
behaviors of a human remain quite complex. Entropy
or 1. Two bits can describe four different behaviors (00, 01,
corresponds to the amount of complexity at the smallest
10, or 11), three bits can describe eight behaviors, and so
scale, but characterizing a system requires understanding its
on. Mathematically, we can write C � log2 N, where C is
complexity across multiple scales. A system’s complexity
the complexity of a system and N is its number of possible
profile is a plot of the system’s complexity as a function of
behaviors (technically, log2 N is actually an upper bound
scale [9]. In the examples below, scale will be taken to be
for the system’s complexity since if some behaviors are
length, but fundamentally, the scale of a behavior is equal to
more likely than others, the average length of the system’s
the number of coordinated components involved in the
description can be reduced by using shorter descriptions
behavior, for which physical length is a proxy. A gas is very
for the more common behaviors and longer descriptions
simple at the scale of human perception because at this scale,
for the less common ones—lossless compression algo-
only behaviors involving trillions of molecules are relevant,
rithms rely on this logic), but for our purposes here, it is
and there are relatively few distinguishable behaviors of a gas
sufficient to state that the greater the number of possible
involving so many molecules.
behaviors, the greater the complexity.
As shown in Figure 2, random, coherent, and correlated
It is important to note that one must carefully define the
systems (see Figure 1) have qualitatively different complexity
space of possible behaviors. For instance, if we are interested
profiles. Random systems have the most complexity at the
in a light bulb already in a socket, the light bulb has two
smallest scale (finest granularity/most detail), but the
possible behaviors, as above, but if we are instead interested
amount of complexity rapidly drops off as the scale is in-
in the complexity of building a light bulb, the space of
creased and the random behaviors of the individual com-
possible behaviors might include all of the ways in which its
ponents are averaged out. A coherent system has the same
parts could be arranged. As another example, consider
amount of complexity at small scales as it does at larger
programming a computer to correctly answer a multiple-
scales because describing the overall behavior of the system
choice question with four choices. At first glance, this task is
(e.g., the position and velocity of a cannonball) also describes
very simple: since there are four possible behaviors, only two
the behavior of all the components (e.g., the positions and
bits are required. Nonetheless, we have the sense that
velocities of all the atoms). Note that complexity tends to
programming a computer to score perfectly on a multiple-
increase (or remain the same) as the scale decreases, since
choice test would be quite difficult. This apparent paradox is
looking at a system in more detail (while still including the
resolved, however, when we recognize that such a task is
whole system in the description) tends to yield more in-
difficult only because we do not a priori know what questions
formation. For a correlated system, various behaviors occur
will be on the test, and thus, the true task is to be able to
at various scales, and so the complexity gradually increases
correctly answer any multiple-choice question. This task is
as one examines the system in greater and greater detail. For
quite complex, given the large number of possible ways the
instance, from very far away, a human, being barely visible,
program could respond to a string of arbitrary multiple-
has very little complexity. As the level of detail is gradually
choice questions.
increased, the description will first include the overall po-
sition and velocity of the human and then the positions and
2.3. Complexity and Scale. Consider a human, and then velocities of each limb, followed by the movement of hands,
consider a gas containing the very same molecules that are in fingers, facial expressions, as well as words that the human
the human but in no particular arrangement. Which system may be saying. Continuing to greater levels of detail, the
is more complex? The gas possesses a greater number of organs and then tissues and patterns within the human brain
possible arrangements of the molecules (i.e., has more en- become relevant, and eventually so do the individual cells. At
tropy, or disorder) and thus would take longer to describe at scales smaller than that of a cell, complexity further increases
a microscopic level. However, when we think of a complex as one sees organelles (cellular substructures), followed by
4 Complexity

Random Many types of goods but

only a few of each type
Complexity

Complexity
Correlated

Coherent Many copies of a few types of goods

Scale
Scale
Figure 2: Representative complexity profiles for random, coherent, Figure 3: The complexity profile of a factory that can produce a large
and correlated systems (see Figure 1). Any given system may have number of copies of a few types of goods, and the complexity profile
aspects of each at various scales. of a factory that can produce many types of goods but not in large
numbers. The number of copies of a good produced is a proxy for
large molecules such as proteins and DNA, and then scale since, given a fixed technology, mass production requires
eventually smaller molecules and individual atoms. At each larger-scale coordinated action in the factory (e.g., an assembly line).
The number of different types of goods that can be produced at a
level, the length of the description grows longer. This in-
given scale is a proxy for the number of different possible behaviors
credible multiscale structure with gradually increasing of the factory—and thus its complexity—at that scale.
complexity is a defining characteristic of complex systems.

2.4. Tradeoffs between Complexity and Scale. The intuition many different types of products, but none at scale. Of
that complex systems require order is not unfounded: for course, a factory may be able to increase both the complexity
there to be complexity at larger scales, there must be be- and scale of its production by adding new machinery or
haviors involving the coordination of many smaller-scale more workers; the precise tradeoff between complexity and
components. This coordination suppresses complexity at scale applies only when considering a fixed set of compo-
smaller scales because the behaviors of the smaller-scale nents with a fixed set of individual behaviors. A subtle point
components are now limited by the interdependencies be- to be made here is that introducing interactions between two
tween them. The tension between small-scale and large-scale parts of a system may in some cases increase the set of
complexity can be made precise: given a fixed set of com- relevant individual behaviors of each part, thereby in-
ponents with a fixed set of potential individual behaviors, the creasing the total area under the complexity profile. For
area under the complexity profile will be constant, regardless example, if two people enter into communication with each
of the interdependencies (or lack thereof ) between the other, the communication itself (e.g., speech) may now be a
components. More precisely, the sum of a system’s com- relevant behavior of each individual person that was not
plexity at each scale (i.e., the area under its complexity there before.
profile) will equal the sum of each individual component’s A corollary of the tradeoff between complexity and scale
complexity [9]. Thus, for any system, there is a fundamental is the tradeoff between adaptability and efficiency [10–15].
tradeoff between the number of behaviors a system can have Adaptability arises when there are many possible actions
and the scale of those behaviors. happening in parallel that are mostly independent from one
For instance, consider a factory consisting of many another, i.e., when the system has high complexity. Effi-
workers [2]. The output of the factory can be characterized ciency, on the other hand, arises when many parts of a
using a complexity profile (Figure 3). The number of dif- system are all working in concert, so that the system can
ferent types of goods that the factory can produce at a given perform the task for which it was designed at the largest
scale is a proxy for the factory’s complexity at that scale, with possible scale. Due to the tradeoff between complexity and
the number of copies of the same type of good that the scale, a system with more adaptability will have a complexity
factory can produce in a given amount of time being a proxy profile with greater complexity but predominantly at smaller
for scale. The fundamental tradeoff is evident in the fact that scales, while a system with more efficiency will have a
if the factory wants to be able to churn out many copies of a complexity profile with lower complexity but extending to
single type of good in a short amount of time, it will have to larger scales. Thus, a very efficient system will, due to its
coordinate all of its workers (perhaps having them work on necessarily lower complexity, not be as adaptable to un-
an assembly line), thereby reducing their individual freedom foreseen variations within itself or its environment, while a
to make many different kinds of goods. The factory’s pro- very adaptable system, designed to handle all sorts of shocks,
duction would then have low complexity but at a large scale will necessarily have to sacrifice some larger-scale behaviors.
(e.g., churning out many identical Model-T Fords—“Any The Soviets thought they could have their cake and eat it, too:
customer can have a car painted any color that he wants so they originally believed that their economy would outper-
long as it is black”). On the other hand, if the factory’s form capitalist ones because capitalist economies have so
employees work independently, they will be able to create much waste related to multiple businesses competing to do
Complexity 5

the same thing (Chapter 16 in [16]). It would be far more As another example, healthcare involves both small-scale
efficient to coordinate all economic production. But in tasks with high overall complexity such as case management
creating such large-scale economic structures, lower-scale and large-scale, lower-complexity tasks such as manufacturing
complexity was sacrificed, resulting in a nonadaptive system. and delivering vaccines [24]. Vaccinations are lower com-
Improperly regulated capitalist systems may also sacrifice plexity but larger scale because essentially the same actions are
redundancy and adaptability for efficiency, resulting in, for performed for nearly every patient. Large-scale top-down
instance, excessive concentrations of market power, harmful organizations and initiatives are suited for large-scale, lower-
feedback loops, and herd-like behaviors [17–21]. complexity tasks, but tasks like case management require
Due to the tradeoff between complexity and scale, any health systems with a high degree of small-scale (i.e., local)
mechanism that creates larger-scale complexity—whether complexity.
market or government or otherwise—will necessarily reduce The eurozone provides a potential illustration of a
individual complexity. This is not to say that larger-scale multiscale complexity mismatch. Fiscal policy is made
complexity is always harmful; it is often worth trading some predominantly at the scale of individual countries and thus
individual-level freedoms for larger-scale cooperation. has a higher complexity at the country scale but relatively
When, then, is complexity at a particular scale desirable? little complexity at the scale of the entire eurozone, while
monetary policy is made at the scale of the entire eurozone
and thus has some complexity at the scale of the eurozone
2.5. Why Be Complex? A determination of when complexity but lacks the ability to vary (i.e., lacks complexity) at the scale
is desirable is provided by the Law of Requisite Variety [22]: of individual countries. Many have argued that economic
to be effective, a system must be at least as complex as the difficulties within the eurozone have arisen because this
environmental behaviors to which it must differentially mismatch has precluded effective interactions between fiscal
react. If a system must be able to provide a different response and monetary policy [25–29].
to each of 100 environmental possibilities and the system has Problems arise not from too much or too little com-
only 10 possible actions, the system will not be effective. At plexity (at any scale) per se but rather from mismatches
the very least, the system would need 100 possible actions, between the complexities of a task to be performed and the
one for each scenario it could encounter. (The above con- complexities of the system performing that task. (Inciden-
dition is necessary but of course not sufficient; a system with tally, human emotions appear to reflect this principle: we are
sufficiently many actions may still not take the right actions bored when our environment is too simple and over-
in the right circumstances.) Note that the environment to whelmed when it is too complex [30].) Note that the system
which a system must react is itself also a system and will in one scenario may be the task/environment in another; for
sometimes be referred to as such. instance, the same complexity that helps a system interact
Since complexity is defined only with respect to a par- with its environment may prevent its effective management
ticular scale, we can refine the Law of Requisite Variety: to be by other systems. In none of the above examples have the
effective, a system must match (or exceed) the complexity of complexity profiles been precisely calculated, nor have scales
the environmental behaviors to which it must differentially been precisely defined. Instead, proxies for scale are used and
react at all scales for which these behaviors occur [9]. To estimated comparisons of complexity made. Such an ap-
illustrate this multiscale version of the Law of Requisite proach cannot yield precise results (indeed, no approach
Variety, we consider military conflict [23] (see Figure 4). can, given the complexity a full description of such systems
Here, one military can be considered as the system, while the would require), but additional precision is not needed when
other military is part of the environment with which the even the approximate analysis reveals large mismatches in
system must interact. For two militaries of equal complexity, complexity. (To remedy the diagnosed mismatches, more
i.e., with the same number of behaviors, but with one detailed analyses may be required.) While it may be
military operating at a larger scale (e.g., two very tightly tempting to attribute the problems arising from a complexity
controlled armies, but with one army larger than the other), mismatch to particular proximate causes and chains of
the larger-scale military will likely win. For two militaries of events, problems of one form or another will be inevitable
equal scale but unequal complexity (e.g., two equally sized unless the underlying mismatch is addressed.
and equally powered fleets, but with one being more ma-
neuverable than the other), the higher-complexity military
will likely win, since the high-complexity military has an 2.6. Subdivided Systems. Even if the complexity of the
action for every action of the lower-complexity military but system matches that of its environment at the appropriate
not vice versa. When a military with high complexity at a scales, there is still the possibility of a complexity mismatch.
smaller scale (e.g., a guerrilla force) conflicts with a military Consider two pairs of friends—four people total, each of
with larger-scale behavior but lower complexity (e.g., the US whom can lift 100 pounds—and consider two 200-pound
army in Vietnam or the Soviet army in Afghanistan), the couches that need to be moved. Furthermore, assume that
terrain, which constrains the scale of the conflict, plays an each person is able to coordinate with her friend but not with
important role. In an open field, or in open waters, the either of the other two people. Overall then, the system of
military that has more complexity at the larger scales is people has sufficient complexity at the appropriate scales to
favored, while in the jungle or in the mountains, higher move both couches since each pair of friends can lift one of
complexity at smaller scales is favored. the 200-pound couches. However, were one person from
6 Complexity

Higher complexity
Complexity

Complexity
Larger scale

Scale Scale

(a) (b)

Higher complexity
at smaller scales
Complexity

Higher complexity
at larger scales

Scale

(c)

Figure 4: Schematic complexity profiles of militaries in conflict. (a) If two armies are operating with the same number of possible behaviors
but at different scales, the larger-scale one is favored. (b) If two armies are operating at the same scale but with different numbers of possible
behaviors, the higher-complexity one is favored. (c) If two armies are operating at different scales and with different numbers of possible
behaviors, which one is favored depends on the terrain (see text). Note that these profiles are simplified to highlight the key concepts; actual
militaries operate at multiple scales. More generally, (a) and (b) depict conflicts in which one army has at least as much complexity as the
other at every scale.

each pair of friends to be assigned to each couch, they would with each subdivision of problems requiring particular types
not be able to lift the couches because the two people lifting of coordinated knowledge and effort in order to be solved.
each couch would not belong to the same pair of friends and Academia’s complexity across multiple scales allows it to
thus would not be able to coordinate their actions. The effectively work on many of these problems. However, there
problem here is that while the pairs of friends possess may exist problems that academia, despite having sufficient
enough overall complexity at the right scales to lift the overall multiscale complexity, is nonetheless unable to solve
couches, the subdivision within the system of friends is not because the subdivisions within the problem do not match
matched to the natural subdivision within the system of the subdivisions within academia. The increase in inter-
couches. The mismatch in complexity can be seen if we focus disciplinary centers and initiatives over the past few decades
our attention on just a single couch: while the couch requires suggests the perception of such a mismatch; however, the
coordinated action at the scale of 200 pounds, the two people structure of the academic system as a whole may still hinder
lifting it are capable only of two independent actions, each at progress on problems that do not fall neatly within a dis-
the scale of 100 pounds. cipline or subdiscipline [31–36].
The way in which academic departments are organized The above examples provide an illustration of the
provides a more realistic example of the potential of sub- principle that in order for a system to differentially react to a
division mismatch. Academia has multiple levels of subdi- certain set of behaviors in its environment, not only must the
vision (departments, subfields, etc.) in order to organize system as a whole have at least as much complexity at all
knowledge and coordinate people, resulting in a high overall scales as this set of environmental behaviors (as described in
degree of complexity across multiple scales, where scale Section 2.5.), but also each subset of the system must have at
could refer to either the number of coordinated people or the least as much complexity at all scales as the environmental
amount of coordinated knowledge, depending on which behaviors corresponding to that subset. A good rule of
aspect of the academic system is under consideration. thumb for applying this principle is that decisions con-
Similarly, there are multiple levels of natural subdivision in cerning independent parts or aspects of a system should be
the set of problems that academia can potentially address, able to be made independently, while decisions concerning
Complexity 7

dependent parts of the system should be made dependently.

It follows that the organizations that make such decisions
should be subdivided accordingly, so that their subdivisions Lower levels given
some autonomy
match the natural divisions in the systems with which they
interact. The subdivisions present in the human brain and

Complexity
the analysis of subdivisions in neural networks more gen-
erally (Chapters 2.4-2.5 in [1]) demonstrate how systems
that are subdivided so as to match the natural subdivisions in
Tightly controlled
their environments outperform those with more internal
connectivity.

Scale
2.7. Hierarchies. A common way in which systems are or- Figure 5: Complexity profiles of two hierarchies, each with the
ganized is through hierarchies. In an idealized hierarchy, same number of people. Here, the scale is the number of coor-
there are no lateral connections: any decision that involves dinated man-hours. In one hierarchy, all decisions, regardless of the
multiple components of the hierarchy must pass through a scale, are made by a single person, while in the other, different
common node under whose control these components all decisions are made at various levels of the hierarchy.
(directly or indirectly) lie. The complexity profile of such a
hierarchy depends on the rigidity of the control structure
(Figure 5). At one extreme, every decision, no matter how likewise, perhaps some of the powers that some argue should
large or small, is made by those at the top of the hierarchy. be devolved from the federal to the state level should in fact
This hierarchy has the same amount of complexity across all be devolved to the local level).
its scales: namely, the complexity of whatever decisions are It is important to distinguish between the complexity of a
being made at the top. At the other extreme, there is no hierarchy and the complexity of the decisions that the people
communication within the hierarchy, and every individual within the hierarchy are capable of making. For instance,
acts independently. This hierarchy has very little complexity one could design a tightly controlled hierarchy that could
beyond the individual level. Between these two extremes is a take a large number of large-scale actions (i.e., high com-
typical hierarchy, in which different decisions are made at plexity at its largest scale), but since the decision-making
different levels. abilities of even the most capable humans are of finite
No type of hierarchy is inherently better than any other. complexity, the individuals at the top may be fundamentally
For a particular environment, the best hierarchy is one for unable to correctly choose from among these actions. This
which the complexity profile matches that of the tasks brings us to an important limitation of hierarchies: the
needed to be performed. A tightly controlled (top-heavy) complexity of the decisions concerning the largest-scale
hierarchy is not well suited to environments in which there is behaviors of a hierarchy—the behaviors involving the entire
a lot of variation in the systems with which the lower levels of organization—is limited by the complexity of the group of
the hierarchy must interact; neither is a very loosely con- people at the top [2]. Thus, a hierarchy will necessarily fail
trolled hierarchy well suited to environments that require when the complexity of matching its largest-scale behaviors
large-scale coordinated action. For example, centralizing too to those of its environment is higher than the complexity of
much power within the US governance system at the federal decision-making that is achievable by any individual or
(as opposed to the local or state) level would not allow for committee. (Note that the complexity of deciding in which
sufficient smaller-scale complexity to match the variation behaviors of a system should correspond to which behaviors
among locales; too decentralized a governance system would of its environment is generally much greater than the
not allow for sufficient larger-scale complexity to engage complexity of either the system or the environment alone:
with problems that require nationally coordinated re- for example, if both the system and environment have 10
sponses. Assigning decisions to higher levels in hierarchies possible behaviors, the system has enough complexity to
allows for more efficiency and scale but less adaptability and match the environment, but properly deciding which be-
variation. haviors of the system should correspond to which envi-
We should also consider not just the overall complexity ronmental conditions requires correctly choosing one
profile of governance systems but how well the subdivisions option out of a space of 10 factorial or 3,628,800 possibil-
in governance systems match those within their territories ities.) The failure of command economies provides a stark
(Section 2.6.). Metropolitan areas are in some ways more example: the allocation of resources and labor is too complex
similar to one another than they are to the rural areas of their a problem for any one person or group of people to un-
respective states. So while dividing the US into 50 states derstand. Markets allocate resources via a more networked
provides substantial lower-scale governmental complexity, system: decisions regarding how to allocate resources are
this complexity is not necessarily well matched to natural made without any individual making them, just as decisions
urban-rural divides. To the extent that such a mismatch are made in the human brain without any neuron making
exists, there may be issues currently handled at the state level them. (Whether or not these market allocations are desirable
that would be better handled at the local level, thereby depends in part on the way in which the market is structured
allowing for different policies in urban and rural areas (and and regulated.)
8 Complexity

We began by considering idealized hierarchies with only 3.2. When Mean-Field Theory Breaks Down. The systems for
vertical connections, but lateral connections provide another which mean-field theory applies exhibit large-scale be-
mechanism for enabling larger-scale behaviors. For instance, haviors that are the average of the behaviors of their
cities can interact with one another (rather than interacting components. They must possess a separation of scales,
only with their state and national governments) in order to which arises when the statistical fluctuations of their
copy good policies and learn from each other’s mistakes. components are sufficiently independent from one an-
Through these sorts of evolutionary processes (described other above a certain scale. Mean-field theory may hold
further in Section 4), large-scale decisions (large-scale because even in the presence of strong interactions, so long as the
policies may be copied by multiple cities) that are more effect of those strong interactions can be captured by the
complex than any individual component can be made. Such average behavior of the system—that is, so long as each
lateral connections can exist within a hierarchical framework in component of the system can be modeled as if it were
which the top of the hierarchy (in this example, the national interacting with the average (i.e., mean field) of the sys-
government) maintains significant control, or they can exist tem. For example, the large-scale motion of solids is well
outside of a strictly hierarchical structure, as in the human described by mean-field theory, even though the mole-
brain. Furthermore, these lateral connections can vary in cules in a solid interact with one another quite strongly,
strength. Overly strong connections lead to herd-like behaviors because the main effect of these interactions is to keep
with insufficient smaller-scale variation, such as groupthink each molecule at a certain distance and orientation from
[37–39] (no system is exempt from the tradeoff described in the average location (center of mass) of the solid. Like-
Section 2.4.), while overly weak connections result in mostly wise, under some (but certainly not all) conditions,
independent behavior with little coordination. economic markets can be effectively described by mod-
eling each market actor as interacting with the aggregate
3. Analyzing Complex Systems forces of supply and demand rather than with other in-
dividual market actors.
The previous section has examined some of the general However, when there are sufficiently strong correla-
properties of systems with many components. But how do tions between the components of the system, i.e., when
we study particular systems? How do we analyze data from the interactions between a component of the system and a
complex systems, and how do we choose which data to specific set of other components (as opposed to its general
analyze? interaction with the rest of the system) cannot be
neglected, mean-field theory will break down. These
systems will instead exhibit large-scale behaviors that
3.1. How Do We Understand Any System? In a sense, it is arise not solely from the properties of individual com-
surprising that we can understand any macroscopic system ponents but also from the relationships between com-
at all, as even a very simple mechanical system has trillions ponents. For example, while the behavior of a muscle can
upon trillions of molecules. We are able to understand such be roughly understood from the behavior of an individual
systems because they possess a separation of scales [40], muscle cell, the behavior of the human brain is funda-
meaning that the macroscopic behavior we are interested in mentally different from that of individual neurons, be-
occurs at a far larger scale than the behavior of the individual cause cognitive behaviors are determined largely by
molecules, with not much behavior occurring in between variations in the synapses between neurons. Similarly, the
these two scales (see Figure 6). This separation allows us to complex ecological behaviors of a forest cannot be de-
treat the macroscopic and microscopic behaviors separately: termined by the behaviors of its constituent organisms in
for mechanical systems, we treat the macroscopic behavior isolation.
explicitly with Newtonian mechanics, while the microscopic Because their small-scale random occurrences are not
behavior is considered in aggregate using thermodynamics. statistically independent, complex systems often exhibit
More generally, the approach described above is an large-scale fluctuations not predicted by mean-field theory,
example of a mean-field theory [41], in which the average such as forest fires, viral content on social media, and crashes
behaviors of a system’s components are explicitly modeled in economic markets. Sometimes, these large-scale fluctu-
and the deviations of the individual components from this ations are adaptive: they enable a system to collectively
average are treated as statistically independent random respond to small inputs [42]. For instance, humans respond
fluctuations. This approach works very well for systems such strongly to minor disturbances in the density of air, such as
as computers, cars, airplanes, and buildings, in which the the sound of their own names. However, some large-scale
motions of individual molecules are—apart from some fluctuations pose systemic risks.
mostly uncorrelated fluctuations—well described by the
motion of the piece of material to which they belong. Mean-
field assumptions are also often employed in analyses of 3.3. Fat-Tailed Distributions and Systemic Risk. When the
biological, social, and economic systems; these assumptions components of a system are independent from one another
work well in many cases, but, as we will see, they are not above a certain scale, then at much larger scales, the mag-
always appropriate for complex systems. It is important, nitudes of the fluctuations of the system follow a normal
therefore, to determine under what conditions mean-field distribution (bell curve), for which the mean and standard
theory holds. deviation are well defined and for which events many
Complexity 9

1.2

1.0

Probability density
Thin-tailed
Complexity

0.8

0.6

0.4

0.2
Fat-tailed
s0 Scale
0 2 4 6 8
Figure 6: A complexity profile of a system with a separation of Scale
scales. A separation of scales implies that the behaviors occurring Figure 7: A normal distribution (thin-tailed) and a distribution
below a certain scale (s0 in the above figure) are at larger scales with a power-law decay (fat-tailed). The fat-tailed distribution may
mostly independent from one another, and that therefore, at these appear more stable, due to the lower probability of small-scale
larger scales, only the average effects of the small-scale behaviors fluctuations and the fact that samples from the distribution may not
are relevant. contain any extreme events. However, sooner or later, a fat-tailed
distribution will produce an extreme event, while one could wait
standard deviations above the mean are astronomically im- thousands of lifetimes of the universe before a normal distribution
probable. Interdependencies, however, can lead to a distri- produces a similarly extreme event. Note that the axes of this graph
are truncated; the illustrated fat-tailed distribution can, with small
bution of fluctuations in which the probability of an extreme
but nonnegligible probability (0.04%), produce events with a scale
event, while still small, is not astronomically so. Such dis- of one million or more.
tributions are characterized as fat-tailed—see Figure 7. For
example, while human height follows a thin-tailed distribu-
tion, with no record of anyone over twice as tall as the average 3.4. Understanding Complex Systems. Because it is usually
human, human wealth—due to the complex economic in- easier to collect data regarding components of a system
teractions between individuals—follows a fat-tailed distri- than it is to collect data regarding interactions between
bution, with multiple individuals deviating from the average components, studies often fail to capture the information
by factors of more than one million [43]. relevant to complex systems, since complex large-scale
One danger of interdependencies is that they may make behaviors critically depend on such interactions. Fur-
systems appear more stable in the short term by reducing the thermore, as discussed in Section 3.3., data analysis can
extent of small-scale fluctuations, while actually increasing the severely underestimate the probability of extreme events
probability of catastrophic failure [44–47]. This danger is (tail risk). Finally, analyses often (implicitly) assume lin-
compounded by the fact that when underlying probability earity, i.e., they assume that the total impact of a set of
distributions have fat tails (a situation made more likely by factors is equal to the sum of the impacts of each individual
interdependencies), standard statistical methods often break factor, an assumption that often breaks down for complex
down, leading to potentially severe underestimates of the systems, which may possess feedback loops, abrupt tran-
probabilities of extreme events [48]. As a thought experiment, sitions (tipping points), and other highly nonlinear be-
imagine 100 ladders, each with a 1/10 probability of falling. If haviors [57–64].
the ladders are independent from one another, the probability How can we understand the systems for which these
that all of them fall is astronomically low (literally so: there is standard approaches do not apply? Our understanding of
about a 1020 times higher chance of randomly selecting a all systems with many components depends on universality
particular atom out of all of the atoms in the known universe). [65], i.e., the existence of large-scale behaviors that do not
If we tie all the ladders together, we will have made them safer, depend on the microscopic details. The standard ap-
in the sense that the probability of any individual ladder proaches are predicated on the assumption of sufficient
falling will be much smaller, but we will have also created a independence between components, which allows large-
nonnegligible chance that all of the ladders might fall down scale behaviors to be determined without a full accounting
together. Other examples include the interconnectedness of of the system’s details via mean-field theory and/or normal
our financial systems resulting in the possibility of global distributions. But mean-field theory is just one example of
market crashes [49–54] and the interconnectedness of travel universality.
routes increasing the probability of pandemics such as the Sound is another example: all materials, regardless of their
Spanish flu and COVID-19 [55, 56]. When such crises do composition, allow for the propagation of sound waves. Sound
occur, they are often attributed to proximate causes or chains behaves so similarly in all materials because at the length scales
of events, and measures are then implemented to ensure that relevant to sound waves, which are far larger than the sizes of
those particular chains of events will not occur again. But individual atoms and molecules, the effect of the microscopic
unless the underlying systemic instabilities are addressed, parameters is merely to set the speed of the sound. Note that
another crisis is bound to happen sooner or later, even if its sound waves cannot be understood as a property of the average
precise form cannot be predicted. behavior—in this case, average density—of a material, since it is
10 Complexity

precisely the systematic correlations in the deviations from that elements that were not considered ahead of time. It should
average that give rise to sound. Nor is sound best understood also be noted that in a functional system with a high degree
by focusing on the small-scale details of atomic motion: sci- of complexity, the potential positive impact of a change is
entists understood sound even before they learned what atoms generally much smaller than its potential negative impact.
are. The key to understanding sound waves is to recognize that For example, a small change to the wiring in a computer is
they have a multiscale structure—with larger-scale fluctuations unlikely to dramatically improve the computer’s perfor-
corresponding to lower frequencies and smaller-scale fluctu- mance, but it could cause the computer to crash. Airplanes
ations corresponding to higher frequencies—and to model are another example. This phenomenon is a consequence of
them accordingly. the fact that, by definition, a high degree of complexity
Lim et al. apply this approach to studying ethnic violence implies that there are many system configurations that will
[66]. They built a predictive model to analyze where ethnic not work for every one configuration that will.
violence has the potential to occur and applied their model Given the absence of perfect knowledge, how can the
to India and to what was Yugoslavia. Ethnic violence has success of systems we design or are part of be assured?
many causes, but rather than focusing on specific, culturally While the success of many systems rests on the assumption
dependent mechanisms or on the average properties of that good decisions will be made, some systems do not
regions, such as demographic or economic statistics, the depend on individual understanding and can perform well
authors instead considered the multiscale patterns in how in spite of the fallibility of decision-makers (whether due to
ethnic groups were geographically distributed (Figure 8). corruption, subconscious bias, or the fundamental limi-
They found that ethnic violence did not occur when the tations of human minds). The study of complex systems
ethnic groups were either well mixed or well separated but approaches this observation scientifically by (implicitly or
rather occurred only when ethnic groups separated into explicitly) considering the decision-makers themselves as
geographic patches (this separation falls into the same part of the system and of limited complexity/decision-
universality class as the separation of oil and water), with the making ability. The question thus becomes: how do we
violence most likely to occur for geographic patches of a design systems that exceed the complexity of the decision-
particular size. This analysis implies that ethnic violence can makers within them?
be prevented by the use of well-placed political boundaries,
as in Switzerland [67]. Although not explicitly included in
the analysis, specific details of a region are relevant insofar as 4.1. Evolutionary Processes. While uncertainty makes most
they are either a cause or an effect (or both) of the patch systems weaker, some systems benefit from uncertainty and
size—for instance, animosity between two ethnic groups, variability [69–72]. The common characteristic of these
though not explicitly considered, may be a cause as well as a systems is their embodiment of some sort of evolutionary
consequence of geographic segregation [68]. process, i.e., a process in which successful changes are copied
Understanding all the details of any complex system is (and further modified) while unsuccessful changes are not.
impossible, just as it is for most systems with a separation of The classic evolutionary processes are biological: due to
scales; there is just too much complexity at the smallest scale. variability introduced by random mutations, organisms with
However, unlike the behaviors of systems with a separation the complexity and scale of humans evolved from single-
of scales, the important large-scale behaviors of complex celled organisms. Furthermore, humans themselves have the
systems are not simply the average of their small-scale be- property of benefiting from exposure to random shocks
haviors. The interdependencies at multiple scales can make it (provided the shocks are not too strong). Immune system
difficult or impossible to precisely understand how small- performance is improved by early exposure to nonlethal
scale behaviors give rise to larger-scale ones, but even for pathogens [73, 74]; muscles and bones are strengthened by
complex systems, there is much less complexity at the larger microtears and microfractures, respectively; we learn by
scales than there is at the smaller scales. Thus, there will exposure to new information and problem-solving; and our
always be large-scale behaviors that do not depend on most psychologies are strengthened by exposure to adversity,
of the system’s details (see Figure 9). The key to analyzing provided the adversity is not too severe [75, 76].
these behaviors is to find the appropriate mathematical (or Competitive market economies provide another exam-
conceptual) description, i.e., to identify variables that de- ple of how systems can thrive on uncertainty. Due to our
scribe the relevant space of possible (large-scale) behaviors, ignorance of which will succeed, many potential innovations
which for complex systems is neither a simple average nor a and businesses must be created and improved upon in
full account of all the details. For additional examples of this parallel, the successful ones expanding and the unsuccessful
multiscale approach, see [40]. ones failing. The successful among these can then be im-
proved upon in the same manner—with many approaches
4. Complex Systems and Uncertainty being applied at once—and so on. (However, without ef-
fectively regulated multiscale cooperative frameworks—see
Although the principles discussed throughout Sections 2 and Section 4.2.—large-scale parts of the economic system may
3 help us recognize the fundamental properties and limi- optimize for the wrong goals, settling into harmful societal
tations of systems, our understanding of most complex equilibria [77, 78].)
systems will inevitably be imperfect. And regardless of how Likewise, the internal processes of large organizations
well considered a plan is, a truly complex system will present may follow an evolutionary pattern in which small parts of
Complexity 11

Hungarian, 1.9 Other, 3.9

Montenegrin, 2.5 Serb, 36.3
Yugoslav, 5.4∗
Macedonia, 5.9
Albanian, 7.7

Slovene, 7.8

Muslim, 8.9 Croat, 19.7

∗
Self-identified as Yugoslavs.
Dispersed around country.

Albanian Muslim
Bulgarian Serb
Croat Slovak
Serbs Muslims
Hungarian Slovene Croats Albanians
Macedonian No majority
Montenegrin present
Based on opstina data from 1991 census
0 100 kilometers
0 100 miles

(a) (b)

0 1
Conflict Muslims
Serbs Albanians
Croats
(c) (d)

Figure 8: A figure from Lim et al.’s paper on ethnic violence [66]. The sites where their model predicts a potential for ethnic violence are
shown in red in (c) and (d) with confirmed reports of ethnic violence depicted by the yellow dots in (d).

the organization can fail and thus be improved upon; without In order to thrive in uncertainty and exceed the com-
such flexibility, the entire organization may fail at once in the plexity of individual decision-making, systems can incor-
face of a changing internal or external environment. In some porate evolutionary processes so that they, even if very
cases, the failure of the entire organization makes room for limited at first, will naturally improve over time. The first
more effective organizations to take its place (assuming the step is to allow for enough variation in the system, so that the
economy is sufficiently decentralized and competitive so that system can explore the space of possibilities. Since a large
the organization in question is not “too big to fail”). The amount of variation means a lot of complexity and com-
collapse of government is generally not one of those cases, plexity trades off with scale (Section 2.4.), such variation
however [79], so it is especially important that governance must occur at smaller scales (in both space and time). For
systems possess the flexibility to internally benefit from example, in the case of governance, enabling each city to
randomness and uncertainty. Perhaps counterintuitively, not experiment independently allows for many plans to be tried
allowing small failures to occur may weaken systems in the out in parallel and to be iterated upon. The opposite strategy
long run by halting evolutionary processes and by creating would be to enact one national plan, the effects of which will
interdependencies that lead to systemic risk (Section 3.3.). not be able to be comparatively evaluated.
12 Complexity

Full description may cooperate with each other to form groups, which in turn
of the system
may cooperate to form even larger groups, and so on. Thus, a
complex network of cooperation and competition among
groups of various sizes (scales) can naturally evolve.
In order to promote effective group cooperation, com-
Complexity

petition must be properly structured. A soccer team in which

the players compete with their own team members to score
Description of
the system’s
goals will not be effective, but one in which the players
most important compete for the title of the most fit may be. The framework
(largest scale) in which competition occurs must be structured so that the
behaviors
competitors are incentivized to take actions that are net good
Scale for the group; otherwise, a kind of tragedy-of-the-commons
situation occurs. The potential for competition to go awry
Figure 9: A representative complexity profile of a complex system. highlights the importance of having a multiscale structure
Understanding all the details (i.e., all of the small-scale behaviors) is
with competition occurring on multiple levels, rather than
impossible and unnecessary; the most important information is
contained in the large-scale behaviors. However, for systems for having everyone in the system compete with everyone else.
which mean-field theory does not apply, characterizing these be- With the multiscale structure, groups with unhealthy evo-
haviors will involve more than a simple average. lutionary dynamics are selected against, while groups with a
healthy mix of competition and cooperation that benefits the
entire group are selected for. There is evidence that the
The second step is to allow for a means of communi- geographic nature of evolution—in which organisms evolve
cation between various parts of the system so that successful in somewhat separated environments and mean-field theory
choices are adopted elsewhere and built upon (e.g., cities does not apply—has resulted in precisely this multiscale
copying the successful practices of other cities). Plans will structure and has therefore allowed for the evolution of
always have unintended consequences; the key is to allow genuine (e.g., not reciprocal) altruistic behavior [81, 82].
unintended consequences to work for rather than against the Likewise, market economic systems are successful not be-
system as a whole. Systems can explicitly design only systems cause free markets produce optimal outcomes (real-world
of lesser complexity since an explicit design is itself a be- markets often sharply deviate from the assumptions of free-
havior of the first system. However, systems that evolve over market models, and externalities abound) but rather be-
time can become more complex than their designers. The cause, at their best, appropriately regulated market systems
desire for direct control must therefore be relinquished in allow for multiscale evolutionary processes to naturally arise,
order to allow complexity to autonomously increase over resulting in innovations and complexity far beyond what
time. anyone could have imagined, let alone designed.

5. Further Reading
4.2. Multiscale Evolutionary Processes. Successful evolu-
tionary processes generally do not consist of unbridled Complex systems science, also known as complexity science,
competition but rather contain both competition and co- contains many subfields. One starting point for exploring
operation, each occurring at multiple scales [80]. For ex- complex systems more broadly is this clickable map [83] of
ample, cells cooperate within multicellular organisms in complex systems science and related fields. Encyclopedias
order to more effectively compete with other organisms, and [84, 85] and textbooks [1, 86–90] provide a range of per-
organisms cooperate both within and between species in spectives. In addition to the topics and references discussed
order to more effectively compete with other species. throughout this introduction, we provide a selection among
Competition at larger scales naturally breeds cooperation at the many works applying complex systems science to social
smaller scales because in order for a group to effectively systems and policy [91–105] and management [106–109].
compete with another group (large-scale competition), there Complex systems science includes, among others, the fields
must be cooperation within the group. Cooperation can also of system dynamics [110], evolutionary dynamics
breed competition since sometimes the best way for the [4, 111, 112], network science [113], fractals and scaling
group to achieve its shared goals is to facilitate some healthy [114–117], urban science [118], pattern formation [119, 120],
competition among its subgroups. Those subgroups must econophysics [121], and nonlinear dynamics and chaos
foster cooperation within themselves in order to effectively [122, 123]. Book series on complex systems topics include
compete with each other, and they too may be able to in- the Santa Fe Institute Series and Unifying Themes in Complex
crease the effectiveness of their internal cooperation by Systems.
introducing some healthy competition among their mem-
bers (Figure 10 provides an example). If these members are 6. Summary
themselves groups, the process of competition begetting
cooperation that begets more competition can continue to Systems with many components often exhibit emergent large-
even smaller scales. This process can work in reverse as well: scale behaviors that cannot be directly inferred from the
in order for individuals to compete more effectively, they behaviors of their components. However, an early insight of
Complexity 13

Competition
between sports

Team collaboration Competition between sports

enables the sport for fan attention and money
to exist and compete increases team collaboration

Collaboration Competition
between teams between teams

Collaboration of Competition between teams

players enables causes selection of teams
teams to compete with collaborating players

Collaboration Competition
between players between players
Figure 10: An illustration from Chapter 7 in [16], showing the interplay between cooperation and competition in the context of sports teams
and leagues.

statistical physics is that in spite of the impossibility of de- importance of interdependencies and the complexity that
scribing the details of trillions of molecules, the macroscopic arises from these interdependencies. To some extent, these
properties of the molecules can be well understood by ana- problems can be mitigated by matching the data analysis or
lyzing their space of possible behaviors, rather than their organizational structure to natural divisions within the
specific configurations and motions. While many macro- system of interest. Since complex systems are those for
scopic properties can be described in terms of the average which behaviors occur over multiple scales, successful or-
behaviors of the molecules, the macroscopic properties of ganizations and analyses for complex systems must also be
certain physical phenomena, such as phase transitions, cannot multiscale in nature. However, even when armed with all the
be understood by averaging over system components; ac- proper information and tools, human understanding of
cordingly, physicists were forced to develop new, multiscale most complex systems will inevitably fall short, with un-
methods. Likewise, while standard statistical methods—which predictability being the best prediction. To confront this
infer the average properties of a system’s many compo- reality, we must design systems that are robust to the ig-
nents—can successfully model some biological and social norance of their designers and that, like evolution, are
systems, they fail for others, sometimes spectacularly so. strengthened rather than weakened by unpredictability.
Taking a systemic view by considering the space of Such systems are flexible with multiple processes occurring
possible behaviors can yield insights that cannot be gleaned by in parallel; these processes may compete with one another
considering only the proximate causes and effects of par- within a multiscale cooperative framework such that ef-
ticular problems or crises. A system’s complexity—which fective practices are replicated. Only these systems—that
depends on its number of distinct potential behaviors (i.e., on grow in complexity over time from trial and error and the
the space of possibilities)—is a starting point from which to input of many—exhibit the necessary complexity to solve
get a handle on its large-scale properties, in the same way that problems that exceed the limits of human comprehension.
entropy is the starting point for statistical physics. Because the
number of distinct behaviors of a system depends on the level Conflicts of Interest
of detail (behaviors that appear the same at lower resolution
may be distinct at higher resolution), complexity depends on The authors declare that they have no conflicts of interest.
scale. Interdependencies between components reduce com-
plexity at smaller scales by restricting the freedom of indi- Acknowledgments
vidual components while creating complexity at larger scales
by enabling behaviors that involve multiple components This material is based upon work supported by the National
working together. Thus, for systems that consist of the same Science Foundation Graduate Research Fellowship Program
components, there is a fundamental tradeoff between the under Grant no. 1122374 and by the Hertz Foundation. The
number of behaviors at smaller and larger scales. This tradeoff authors thank Uyi Stewart for discussions that led to the
among scales is related to the tradeoff between a system’s writing of this paper, Gwendolyn Towers for editing early
adaptability, which depends on the variety of different re- drafts of the manuscript, and Robi Bhattacharjee for helpful
sponses it has to internal and external disturbances, and its discussions regarding complexity and scale.
efficiency, which depends on its operating scale. There is no
ideal scale at which a system should possess complexity; References
rather, the most effective systems are those that at each scale [1] Y. Bar-Yam, Dynamics of Complex Systems, Addison-Wes-
match the complexity of their environments. ley, Boston, MA, USA, 1997.
When analyzing data or creating organizational struc- [2] Y. Bar-Yam, “Complexity rising: from human beings to
tures, standard methods fail when they underestimate the human civilization, a complexity profile,” in Encyclopedia of
14 Complexity

Life Support Systems, R. B. Pimentel, R. C. Elliot, R. Holton, [22] W. R. Ashby, “Requisite variety and its implications for the
P. Lorenzano, and H. Arlt, Eds., Vol. 1, EOLSS UNESCO control of complex systems,” in Facets of Systems Science,
Publishers, Oxford, UK, 2002. G. J. Klir, Ed., pp. 405–417, Springer Science & Business
[3] P. W. Anderson, “More is different,” Science, vol. 177, Media, Berlin, Germany, 1991.
no. 4047, pp. 393–396, 1972. [23] Y. Bar-Yam, “Complexity of military conflict: multiscale
[4] S. A. Kauffman, The Origins of Order: Self-Organization and complex systems analysis of littoral warfare,” Report to
Selection in Evolution, Oxford University Press, New York, Chief of Naval Operations Strategic Studies Group, 2003.
NY, USA, 1993. [24] Y. Bar-Yam, “Improving the effectiveness of health care and
[5] M. Eigen, “Selforganization of matter and the evolution of public health: a multiscale complex systems analysis,”
biological macromolecules,” Die Naturwissenschaften, American Journal of Public Health, vol. 96, no. 3, pp. 459–
vol. 58, no. 10, pp. 465–523, 1971. 466, 2006.
[6] M. C. Cross and P. C. Hohenberg, “Pattern formation [25] M. B. Canzoneri, R. E. Cumby, and B. T. Diba, “How do
outside of equilibrium,” Reviews of Modern Physics, vol. 65, monetary and fiscal policy interact in the European mone-
no. 3, pp. 851–1112, 1993. tary union?” Tech. Rep., National Bureau of Economic
[7] H. Haken, Information and Self-Organization: A Macroscopic Research, Cambridge, MA, USA, 2005.
Approach to Complex Systems, Springer Science & Business [26] W. Semmler and W. Zhang, “Monetary and fiscal policy
Media, Berlin, Germany, 2006. interactions in the euro area,” Empirica, vol. 31, no. 2-3,
[8] T. M. Cover and J. A. Thomas, Elements of Information pp. 205–227, 2004.
Theory, John Wiley & Sons, Hoboken, NJ, USA, 2012. [27] P. Alessandrini and M. U. Fratianni, “In the absence of a
[9] B. Allen, B. Stacey, and Y. Bar-Yam, “Multiscale information fiscal union, the eurozone needs a more flexible monetary
theory and the marginal utility of information,” Entropy, policy,” PSL Quarterly Review, vol. 68, 2015.
vol. 19, no. 6, p. 273, 2017. [28] H. Dan, “The euro zone–between fiscal heterogeneity and
[10] R. E. Ulanowicz, “The balance between adaptability and monetary unity,” Transylvanian Review of Administrative
adaptation,” BioSystems, vol. 64, no. 1–3, pp. 13–22, 2002. Sciences, vol. 43E, pp. 68–84, 2014.
[11] J. Korhonen and T. P. Seager, “Beyond eco-efficiency: a [29] F. Drudi, A. Durré, and F. P. Mongelli, “The interplay of
resilience perspective,” Business Strategy and the Environ- economic reforms and monetary policy: the case of the
ment, vol. 17, no. 7, pp. 411–419, 2008. eurozone,” JCMS: Journal of Common Market Studies,
[12] R. E. Ulanowicz, “The dual nature of ecosystem dynamics,” vol. 50, no. 6, pp. 881–898, 2012.
Ecological Modelling, vol. 220, no. 16, pp. 1886–1892, 2009. [30] A. W. Davidson and Y. Bar-Yam, “Environmental com-
[13] C. Weigelt and M. Sarkar, “Performance implications of plexity: information for human-environment well-being,” in
outsourcing for technological innovations: managing the Unifying Themes in Complex Systems, A. A. Minai and Y. Bar-
Yam, Eds., vol. IIIB, pp. 157–168, Springer, Berlin, Germany,
efficiency and adaptability trade-off,” Strategic Management
2006.
Journal, vol. 33, no. 2, pp. 189–216, 2012.
[31] M. Nissani, “Ten cheers for interdisciplinarity: the case for
[14] M. Pizzol, M. Scotti, and M. Thomsen, “Network analysis as a
interdisciplinary knowledge and research,” The Social Science
tool for assessing environmental sustainability: applying the
Journal, vol. 34, no. 2, pp. 201–216, 1997.
ecosystem perspective to a Danish Water Management
[32] G. D. Brewer, “The challenges of interdisciplinarity,” Policy
System,” Journal of Environmental Management, vol. 118,
Sciences, vol. 32, no. 4, pp. 327–337, 1999.
pp. 21–31, 2013. [33] I. Feller, “New organizations, old cultures: strategy and
[15] V. Panyam, H. Huang, B. Pinte, K. Davis, and A. Layton,
implementation of interdisciplinary programs,” Research
“Bio-inspired design for robust power networks,” in Pro- Evaluation, vol. 11, no. 2, pp. 109–116, 2002.
ceedings of the 2019 IEEE Texas Power and Energy Conference [34] D. Rhoten and A. Parker, “EDUCATION: risks and rewards
(TPEC), IEEE, College Station, TX, USA, pp. 1–6, February of an interdisciplinary research path,” Science, vol. 306,
2019. no. 5704, p. 2046, 2004.
[16] Y. Bar-Yam, Making Things Work: Solving Complex Problems [35] S. Lélé and R. B. Norgaard, “Practicing interdisciplinarity,”
in a Complex World, Knowledge Press, Manchester, UK, BioScience, vol. 55, no. 11, pp. 967–975, 2005.
2004. [36] J. A. Jacobs and S. Frickel, “Interdisciplinarity: a critical
[17] J. B. De Long, A. Shleifer, L. H. Summers, and assessment,” Annual Review of Sociology, vol. 35, no. 1,
R. J. Waldmann, “Positive feedback investment strategies pp. 43–65, 2009.
and destabilizing rational speculation,” The Journal of Fi- [37] I. L. Janis, “Groupthink,” IEEE Engineering Management
nance, vol. 45, no. 2, pp. 379–395, 1990. Review, vol. 36, no. 1, p. 36, 2008.
[18] T. Lux and M. Marchesi, “Scaling and criticality in a sto- [38] B. Trueman, “Analyst forecasts and herding behavior:,”
chastic multi-agent model of a financial market,” Nature, Review of Financial Studies, vol. 7, no. 1, pp. 97–124, 1994.
vol. 397, no. 6719, pp. 498–500, 1999. [39] W. Pan, Y. Altshuler, and A. Pentland, “Decoding social
[19] B. Lietaer, R. E. Ulanowicz, S. J. Goerner, and N. McLaren, influence and the wisdom of the crowd in financial trading
“Is our monetary structure a systemic cause for financial network,” in Proceedings of the 2012 International Conference
instability? Evidence and remedies from nature,” Journal of on Privacy, Security, Risk and Trust and 2012 International
Futures Studies, vol. 14, pp. 89–108, 2010. Confernece on Social Computing, IEEE, Amsterdam, Neth-
[20] J.-P. Bouchaud, “Crises and collective socio-economic erlands, pp. 203–209, September 2012.
phenomena: simple models and challenges,” Journal of [40] Y. Bar-Yam, “From big data to important information,”
Statistical Physics, vol. 151, no. 3-4, pp. 567–606, 2013. Complexity, vol. 21, no. S2, pp. 73–98, 2016.
[21] D. Harmon et al., “Anticipating economic market crises [41] L. P. Kadanoff, “More is the same; phase transitions and
using measures of collective panic,” PLoS One, vol. 10, no. 7, mean field theories,” Journal of Statistical Physics, vol. 137,
Article ID e0131871, 2015. no. 5-6, pp. 777–797, 2009.
Complexity 15

[42] T. Mora and W. Bialek, “Are biological systems poised at [63] K. Wiesner et al., “Stability of democracies: a complex
criticality?” Journal of Statistical Physics, vol. 144, no. 2, systems perspective,” European Journal of Physics, vol. 40,
pp. 268–302, 2011. Article ID 014002, 2018.
[43] J. B. Davies, R. Lluberas, and A. F. Shorrocks, “Estimating the [64] A. F. Siegenfeld and Y. Bar-Yam, “Negative representation
level and distribution of global wealth, 2000–2014,” Review of and instability in democratic elections,” Nature Physics,
Income and Wealth, vol. 63, pp. 731–759, 2017. vol. 16, no. 2, pp. 186–190, 2020.
[44] S. V. Buldyrev, R. Parshani, G. Paul, H. E. Stanley, and [65] M. Kardar, Statistical Physics of Fields, Cambridge University
S. Havlin, “Catastrophic cascade of failures in interdependent Press, Cambridge, UK, 2007.
networks,” Nature, vol. 464, no. 7291, pp. 1025–1028, 2010. [66] M. Lim, R. Metzler, and Y. Bar-Yam, “Global pattern for-
[45] A. Vespignani, “The fragility of interdependency,” Nature, mation and ethnic/cultural violence,” Science, vol. 317,
vol. 464, no. 7291, pp. 984-985, 2010. no. 5844, pp. 1540–1544, 2007.
[46] D. Helbing, “Globally networked risks and how to respond,” [67] A. Rutherford, D. Harmon, J. Werfel et al., “Good fences: the
Nature, vol. 497, no. 7447, pp. 51–59, 2013. importance of setting boundaries for peaceful coexistence,”
[47] J. Balsa-Barreiro, A. Vié, A. J. Morales, and M. Cebrián, PLoS One, vol. 9, no. 5, Article ID e95660, 2014.
[68] T. C. Schelling, “Dynamic models of segregation,” The Journal
“Deglobalization in a hyper-connected world,” Palgrave
of Mathematical Sociology, vol. 1, no. 2, pp. 143–186, 1971.
Communications, vol. 6, pp. 1–4, 2020.
[69] A. M. Turing, “Computing machinery and intelligence,” in
[48] N. N. Taleb, “How much data do you need? An operational,
Parsing the Turing Test, R. Epstein, G. Roberts, and G. Beber,
pre-asymptotic metric for fat-tailedness,” International
Eds., pp. 23–65, Springer, Berlin, Germany, 2009.
Journal of Forecasting, vol. 35, no. 2, pp. 677–686, 2019. [70] M. D. McDonnell and L. M. Ward, “The benefits of noise in
[49] R. M. May, S. A. Levin, and G. Sugihara, “Ecology for neural systems: bridging theory and experiment,” Nature
bankers,” Nature, vol. 451, no. 7181, pp. 893-894, 2008. Reviews Neuroscience, vol. 12, no. 7, pp. 415–425, 2011.
[50] D. Harmon, B. Stacey, Y. Bar-Yam, and Y. Bar-Yam, [71] N. Goldenfeld and C. Woese, “Life is physics: evolution as a
“Networks of economic market interdependence and sys- collective phenomenon far from equilibrium,” Annual Re-
temic risk,” 2010, http://arxiv.org/abs/1011.3707. view of Condensed Matter Physics, vol. 2, no. 1, pp. 375–399,
[51] A. G. Haldane and R. M. May, “Systemic risk in banking 2011.
ecosystems,” Nature, vol. 469, no. 7330, pp. 351–355, 2011. [72] N. N. Taleb, Antifragile: Things that Gain from Disorder,
[52] J. D. Farmer, M. Gallegati, C. Hommes et al., “A complex Random House Incorporated, New York, NY, USA, 2012.
systems approach to constructing better models for man- [73] T. Olszak, D. An, S. Zeissig et al., “Microbial exposure during
aging financial markets and the economy,” The European early life has persistent effects on natural killer T cell
Physical Journal Special Topics, vol. 214, no. 1, pp. 295–324, function,” Science, vol. 336, no. 6080, pp. 489–493, 2012.
2012. [74] L. F. Su, B. A. Kidd, A. Han, J. J. Kotzin, and M. M. Davis,
[53] F. Schweitzer, G. Fagiolo, D. Sornette, F. Vega-Redondo, “Virus-specific CD4+ memory-phenotype T cells are
A. Vespignani, and D. R. White, “Economic networks: the abundant in unexposed adults,” Immunity, vol. 38, no. 2,
new challenges,” Science, vol. 325, no. 5939, pp. 422–425, pp. 373–383, 2013.
2009. [75] M. D. Seery, “Resilience,” Current Directions in Psychological
[54] D. Sornette, Why Stock Markets Crash: Critical Events in Science, vol. 20, no. 6, pp. 390–394, 2011.
Complex Financial Systems, Princeton University Press, [76] G. Lukianoff and J. Haidt, “The coddling of the American
Princeton, NJ, USA, 2017. mind,” The Atlantic, vol. 316, pp. 42–52, 2015.
[55] E. M. Rauch and Y. Bar-Yam, “Long-range interactions and [77] E. Ostrom, “Beyond markets and states: polycentric gover-
evolutionary stability in a predator-prey system,” Physical nance of complex economic systems,” American Economic
Review E, vol. 73, Article ID 020903, 2006. Review, vol. 100, no. 3, pp. 641–672, 2010.
[56] A. F. Siegenfeld and Y. Bar-Yam, “Eliminating COVID-19: [78] I. Luna-Pla and J. R. Nicolás-Carlock, “Corruption and
the impact of travel and timing,” 2020, http://arxiv.org/abs/ complexity: a scientific framework for the analysis of cor-
ruption networks,” Applied Network Science, vol. 5, pp. 1–18,
2003.10086.
[57] D. Ludwig, B. Walker, and C. S. Holling, “Sustainability, 2020.
[79] A. S. Gard-Murray and Y. Bar-Yam, “Complexity and the
stability, and resilience,” Conservation Ecology, vol. 1, no. 1,
limits of revolution: what will happen to the Arab Spring?,” in
Article ID 7, 1997.
Conflict and Complexity, P. V. Fellman, Y. Bar-Yam, and
[58] H. H. Q. Heng, “The conflict between complex systems and
A. A. Minai, Eds., pp. 281–292, Springer, Berlin, Germany, 2015.
reductionism,” JAMA, vol. 300, no. 13, pp. 1580-1581, 2008.
[80] Y. Bar-Yam, “Engineering complex systems: multiscale
[59] M. Scheffer, J. Bascompte, W. A. Brock et al., “Early-warning
analysis and evolutionary engineering,” in Complex Engi-
signals for critical transitions,” Nature, vol. 461, no. 7260, neered Systems, D. Braha, A. A. Minai, and Y. Bar-Yam, Eds.,
pp. 53–59, 2009. pp. 22–39, Springer, Berlin, Germany, 2006.
[60] M. Scheffer, S. R. Carpenter, T. M. Lenton et al., “Antici- [81] J. Werfel and Y. Bar-Yam, “The evolution of reproductive
pating critical transitions,” Science, vol. 338, no. 6105, restraint through social communication,” Proceedings of the
pp. 344–348, 2012. National Academy of Sciences, vol. 101, no. 30, pp. 11019–
[61] D. Helbing, D. Brockmann, T. Chadefaux et al., “Saving 11024, 2004.
human lives: what complexity science and information [82] D. S. Wilson, Does Altruism Exist?: Culture, Genes, and the
systems can contribute,” Journal of Statistical Physics, Welfare of Others, Yale University Press, London, UK, 2015.
vol. 158, no. 3, pp. 735–781, 2015. [83] B. Castellani, “Map of the complexity sciences,” 2018, https://
[62] H. Rutter, N. Savona, K. Glonti et al., “The need for a www.art-sciencefactory.com/complexity-map_feb09.html.
complex systems model of evidence for public health,” The [84] UNESCO, “Encyclopedia of life support systems,” 2020,
Lancet, vol. 390, no. 10112, pp. 2602–2604, 2017. https://www.eolss.net/.
16 Complexity

[85] R. A. Meyers, Encyclopedia of Complexity and Systems Sci- [105] D. P. Stroh, Systems Thinking for Social Change: A Practical
ence, Springer, Berlin, Germany, 2009. Guide to Solving Complex Problems, Avoiding Unintended
[86] H. A. Simon, The Sciences of the Artificial, MIT Press, Consequences, and Achieving Lasting Results, Chelsea Green
Cambridge, MA, USA, 2019. Publishing, Hartford, VT, USA, 2015.
[87] M. Mitchell, Complexity: A Guided Tour, Oxford University [106] K. E. Weick, Sensemaking in Organizations, Sage, Thousand
Press, Princeton, NJ, USA, 2009. Oaks, CA, USA, 1995.
[88] J. H. Miller and S. E. Page, Complex Adaptive Systems: An [107] R. D. Stacey, Complexity and Creativity in Organizations,
Introduction to Computational Models of Social Life, Berrett-Koehler Publishers, San Francisco, CA, USA, 1996.
Princeton University Press, 2009. [108] D. Hock, Birth of the Chaordic Age, Berrett-Koehler Pub-
[89] H. Sayama, Introduction to the Modeling and Analysis of lishers, San Francisco, CA, USA, 1999.
Complex Systems, Open SUNY Textbooks, New York, NY, [109] M. Hammer and J. Champy, Reengineering the Corporation:
USA, 2015. A Manifesto for Business Revolution, Zondervan, Grand
[90] S. Thurner, R. Hanel, and P. Klimek, Introduction to the Rapids, MI, USA, 2009.
Theory of Complex Systems, Oxford University Press, Oxford, [110] J. Sterman, Business Dynamics, Irwin/McGraw-Hill, New
UK, 2018. York, NY, USA, 2010.
[91] D. S. Byrne, Complexity Theory and the Social Sciences: An [111] E. Sober and D. S. Wilson, Unto Others: The Evolution and
Introduction, Psychology Press, East Sussex, UK, 1998. Psychology of Unselfish Behavior, Harvard University Press,
[92] R. J. Lempert, “A new decision sciences for complex sys- Cambridge, MA, USA, 1999.
tems,” Proceedings of the National Academy of Sciences, [112] R. Axelrod, The Evolution of Cooperation, Basic Books, New
vol. 99, no. 3, pp. 7309–7313, 2002. York, NY, USA, 2006.
[93] S. C. Bankes, “Tools and techniques for developing policies [113] A.-L. Barabási, Network Science, Cambridge University
for complex and uncertain systems,” Proceedings of the Press, Cambridge, UK, 2016.
National Academy of Sciences, vol. 99, no. 3, pp. 7263–7266, [114] P.-G. de Gennes, Scaling Concepts in Polymer Physics,
2002. Cornell University Press, Ithaca, NY, USA, 1979.
[94] S. Glouberman and B. Zimmerman, “Complicated and [115] B. B. Mandelbrot, The Fractal Geometry of Nature, W. H.
complex systems: what would successful reform of Medicare Freeman and Co., New York, NY, USA, 1982.
look like?” in Romanow Papers: Changing Health Care in [116] P. Bak, How Nature Works: The Science of Self-Organized
Canada, vol. 2, pp. 21–53, University of Toronto Press, Criticality, Springer Science & Business Media, Berlin,
Toronto, Canada, 2002. Germany, 2013.
[95] G. R. Teisman and E.-H. Klijn, “Complexity theory and [117] J. H. Brown and G. B. West, Scaling in Biology, Oxford
public management,” Public Management Review, vol. 10, University Press, Oxford, UK, 2000.
no. 3, pp. 287–297, 2008. [118] L. Bettencourt and G. West, “A unified theory of urban
[96] S. Maroulis, R. Guimera, H. Petry et al., “Complex systems living,” Nature, vol. 467, no. 7318, pp. 912-913, 2010.
view of educational policy research,” Science, vol. 330, [119] T. Toffoli and N. Margolus, Cellular Automata Machines: A
no. 6000, pp. 38-39, 2010. New Environment for Modeling, MIT Press, Cambridge, MA,
[97] A. Hall and N. Clark, “What do complex adaptive systems USA, 1987.
look like and what are the implications for innovation [120] T. C. Schelling, Micromotives and Macrobehavior, WW
policy?” Journal of International Development, vol. 22, no. 3, Norton & Company, New York, NY, USA, 2006.
pp. 308–324, 2010. [121] R. N. Mantegna and H. E. Stanley, Introduction to Econo-
[98] X. Bai, R. R. McAllister, R. M. Beaty, and B. Taylor, “Urban physics: Correlations and Complexity in Finance, Cambridge
policy and governance in a global environment: complex University Press, Cambridge, UK, 1999.
systems, scale mismatches and public participation,” Current [122] S. H. Strogatz, Nonlinear Dynamics and Chaos: With Ap-
Opinion in Environmental Sustainability, vol. 2, no. 3, plications to Physics, Biology, Chemistry, and Engineering,
pp. 129–135, 2010. CRC Press, Boca Raton, Fl, USA, 2018.
[99] P. Cairney, “Complexity theory in political science and [123] J. Gleick, Chaos: Making a New Science, Open Road Media,
public policy,” Political Studies Review, vol. 10, no. 3, New York, NY, USA, 2011.
pp. 346–358, 2012.
[100] P. Ball, Why Society Is a Complex Matter: Meeting Twenty-
First Century Challenges with a New Kind of Science, Springer
Science & Business Media, Berlin, Germany, 2012.
[101] L. A. Lipsitz, “Understanding health care as a complex
system,” JAMA, vol. 308, no. 3, pp. 243-244, 2012.
[102] R. Geyer and P. Cairney, Handbook on Complexity and
Public Policy, Edward Elgar Publishing, Cheltenham, UK,
2015.
[103] Y. Bar-Yam, “Complexity theory in applied policy world-
wide,” in Modeling Complex Systems for Public Policies,
B. A. Furtado, P. A. Sakowski, and M. H. Tóvolli, Eds.,
Instituto de Pesquisa Econômica Aplicada, Brası́lia, Brazil,
2015.
[104] B. Edmonds and C. Gershenson, “Modelling complexity for
policy: opportunities and challenges,” in Handbook on
Complexity and Public Policy, R. Geyer and P. Cairney, Eds.,
Edward Elgar Publishing, Cheltenham, UK, 2015.