See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220572508

Complex systems theory: Implications and promises for manufacturing

organisations

Article in International Journal of Manufacturing Technology and Management · January 2000

DOI: 10.1504/IJMTM.2000.001365 · Source: DBLP

CITATIONS READS

51 7,497

3 authors, including:

Ian Paul McCarthy

Simon Fraser University
107 PUBLICATIONS 12,637 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ian Paul McCarthy on 22 May 2014.

The user has requested enhancement of the downloaded file.

Int. J. Manufacturing Technology and Management, Vol. 2, Nos. 1–7, 2000 559

Complex systems theory: implications and promises

for manufacturing organizations

Ian P. McCarthy, Thierry Rakotobe-Joel

International Manufacturing Centre, University of Warwick, Coventry
CV4 7AL, UK
Fax: +44 (0) 247 652 4307 E-mail: [email protected]
E-mail: [email protected]

Gerry Frizelle
Institute for Manufacturing, Department of Engineering, University of
Cambridge, Mill Lane, Cambridge CB2 1RX, UK
Fax: +44 (0)122 333 8076 E-mail: [email protected]

Abstract: This paper discusses the body of knowledge known as complex

systems theory and its relevance to manufacturing organizations. It begins by
addressing the development of systems thinking, in order to explore the
different views on how systems should be studied. This initial commentary
provides an introduction to the concepts of the complex systems theory
approach, along with its various metaphors, tools and techniques. This is
followed by a discussion on how manufacturing organizations conform to
complex systems theory, and in particular complex adaptive systems. An
account of the complex systems characteristics of manufacturing organizations,
and the relationship between manufacturing complexity is also given. Finally,
the paper concludes by discussing the potential application and advantage for
manufacturing organizations from using a complex systems theory approach to
model, understand and in some cases optimize manufacturing operations and
strategy.

Keywords: Complex Systems Theory; complexity; manufacturing

organizations; strategy.

Reference to this paper should be made as follows: McCarthy, I.P., Rakotobe-

Joel, T. and Frizelle, G. (2000) ‘Complex systems theory: implications and
promises for manufacturing organizations’, Int. J. Manufacturing Technology
and Management, Vol. 2, Nos. 1–7, pp.559–579.

Biographical notes: Dr. Ian McCarthy is a Principal Fellow at the Warwick

Manufacturing Group of the University of Warwick. He heads the
Organizational Systems and Strategy Unit within the International
Manufacturing Centre. The activities of the group focus on industrial strategy
and systems, and draw upon biological and evolutionary models.

Dr. Thierry Rakotobe-Joel is a Research Fellow at the International

Manufacturing Centre of the University of Warwick. He was a Fulbright
scholar in the Industrial Engineering program of the University of Cincinnati,
Ohio (USA). He has published articles in the area of supply chain management,
industrial engineering, and strategic management. His research interest is on the
application of complex systems theory to operations strategies.

Copyright © 2000 Inderscience Enterprises Ltd.

560 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

Mr. Gerry Frizelle is an Assistant Director of Research within the Institute for
Manufacturing at Cambridge University. His major interest is developing
measures for complex systems in manufacturing. Prior to that, he set up the
European operation of Scheduling Technology. This followed 15 years with
Cape where he was ultimately Managing Director of three subsidiary
companies. He is the author of the book The Management of Complexity in
Manufacturing published by Business Intelligence. He holds degrees in Physics
and Pure Mathematics and is a Chartered Engineer.

1 Introduction

JIT, TQM, Agile Manufacturing, Supply Chain Management, Kaizen, QFD, … are just a
few of the myriad of management techniques, methods and philosophies that
manufacturing organizations have become more or less successfully acquainted to over
the last decade. While these management techniques vary in their resulting effects, a
single fact remains – the manufacturing environment, with its technology intensive and
innovation thirsty nature, will always be a favorite guinea pig of management scientists.
Recently, the results of several interdisciplinary scientific efforts have led to the
resurgence of a body of knowledge known as the complex systems movement or often
referred to as ‘complex systems theory’. Many businesses and consultancies have started
to embrace complex systems theory, forming departmental units that develop tools and
methods for application to their organizations.
It is important to note that complex systems theory is not about the study of
‘complexity’. It is concerned with understanding the behaviour of systems – often
complex – that evolve, learn and adapt [1]. To manufacturing organizations, complex
systems theory could provide an alternative perspective for conceptualizing and
modelling their internal dynamic. In this paper, we present manufacturing organizations
as complex adaptive systems that are able to more or less satisfy their customers by
continuously adapting themselves to meet the needs and expectations of the market of the
latter. The industrial revolution, the quality revolution, the lean movement and the agile
age are representative of such adaptive changes. In addition, complexity, in its various
forms, has always existed in manufacturing and other organizations, and this is one of the
main attractions of complex systems theory to the manufacturing arena.
The first section of this paper provides an introduction to complex systems theory and
its development within systems sciences. It then provides a discussion on the different
approaches used in complex systems theory. The second section provides an analysis of
the manufacturing organization as a complex adaptive system. While, the third section
concludes with an evaluation of the relevance and application of the complex systems
theory to manufacturing organizations.

2 Sciences in motion

The development of complex systems theory has been influenced by new thinking and
rationales for studying the emergent properties of systems. Often, complex systems
theory is associated with the study of self-organizing systems, in disciplines such as
biology, physics and chemistry. While this notion is true, the advent of complex systems
theory can be traced back through various periods of systems thinking. In order to
 Complex systems theory 561

understand the development of systems theory and its relevance to complex systems
theory, a review of the evolution of knowledge that led to the complex systems theory is
given.
In his eloquent review of the evolution of systems process thinking, Capra [2]
explored the path to complex systems thinking by reviewing the Galilean, Newtonian and
Cartesian approaches to understanding living systems and their environment. This path to
complex systems theory is paved with developments that are motivated by mankind’s
desire to understand the universe and the origins of life. These developments did not
occur instantaneously, but rather as a series of significant advances over many hundreds
of years, in the concepts and rationales used to model and understand systems. At the
heart of this development has been a debate about role of the whole and the part. There
have been two opposing views from the reductionist and the organic proponents. The
reductionist view, also referred to as the mechanistic view, sought to understand system
behaviour by reducing the system to its constituent parts and studying the behaviour of
and relationship between the individual parts. This view believed that a detailed
knowledge about the parts would be representative of the behaviour of the whole system.
The opposing view, referred to as the organic view, believes that that you cannot infer
accurate predictions about the whole system by studying the systems parts. This is
because the parts often have degrees of self-organization and independence. Thus, the
fundamental difference between the two views is the role of the parts as either fully
dependent building blocks of the whole system, or as required but autonomous factors
that make up the system.

2.1 Cartesian mechanism

The Cartesian view (mechanistic and reductionist) opposed the Aristotelian philosophy
and Christian theology that the universe and its systems were organic, living and
spiritual. This view considers such systems analogous to machines with clockwork like
behaviour [3]. Scientific discoveries in the areas of mathematics, physics and astronomy
by people such as Newton, Descartes and Galileo helped develop and support this view.
This was a period where phenomena could be measured and quantified precisely and
many scientific laws were proven and represented by using mathematics and physics [4].
This era reached a high point with Isaac Newton’s Principles of Mechanics.

2.2 Romantic movement and 19th century mechanism

During this period, there was a resurgence of the Aristotelian view of systems i.e. that
systems were organic in their characteristics and behaviour. Goethe’s view of biological
entities, as reported in Hoffman [5] and Brady [6], is “… a pattern of relationships within
an organized whole …” is representative of this period, as researchers sought to study
natural systems as qualitative patterns through forms. The emphasis was a qualitative
approach, rather than the quantitative one of the Newtonian and Cartesian mechanists. In
the same framework, Kant contributed significantly to the Romantic view of the world by
asserting that all natural phenomena could not be explained by science [7]; and that
nature should be studied by trying to understand its purpose, i.e. by gathering
confirmatory and exploratory data that would complement theoretical assertion [8].
In his argument on the difference between machines and living systems, Kant [7,9]
first used the term self-organizing wholes in an attempt to show that machines have parts
562 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

that exist for each other, whilst living organisms have parts that exist by means of each
other. He suggested that living organisms have parts that must be thought of as an organ
that produces the other parts; consequently, the organism would therefore be an
organized and self-organizing being.

2.3 General systems science

During this period there was yet again a shift in view about the parts and their
relationship to the whole. The focus is on the behaviour and characteristics of the whole
(i.e. the whole system). Such thinking was the foundation of Von Bertalanffy’s [10]
general systems science. While the notion of the system and system thinking were
already being used by many scientists prior to Bertalanffy’s work, it was his research
about open and closed systems that gave momentum to general systems science. Indeed,
ever since the introduction of evolutionary thinking in the 14th century, scientists had
grappled with the dilemma about whether systems would be closed or open [11]. One
view was that the living world was following a physical trend from order to disorder, as
exhibited by a closed system. The second law of thermodynamics, also known as entropy
theory, is at the basis of this closed system view of the world. It asserts that the physical
world is continuously dissipating. While the other view, is that the world and its system
are constantly moving from disorder to a more complex order and the theoretical basis for
this view is based on aspects of evolutionary theory.
While Bertalanffy recognized the legitimacy of both views, he could not apply the
concepts of either for describing living systems. Instead, he adopted the view of living
systems as open systems that maintain themselves in a quasi-state of equilibrium, through
exchanges of material with the environment. He concluded that living systems are open
systems, which are far from equilibrium. This is in contrast to the second law of
thermodynamics, which suggested that living systems are at, or near equilibrium. The
outcome was that a new thermodynamics of open systems was necessary to address the
problems of living systems [10,11]. Prigogine [12] achieved the reconciliation of the two
dissenting views in his elaboration of the self-organization of dissipative structures and
thus introduced an important and unifying approach to studying living systems.

2.4 Cybernetics
Parallel to the development of the general systems science theories, was the work of a
group of engineers, mathematicians, neuroscientists and social scientists. Defined by
Wiener [13], Cybernetics (Greek: Kybernetes = Steersman) is the science of control and
communication in animal and the machine. One of the aims of cybernetics is to create
cognitive and mathematical models of natural phenomena. Scientists such as Von
Neumann and Wiener were at the forefront of the cybernetics movement. The Macy
conferences in New York, in the 1940s, were the main forum for cybernetics followers
[14] and led to various developments in the cognitive sciences [15,16], information
theory and computer modelling [17]. A notable development of the cyberneticists was on
the notion of feedback as the conveyance of information about the outcome of any
process or activity to its source. Self-balancing and self-regulating feedback are different
forms of this phenomenon.
Another development by the cybernetics movement was the formulation of theories to
explain self-organization. Capra [2] summarized the characteristics of self-organization
as the spontaneous emergence of new structures and new forms of behaviour in open
 Complex systems theory 563

systems far from equilibrium. Such systems include internal feedback loops and can be
described mathematically using non-linear equations. In summary, a significant step to
complex systems theory had been made.
From a manufacturing perspective, organizations are continually self-organizing. This
leads to an evolution of the system i.e. from craft production, to mass production, to lean
production and to agile production. These systems also co-evolve with other systems in
the industrial environment. There is also a strong link between cybernetics and
manufacturing control systems as shown in Table 1. This table has been adapted from
Beers [18] analogy of a business system and the human nervous system.

Table 1 Links between cybernetics and manufacturing control

System Level Physical Analogy Function Manufacturing Analogy

System 1 Organ, Muscle Action Manufacturing Activities
System 2 Spinal Cord Signal Processing Shopfloor communication
and Transformation
System 3 Muscle Fibers Control MRPII and ERP systems
System 4 Nervous System Sensing and Risk Management,
Monitoring Maintenance and Process
Monitoring
System 5 Cortex Policy and Strategy Manufacturing Strategy
Formation Process

2.5 Complex systems movement

This systems movement is influenced by the progress that took place in cybernetics and
general systems science. As the development of systems thinking and associated
mathematics allowed the study of large aggregate systems rather than individual parts,
the notion of complex systems theory started to take form. This allowed researches to
infer commonalties about the collective behaviour of the system and the system’s
interaction with its environment. The field of complex systems has multi-disciplinary
origins. Its application cuts across the pure science disciplines and starts to embrace
engineering and management disciplines.
However, the introduction of the term complex to systems thinking created a new
dilemma about the definition and use of the word complex and its relation to terms such
as simple, complex and complicated. In fact, there is confusion and concern that complex
systems theory is about the study of complicated or difficult situations. There is also a
perception that complex systems theory, when applied to management, advocates that
business and industrial systems should have high levels of complexity, when most
organizations would prefer to have simplicity. Neither of these views is correct. Although
complex systems theory can certainly be used to model and understand industrial and
business systems that have high levels of complexity, this body of systems knowledge is
concerned with:
1 Recognizing that business and industrial problems are difficult to comprehend
because of complex system behaviour i.e. understanding how the parts of a system
give rise to the collective behaviours of the system.
564 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

2 Recognizing that many business and industrial systems can be classified as complex
systems, because they exhibit degrees of self-organization, emergence, innovation,
learning and adaptation.
3 To acknowledge that the study and modelling of such systems, requires metaphors
and tools which are able to capture such characteristics and overcome the limitations
of reductionist approaches
A popular definition, which reflects the above concepts, is given by Sherman and Shultz
[19], from the Santa Fe Institute. It is however important to note that the term complexity
is used, rather than complex systems theory, and that this is a classic example of the
confusion between the terms complex, simple and complicated
“Complexity refers to the condition of the universe which is integrated and yet
too rich and varied for us to understand in simple common mechanistic or
linear ways. We can understand many parts of the universe in these ways, but
the larger and more intricately related phenomena can only be understood by
principles and patterns – not in detail. Complexity deals with the nature of
emergence, innovation, learning and adaptation”. [19]
This section will conclude by trying to resolve the issue of definitions before discussing
manufacturing as a complex system and in particular a complex adaptive system. This is
important in order to understand manufacturing as a complex adaptive system and the
various tools and metaphors presented in section 4 of this paper.
• Complexity. The Physicist Larry Smarr [20] pointed out that there is no one right way
to define and measure complexity. It is a multi-dimensional, multi-disciplinary
concept. Indeed, a mathematician might define complexity as the number of degrees
of freedom in computational operations. While a manufacturing engineer would
relate it to very high levels of product variety or sophisticated production schedules.
A good starting point is to consider complexity as a systems adjective, inferring that
the system has a high number of parts and a high number of system states. This
means that systems with high levels of complexity are difficult to describe and to
recreate. In this sense, the term was used by systems and information theorists in the
1960s. Shannon information theory and algorithmic complexity [21] were concerned
with describing systems, whilst computational complexity was related to problems of
recreating systems. In the same way, Frizelle et al. [22,23] adopted ideas from
information theory to capture complexity in manufacturing systems.
• Simple. Etymologically, simple is derived from the Latin word simplex, which
literally means ‘once folded’ [24]. In the context of systems, simple systems would
therefore relate to systems that are easily understood and whose behaviour is
relatively predictable. True simple systems are not common, because in the long term
most systems have relatively high levels of unpredictability. In Casti’s classification
of systems [25], he suggests that simple systems, are the teaching and educational
models found in schools.
• Emergence. This is a property of a complex system result directly from the system’s
evolution and occurs independently. Literally, emergence means ‘to dive out’ or to
come out of the depths. In the context of complex systems, emergence relates to the
apparition of new system behaviour because of: the collective behaviour of the parts,
as opposed to the individual behaviour of each part, and the system’s response to its
environment. Emergence is an important characteristic of complex systems theory as
it allows the identification of new opportunities.
 Complex systems theory 565

• Self-Organization. This refers to the spontaneous creation of increased order,

resulting from the internal dynamics of the system. Self-organization often emerges
from learning, adaptation and innovation. Complex systems that consist of a large
number of interacting elements can display self-organizing behaviour. An example
of self-organization in manufacturing organizations is the implementation of
operating rules in multi-stage manufacturing systems. The interaction of individual
process variations through individual learning, adaptation from the information
gathered about neighbouring processes and the resulting innovation or changes at
local scale can greatly affect the performance of the overall system.
• Adaptive. This refers to the change process in complex system. If the change is in
response to achieving a certain goal or objective, or in response to an environmental
change, then systems that adapt themselves, are known as complex adaptive systems.
This occurs because the system is open, has feedback mechanisms and has the
ability to self-organize. Thus, manufacturing organizations along with most systems
in the universe are complex adaptive systems, as they learn, adapt and evolve over
time [26].
• Chaos. A chaotic system is a deterministic (predictable) system, but it is difficult to
predict, because the future state of chaotic systems are very sensitive to the current
state of the system. This is known as ‘sensitivity to initial conditions’.
In summary, complex systems theory is neither a science nor a technique. It is rather a
field of thinking with a set of techniques and methods that allow people to view systems
as entities that can have emergent behaviour and adapt. This view largely opposing the
reductionist view, but it is important to note that the reductionist view is still important
for understanding certain aspects of a system, especially system structure. Thus, complex
systems theory suggests placing priority on the importance of relationships and patterns
inside the system, and essentially the learning opportunity that the complexity mindset
offers [27].

3 Manufacturing organizations as complex adaptive systems

Gell-Mann [28] noted that Edward Deming recognized firms as complex adaptive
systems, as early as the 1950s when he saw organizations constantly gathering
information about their environment and adapting to it. Kaufman [29] also pointed out
that “… Organisms, artifacts, and organizations are all evolved structures.” A complex
adaptive system (CAS), as defined by Waldrop [30], is an open system:
“… in which complex behaviour of the system as a whole emerges from the
interaction of large numbers of simple components, and in which the system is
able to adapt, to automatically improve its performance (according to some
measure) over time, in response to what has been encountered previously”.

From this definition, there are several important characteristics of a CAS. Firstly, each
system is a network of many individual active elements, referred to as agents that interact
in various ways, using their own internal rules, states, and strategies of past experience.
These internal rules are referred to as schemas. Secondly, a CAS has the ability to learn
and hence adapt to a new environment. The system is constantly revising and re-
organizing its agents as experience is gained from past interactions. From this learning,
the system will develop its strategies for the future, by changing its schema. Without this
566 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

adaptability, the system is likely to face extinction. In addition to these two interesting
characteristics, Waldrop [30] also noted that complex adaptive systems will in one way or
another try to predict the future and then react to the situation regardless of the actual
effects on the environment. A complex adaptive system will make predictions based on
its internal models of the environment. These predictions are based on the internal and
external assumptions of the agents, relative to the environment.
Manufacturing organizations are complex adaptive systems (Figure 1). They consist
of an integrated assembly of interacting elements, designed to carry out cooperatively a
predetermined objective, which is the transformation of raw material into marketable
products. Using a systems definition, a manufacturing organization therefore consists of:
1 Assemblage. A number of distinguishable elements (people, machines, departments,
components, sub-systems, etc.).
2 Relationship. Simply bringing the elements together results in a ‘group’ rather than a
system. For a system to exist the elements must have relationships. The materials and
resources within a manufacturing system must interact to produce a product.
Materials + Resources = Products
3 Objectives. Manufacturing systems have multiple objectives which can result in
conflict. The purpose of a manufacturing system is to organize elements and sub-
systems to satisfy certain objectives. They will vary from organization to
organization, but are generally considered to be those listed below:
• Produce products within defined process and design specifications.
• Produce products within defined time scales.
• Produce products to specified levels of quality.
• Produce products to a cost acceptable to customer and company.
4 Adaptive. A manufacturing system must operate and adapt to market, economic
and political conditions in order to survive and stay competitive.
If a manufacturing organization is now related to the key concepts of complex adaptive
systems (agents, schemas and predictions), the following statements can be made:
• Manufacturing Agents: Any entity (person, machine, supplier and customer) within
the system that can produce an effect (new orders, machining, break downs,
unloading, etc.) Agents have a degree of autonomy (machines are autonomous as
they break down on their own!). Manufacturing organizations consist of a multitude
of agents that tend to be independent in their operations and behaviour, despite the
connections that exist among them and efforts by managers to fully control them. In
summary agents are characterized by:
– Their internal states, such as ‘operating’ or ‘idle’ (e.g. the status of a machine or
worker).
– The input they receive and the output they generate (e.g. processing raw material
or information).
– The process that allows them to make simple decisions (e.g. the functional task
that carries out the process – machining, scheduling, ordering, etc.)
• Manufacturing Schemas: a characteristic of the interaction of the agents is the
creation of schemas inside the system. Schemas are the rules or procedures that can
 Complex systems theory 567

be specific to individual agents (e.g. the mentality and attitude of individual workers
and machines) or can be shared by two or more agents (e.g. this organization
operates a make to order system).
• Manufacturing Predictions: this is when manufacturing organizations attempt to
anticipate the future. It includes forecasting activities such as diagnostics for
maintenance, and statistical process control for quality and sales forecasting.

Figure 1 Manufacturing organizations as open systems

568 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

In summary, like other living systems manufacturing organizations are complex adaptive
systems whose productivity, adaptability and profitability are dependent on the role if its
agents and the resulting schemas. The issue of complexity (or complicatedness) is not
new to manufacturing. It is rather the new perspective that complex systems theory
provides for studying manufacturing organizations, that is considered to be novel and
worthwhile.

4 Implications and promises of the complex systems theory to

manufacturing organizations

This section presents a framework for understanding some of the known and emerging
bodies of work dealing with complex systems theory and how they relate to each other in
terms of knowledge application and levels of manufacturing activity.
The framework is shown in Figure 2 and has two axes: a knowledge axis that is a
continuum extending between abstract and applied knowledge and a manufacturing
activity axis that ranges from strategic to operational. The knowledge and body of work,
which exists within the field of complexity and complex systems theory, is embodied
within a number of different concepts and constructs. The following sections describe the
various ideas, metaphors, theories and tools that constitute complexity and complex
systems knowledge

Figure 2 Complex systems framework and manufacturing issues

 Complex systems theory 569

4.1 Memetics

Memetics is the study of memes. The name originated from Dawkins attempt to
characterize the crossing of ‘memory’ and ‘genes’ [31]. A meme can be thought of as a
unit of knowledge (e.g. an idea, a concept, a form of technology), which evolved. To
manufacturing organizations, memes are analogous to genes in biological organisms. It is
like an organizational gene or blueprint, which contains a manufacturing organization’s
history, its past experiences and the resulting learning process. Memes transmit instituted
past ideas and concepts to improve new working practices. De Geus [32] suggests that
memes enable organizations to learn and are the characteristics of organizational culture.
Thus, one of the main applications of memetics to manufacturing is the ability to
understand the knowledge management processes that exist within the organization.
In an engineering design context, memes are the design knowledge that accompany
designed artifacts as they evolve. For instance, if we consider computational devices,
there are a host of ideas and concepts which have accompanied the evolution of the
abacus, to Babbage’s first computing device, to the first mainframe computers, to the
desktop calculators, to the first personal computers, to the latest palmtop computers. As
each piece of technology has evolved, there is accompanying design knowledge (a
meme).
To understand how memes are transmitted between generations of engineering
designs, or types of manufacturing organizations is a knowledge management issue. The
model that is used to understand the knowledge processes is based on three criteria:
heritability, variability and selection.
• Heritability means that the information must be transmissible in some way. In the
modern world, this criterion is easily satisfied, as information is documented and
published via books, television and the internet. The transmission of information can
take many forms, e.g. reading a book, a training course, an informal conversation
between two people and formal education at schools and universities. A classic
example of heritability is when people move from one job to another in a different
industry. If a person has worked for 20 years in the mining industry and then joins
the automotive industry, that person takes with him 20 years of experience and ideas
(memes) to his new job. The important thing is that this person will transmit or infect
some of his new colleagues with his memes, in the same way that his new working
environment will change him.
• Variability is the notion that there is scope for difference. Memes do not replicate
perfectly. In the above example, even though the worker takes with him experience
and ideas to his new job, much of it will not be relevant to the new job and will be
forgotten or discarded.
• Selection. When memes are popular, successful or competitive, there is a tendency
for them to replicate. That is, if an engineering design is successful it will be
imitated. In terms of organizational philosophy the success of the Toyota Production
System spawned a host of imitators across the world. The variability process created
different versions of this new way of manufacturing. European and North American
systems were slightly different from the original Japanese system, due to the cultural
and political differences that existed in these continents.
570 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

4.2 Chaos theory

To manufacturing organizations, chaotic behaviour is a daily realization in supply chains,

quality control issues, organizational change programs and machine reliability, to name
but a few. Many manufacturing managers believe that the systems under their
management are behaving randomly and out of control, but chaos theory can show that
the system behaviour may have deterministic origins and deterministic processes which
will show patterns of behaviour i.e. they are not random.
For instance, a spring manufacturing company used chaos theory to understand how
defects were occurring during the spring manufacturing process [33,34]. The quality
issue faced by the company was the coilability testing of carbon steel wire. Coilability
testing is an important process in spring manufacturing, as it corresponds to the strength
of the produced spring. The difference between good coilability and bad coilability was
always considered a random variable due to a ‘black art’ manufacturing process. By
using chaos theory, patterns of order were discovered and thus knowledge on how to
minimize poor coilability was gained.

4.3 Edge of chaos

Derived from chaos theory, the edge of chaos metaphor suggests that manufacturing
organizations behave optimally at a state poised between order and chaos, known as the
edge of chaos (EOC). Described by Kaufmann [29] as “… a natural state between order
and chaos, a grand compromise between structure and surprise …”, edge of chaos is a
phase transition between excessive rigidity and lenience. Manufacturing organizations
that exhibit self-organizing characteristics are operating within this small stretch of the
EOC that provides them with high responsiveness to their environments opportunities but
enough structure to act and perpetuate themselves [29]. The edge of chaos is an
organizational state that allows them to have high levels of responsiveness, variety,
creativity and vitality.

Figure 3 Organizational responsiveness at the edge-of-chaos

 Complex systems theory 571

If we consider manufacturing management, the edge of chaos is the situation between

highly disordered and disrupted operations (machine breakdowns, rush orders for
customers, strikes) and highly planned and ordered activities. The ability to rapidly
introduce new products and to respond quickly to changing customer demands is an
example of behaviour at the edge of chaos (manufacturing agility or responsiveness). A
highly unfocused manufacturing strategy would create a level of randomness and
variability that falls into chaotic behaviour, whilst a strategy, which is overly rigid, would
result in an organization unable to adapt i.e. it has too much order.
In summary, the desire by many manufacturing managers to have stability and
equilibrium in their factories and the volatility of the economy call for a strategy at the
edge of chaos. Thus, complex systems theory advocates that manufacturing organizations
should be at a non-equilibrium state. The edge of chaos metaphor states that there is a tri-
phase process to achieve this:
1. A chaotic regime when the disturbances and problems arise e.g. a customer requests
engineering changes or a supplier fails to deliver.
2. An ordered regime when the above problems and issues have been successfully
resolved.
3. An edge-of-chaos complex regime when an emergent behaviour is constantly sought
in order to keep the system alive and dynamic.

4.4 Self-organizing systems

Self-organization is the evolution of a system into an organized form in the absence of

external constraints, and it can only take place when the system is near the edge of chaos.
For organizations, it is believed that there are five levels of self-organization that exist.
Table 2 presents each level in terms of the nature of the processes taking place within the
organization, the ways in which decisions are made, the types of trade-offs taken into
account during decision making, the attributes of the organization’s knowledge and the
types of agent available to the organization.

Table 2 Levels of Self-Organizations. Adapted from Kelly and Allison [35]

Self-Organization Process Decisions Trade-Offs Knowledge Agents

Level
Consciously Enterprise Value-based Mutually Evolutionary Dynamic
competent always beneficial relationships and
improving changing
Quantitatively Based on Data based Anticipated History and Quantified
guided statistics simulation
based
Guided Units that Rule based Objective History and Leveraged
work team based
Conscious Loyal to team Integrative Visible Team based Common
plan and public
Unconscious Agent ad hoc Reactive Unclear Personal Private
572 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

4.5 Cladistics
Cladistics is a method that has been used to study different manufacturing strategies and
the evolution of organizational forms. It is a classification method used primarily in
biology to study diversity by examining the evolutionary relationships between entities
with reference to the common ancestry of the group. The output of a cladistic
classification is a cladogram, which is a tree like diagram that represents different breeds
of manufacturing organization along with their defining characteristics. The value of this
classification method is the information contained within the diagram. It provides a
transparent snapshot of different manufacturing strategies, along with information about
how to formulate each strategy and move from one strategy to strategy to another. As
reported by McCarthy [36–38] this system of coordinating information has application
and value in the areas of change management, benchmarking, strategy formulation.

4.6 Intelligent agents

There is no unified definition of intelligent agents, but it is linked to the term adaptive
agents. Whereas an adaptive agent has operating states, inputs and outputs and a degree
of decision making, an intelligent agent as defined by Jennings and Wooldridge [39] is
some form of computer programs or computational systems capable of flexible
autonomous action in order to meets its objectives. Jennings and Wooldridge refer to the
flexibility of such system as:
• Responsive: agents should perceive their environment (workers and their
departments, cells or assembly lines) and respond in a timely fashion to changes that
occur.
• Proactive: agents should not simply act in response to their environment, they should
be able to exhibit opportunistic and goal directed behaviour, i.e. they are able to take
the initiative.
Several initiatives have used intelligent agents (also known as agent based systems or
intelligent systems ) to help manufacturing organizations continually change, adopt new
technologies, create new structures and manage new working practices. For an agent-
based system to help manufacturing organizations in this way, the system itself must be
capable of self-reconfiguration and change. Manufacturing strategies that adopt such
systems to help new product development and rapid market responsiveness are known as
intelligent manufacturing initiatives. For a state of the art survey of how intelligent
manufacturing initiatives have been adopted by industry, the reader is referred to Shen
and Norrie [40].
In summary, the intelligent agent approach places importance on the ‘behaviour
producing’ aspects of a system, rather than ‘information structure’ aspects of a system.
The behaviour of intelligent agents does not depend on the user’s input or specific
problems that are closely related to stored knowledge. Instead, intelligent agents learn
and attempt to solve the problem according to their perception of the environment and the
individual goals of the agents. Applications have shown that agent based approaches have
the following advantages for enterprise integration and supply chain management [40]:
• Increasing the responsiveness of the enterprise to the market requirements;
• Involving customers in total supply chain optimization;
 Complex systems theory 573

• Realizing supply chain optimization through effective resource allocation;

• Achieving dynamic optimization of materials and inventory management;
• Realizing total supply chain optimization including all linked enterprises;
• Increasing the effectiveness of the information exchange and feedback.

4.7 Genetic algorithms

Genetic algorithms (GAs) developed from the use of computers to solve computational
problems. They are based on the idea of a population-based search that mimics the
principles of evolution and are thus a type of evolutionary computing. The development
and utilization of GAs has been the most prominent of the evolutionary computational
techniques.
A GA creates a set (population) of candidate solutions to the problem (usually fixed
length character binary strings). Each solution receives an associated fitness value based
on its ability to solve the problem according to set criteria. A new population (the next
generation) is created using a computational process, which mimics the Darwinian
principle of selection, reproduction and mutation. For a detailed discussion on GA
procedures, see Holland [41].
In terms of application, GAs have a good track record. John Deere, the world’s
leading producers of agricultural machinery, used a standard GA-based optimization tool
(Evolver™, an Excel add-in from Palisade Corporation) to optimize production problems
at some of their factories. The problems included the scheduling of transportation and
delivery, factory scheduling and production capacity. Each solution (binary string) in the
algorithm was made up of manufacturing constraints such as: manufacturing hours
available, sequencing/spacing, changeovers, parts/components availability, priority to
orders (retail and earliest scheduled shipping date) and shipping off the end of the line.

4.8 Dissipative structures

Chaos theory and in particular the edge of chaos metaphor are concerned with system
stability and transformation. Manufacturing organizations are continually trying to re-
invent themselves and therefore are concerned with organizational effectiveness
initiatives such as JIT, TQM, lean, agile and mass customization. Dissipative structure,
initially developed by Prigogine in chemical systems [12], has been used in complex
systems framework to assist organizational change.
As reported by MacLean and MacIntosh [42], when a system moves further from
equilibrium to the point where a ‘descent in to chaos’ ensues and the system structures
are broken down, then at this point, the system becomes open to its environment,
importing energy and exporting entropy. This exporting of entropy is termed
‘dissipative’. It is used as a measure of disorder and corresponds to a new structure,
operations and rules within the system. Thus the concept of ‘dissipative structures,’
proposes that as a stable system becomes chaotic, new order emerges, whilst the ‘edge of
chaos’ metaphor suggests that systems are constantly adapting and self-organizing, but
they do not cross the line into chaos.
Therefore, dissipative structure theory is useful for strategic managers, who are
concerned with radical organizational transformation. If we consider the case of a
574 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

manufacturing organization that is continuously under-performing and is facing pressure

to change. The effect of a business crisis (possible closure, a take over, receivership, etc.)
will generate a chaos factor that could lead to the emergence of new order. MacLean and
MacIntosh [42] suggest a model to describe the application of dissipative structure theory
to organizational change. It has three stages:
1 Conditioning – comprehending the rules and structure that underpin the current
organizational form
2 Create far from equilibrium conditions – begin the radical and chaotic change
program
3 Manage the feedback process – as the new organizational form emerges create
positive and negative feedback to avoid returning to the old organizational form.
The creation and management on these conditions and the resulting state of chaos often
lead to an emerging order that can help organizations to move from a situation of
business closure to one of business viability.

4.9 Fitness landscapes

Kauffman [29] used this notion to investigate the process of self-organization and natural
selection. He noted that adaptation is usually thought to be a process similar to ‘hill
climbing’ where minor variations of the species (from one generation to the next
generation) result in a move towards a peak of high fitness on a fitness landscape. The
concept of natural selection and survival of the fittest will push an organization towards
such peaks. This fitness landscape can be imaged as a series of hills and valleys of
different heights and depths. To represent such landscapes, Kauffman created NK
models, which are derived from the physics spin-glass model. Fitness is then defined as
the ability to successfully navigate such landscapes to survive and compete.
To illustrate NK models, an example concerned with the decision ‘how many pieces
of equipment should be purchased for a customized job’ was used by McCarthy and Tan
[43]. If there are three types of machine that could be purchased to satisfy a machining
requirement, then this set of alternatives is known as N = 3 (Machine A, Machine B and
Machine C). The solution to the problem is to buy one or all three pieces of equipment or
any of the intermediate combinations. For each machine, there are two values: to
purchase the machine (1) or to not purchase the machine (0). This simple example
provides a straightforward binary code of the problem, with the total combination of 23
(i.e. 8 possible combinations). With each combination, a value (from 0 to 1) called fitness
can be allocated according to criteria which could be based on cost, flexibility, speed, etc.
The definition and function of fitness is not simply to optimize. In the definition used
in McCarthy and Tan [43], the term ‘genotype’ was replaced with a manufacturing
solution. Therefore, in a manufacturing context, the concept of fitness is the ability of a
manufacturing organization to survive by inheriting, imitating and searching solutions
(of any form) that produce a desired outcome (measurable or immeasurable) such as
profit, organization goal, purpose etc (i.e. long term survival).
With the machine selection example, a random number has been allocated to each
solution to indicate fitness. The total possible number of combinations and the assigned
fitness for each individual combination is shown in Table 3.
 Complex systems theory 575

Table 3 Combination of machines and the allocated random fitness

Machine A Machine B Machine C Assigned random

fitness
000 Do not purchase Do not purchase Do not purchase 0.0
001 Do not purchase Do not purchase Purchase 0.1
010 Do not purchase Purchase Do not purchase 0.3
011 Do not purchase Purchase Purchase 0.5
100 Purchase Do not purchase Do not purchase 0.4
101 Purchase Do not purchase Purchase 0.7
110 Purchase Purchase Do not purchase 0.8
111 Purchase Purchase Purchase 0.6

Since N=3, a three-dimensional cube is used to represent the possible space of solutions
and how they relate to each other (see Figure 4). The corner of each cube is a possible
solution and neighbouring solutions are said to be a ‘hamming distance’ apart. For
example, solution 000 has the neighbours 001, 010 and 100. With the cube representing a
three dimensional landscape of solutions, one of the solutions is selected as the random
staring point. If the point has a fitter neighbour, i.e. a higher fitness value, the species will
walk (evolve) from the existing point to the fitter point. The arrows shown on the lines of
the cube indicate the ‘uphill walks’ (walking to a fitter immediate neighbour). Within the
whole cube, there is one global optimum 110 (0.8); this is referred to as the ‘fittest peak’.
A ‘local peak’ is a point where there are no fitter neighbours and the current position is
not the fittest peak. Such a point would be 101, where all the arrows point to the
intersection and not to a neighbour

Figure 4 Possibility space and associated fitness

576 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

4.10 Entropic measures

Frizelle and his colleagues [23,44,22] at Cambridge University looked at the problems of
measuring complex adaptive systems and, in particular, manufacturing systems. They
start by simplifying the problem by ignoring adaptation – a reasonable approximation
given the time periods over which they take measurements. Then they adopt ideas from
information theory [21].
It is not exaggerating to say that complex systems ‘radiate’ information and the more
complex the system becomes the more information it radiates. This is because the models
used to describe the measurement process are closely related to those of information
transmission from a source to a receiver. These might be voice messages down a
telephone line, a television picture or radio signals bounced off a satellite.
However there are a number of important differences. The first is that the
transmission of information is usually concerned with ‘intelligible’ information – using
the term in its everyday sense. Information theory, as the field of study is called, does not
limit itself only to intelligible information; any signal will do. Therefore, the ‘snow’ on a
television screen is as much ‘information’ according to the theory as is the picture – even
if a great deal less interesting. The difference between the two resides in the structure of
the information in each case, not in any material difference. This becomes clear in digital
transmissions. Either ‘noise’ or intelligible information can be transmitted as a series of
zeros and ones.
However there is a second important difference between message transmission and
observing complex systems; there is nothing specific about the information radiated by
the latter. If a transmitter sends a receiver a message, then the message is quite specific –
it is the intelligible information that the transmitter wants to impart to the receiver. This is
not the case for a complex system. The information is generated by the system’s variety,
so that any source of variety will do. The difficult question is which source of variety to
take. Moreover, if two or more sources of variety are selected, which is the more
important and how are they related.
The Cambridge group was able to demonstrate that queuing behaviour and its causes
were the important issues. They therefore observed the dynamics of queues along with
the reason queues were generated. The information theoretical approach allowed them to
develop a single index that permitted comparisons to be made across the process. This
index is the average rate at which the information is generated and is called the entropy of
the system. Put another way, it is the amount of information the observer needs to capture
to understand the system.
The theory predicted a number of results that seem to support intuitive ideas of how a
complex manufacturing system might behave. They include one that says the more
complex the operation becomes (complex used in the ‘information’ sense), the longer it
will take parts to get through the system. Another is that as the system becomes more
complex, it becomes harder to make reliable promises. A third is that the most complex
operation within the system is liable to be the bottleneck while a fourth is that
‘simplification’ and/or control are the only options for tackling unwanted complexity –
although it can also be dispersed.
The approach has been applied within factories, to drawing offices and more latterly
to supply chains. The results seem to indicate that management is provided with insight
into the causes of problems within its systems. Having a measure brings two other
benefits. The first is that it allows comparisons to be made, providing a natural ranking
 Complex systems theory 577

for problems. The second is that the findings are accepted by the staff, as they result from
an objective process.

5 Conclusion

Complex systems theory offers opportunity for the management and development of
manufacturing organizations. The main attraction of the complex systems view is that it
does not offer a single fixed approach by which to study manufacturing, but takes into
consideration the variety of possible states that one could encounter in manufacturing
operations. As Abell and Simmons [45] put it:
“… If the short history of modern research in complexity tells us anything, it is
that multiple perspectives are more powerful than single perspectives …”

This helps managers to devise contingency plans in face of extraordinary operational and
strategic issues.
This paper concludes with a note of caution. It is important to recognize that many of
the complex systems advocates fall into the trap of stretching their observations to fit
nicely the problems of manufacturing and business. The result is an inappropriate
understanding of the problem and the creation of a distorted solution. Circumspection and
educated judgment could therefore be the principle to using complex systems theory
inside the manufacturing organization. One always needs to inspect whether the approach
is either a mere metaphorical representation, and therefore a re-packaged version of
existing management tools or a legitimate model to solving the problem at hand.

References
1 McCarthy, I.P. and T. Rakotobe-Joel, (Eds.) (2000) Complexity and Complex Systems in
Industry, University of Warwick: Coventry, UK.
2 Capra, F. (1996) The Web of Life, 1st Anchor Books.
3 Kuhn, T.S. (1962) The Structure of Scientific Revolutions, Chicago: University of Chicago
Press.
4 Capra, F. (1986) ‘The concept of paradigm and paradigm shift’, Re-Vision, Vol. 9, No. 1, p.3.
5 Hoffman, N. (1998) ‘The unity of science and art: Goethean phenomenology as a new
ecological discipline’, in Goethe’s Way of Science: A Phenomenology of Nature, D. Seamon
and A. Zajonc, (Eds.), State University of New York Press: Albany, NY.
6 Brady, R.H. (1998) ‘The idea in nature: re-reading Goethe’s organics’, in Goethe’s Way of
Science: A phenomenology of Nature, D. Seamon and A. Zajonc, Editors, State University of
New York Press: Albany, NY.
7 Kant, I. (1987) Critique of Judgment, 1790, Indianapolis, IN: Hackett.
8 Devall, B. and George, S. (1985) Deep Ecology, Salt Lake City, Utah: Peregrine Smith.
9 Roque, A.J. (1985) ‘Self-organization: Kant’s concept of teleology and modern chemistry’.
Review of Metaphysics, Sept. Vol. 39, pp.107–135.
10 Bertalanffy, L.V. (1968) General System Theory, (Ed.) Braziller, New York.
11 Bertalanffy, L.V. (1950) ‘The theory of open systems in physics and biology’, Science, Vol.
111, pp.23–29.
12 Prigogine, I. (1967) ‘Dissipative structures in chemical systems’, in Fast Reactions and
Primary Processes in Chemical Kinetics, S. Claesson, (Ed.), Interscience: New York.
578 I.P. McCarthy, T. Rakotobe-Joel and G. Frizelle

13 Wiener, N. (1948) Cybernetics or Control and Communications in the Animal and the
Machine, Cambridge, Massachusetts: Technology Press.
14 Heims, S.J. (1991) The Cybernetics Group, Cambridge, Mass.: MIT Press.
15 Ashby, R. (1952) Design for a brain, New York: John Wiley.
16 Ashby, R. (1956) Introduction to Cybernetics, New York: John Wiley.
17 Weizenbaum, J. (1976) Computer Power and Human Reason, New York: Freeman.
18 Beer, S. (1972) Brain of the Firm: The Managerial Cybernetics of Organization, London:
Allen Lane.
19 Sherman, H. and Shultz, R. (1998) Open Boundaries: Creating Business Innovation through
Complexity, Reading, Massachusetts: Perseus Books.
20 Smarr, L. (1985) ‘An approach to complexity: numerical computations ’. Science, April 26,
Vol. 228, pp.403–408.
21 Shannon, C.E. (1949) The Mathematical Theory of Communication, Urbana, IL: University of
Illinois Press.
22 Frizelle, G. and Gregory, M.J. (2000) ‘Complexity and the impact of introducing new
products’, in Complex Systems and Complexity in Industry, I.P. McCarthy and T. Rakotobe-
Joel, (Eds.) University of Warwick, UK. p.247–259.
23 Frizelle, G. and Woodcock, E. (1995) ‘Measuring complexity as an aid to developing
operational strategy’, International Journal of Operations Management, Vol. 15.
24 Gell-Mann, M. (1996) ‘The simple and the complex’, in Complexity, Global Politics and
National Security, D. Alberts and T. Czerwinski, (Eds.) Washington, DC.
25 Casti, J. (1998) Manufacturing Complexity Network Seminar Report –
http://www.wmg.org.uk/mcn/1stmeeting.pdf, EPSRC Manufacturing Complexity Network.
26 Kauffman, S. and MacReady, W. (1995) ‘Technological evolution and adaptive
organizations’, Complexity, Vol. 1, No. 2 pp.26–43.
27 Battram, A. (1998,) Navigating Complexity, London: The Industrial Society.
28 Gell-Mann, M. (1997) ‘Murray Gell-Mann on the complex adaptive business’, Perspectives
on Business Innovation – http://www.businessinnovation.ey.com/journal/issue1, Vol. 1.
29 Kauffman, S. (1995) At Home in the Universe, Oxford: Oxford University Press, Inc.
30 Waldrop, M. (1992) Complexity, London: Viking-Penguin.
31 Dawkins, R. (1989) The Selfish Gene, Oxford University Press.
32 De Geus, A. (1988) ‘Planning as learning’. Harvard Business Review, March–April,
pp.70–74.
33 Muldoon, M., et al. (1995) ‘Chaos theory in quality control of spring wire – Part 1’. Wire
Industry, June, pp.309–311.
34 Muldoon, M., et al. (1995) ‘Chaos theory in quality control of spring wire – Part 2 and 3’,
Wire Industry, Sept., pp.491–495.
35 Allison, M.A. and Kelly, S. (1999) The Complexity Advantage: How the Science of
Complexity Can Help your Business Achieve Peak Performance. Business Week Books, New
York, NY: McGraw Hill Companies.
36 McCarthy, I., et al. (1997) ‘Building a manufacturing cladogram ’, Int. J. of Technology
Management, Vol. 13, No. 3, pp.269–286.
37 McCarthy, I.P., et al. (2000) ‘Organizational diversity, evolution and cladistic classifications’,
The International Journal of Management Science – OMEGA, Vol. 28, pp.77–95.
38 McCarthy, I. and Ridgway, K. (2000) ‘Cladistics: a taxonomy for manufacturing
organizations’, The International Journal of Manufacturing Technology Management –
Integrated Manufacturing Systems, Vol. 11, No. 1, pp.16–29.
39 Wooldridge, M. and Jennings, N.R. (1995) ‘Intelligent agents: theory and practice’, The
Knowledge Engineering Review, Vol. 10, No. 2, pp.115–152.
 Complex systems theory 579

40 Shen, W. and Norrie, D.H. (1999) ‘Agent-based systems for intelligent manufacturing: a State
of the art survey’, Knowledge and Information Systems, an International Journal, Vol. 1,
No. 2, pp.129–156.
41 Holland, J.H. (1975) Adaptation in Natural and Artificial Systems, University of Michigan
Press.
42 MacLean, D. and MacIntosh, R. (1998) ‘Transforming organizations (some) insights from
complexity theory’, in Strategic Management of the Manufacturing Value Chain, U.S. Bititci
and A.S. Carrie, (Eds.) Kluwer Academic Publishers: London. pp.45–52.
43 McCarthy, I. and Tan, Y.K. (2000) ‘Manufacturing competitiveness and fitness landscape
theory’, Journal of Materials Processing Technology – Special Issue on Agile/Responsive
Manufacturing on a Global Basis, Vol. 107, Nos. 1–3, pp.347–352.
44 Frizelle, G. (1996) An Entropic Measurement of Complexity in Manufacturing Operations –
Research Report, Department of Engineering – University of Cambridge, UK.
45 Abell, B. and Simmons, M. (2000) Review of the Reviews – http://www.santafe-
strategy.com/News_About_SFCES/Emergence/emergence.html. Consulted in March 2000.

View publication stats