A Complex Adaptive Systems Model of Organization Change
ABSTRACT The study of complex adaptive systems has yielded great insight into how complex, organic-like structures can evolve order and purpose over time. Business organizations, typified by semi-autonomous organizational members interacting at many levels of cognition and action, can be portrayed by the generic constructs and driving mechanisms of complex adaptive systems theory. The purpose of this paper is to forge a unified description of complex adaptive systems from several sources, and then investigate the issue of change in a business organization via the framework of complex adaptive systems. The theory of complex adaptive systems uses components from three paradigms of management thought: systems theory, population ecology, and information processing. Specific propositions regarding the nature of dynamical change will be developed, driven by the complex adaptive systems model. Supporting evidence for these propositions is then sought within the existing management theory literature. In doing so, the complex adaptive systems approach to understanding organization change will be better grounded in domain-specific theory, and new insights and research areas will come to light.
- SourceAvailable from: Arwin Van BuurenComplexity and Planning: Systems, Assemblages and Simulations (New directions in planning theory, Edited by G. de Roo, J. Hillier, J van Wezemael, 01/2012: chapter Climate adaptation in complex governance systems. Governance systems between inertia and adaptability: pages 221-242; London: Ashgate Publishing.
- Handbook of research methods in social and personality psychology, 2nd edited by Harry T. Reis, Charles M. Judd, 02/2014: chapter Complex dynamical systems in social and personality psychology: Theory, modeling and analysis.: pages 251-280; Cambridge University Press., ISBN: 978-1107600751
- [Show abstract] [Hide abstract]
ABSTRACT: Our research addresses one of the most vexing issues in IS, that of how IS alignment occurs. Even more vexing, and largely unaddressed, is the issue of sustainable IS alignment. We address sustainable IS alignment as a dynamic, multi-faceted, and non-deterministic process based on the complexity theory worldview. The complexity theory worldview conceives of organizations and IS as complex adaptive systems (CAS) that co-evolve over time. Sustainable IS alignment results when an organization’s complex adaptive IS adapt to remain in alignment with the constantly-changing (that is, evolving) organization’s goals. Our co-evolutionary theory of IS alignment links bottom-up, emergent processes that foster adaptivity with top-down, formal organizational processes essential to established organizations. We illustrate the theory by applying it to the co-evolution and therefore adaptation of enterprise architectures and IS development projects. Our research on the role of co-evolution in sustainable IS alignment contributes to IS research in general and to prior research on IS alignment in particular, and has implications for achieving sustainable IS alignment. We believe that portraying organizations and their IS as complex adaptive systems that co-evolve provides both research and practice with a way toJournal of the Association of information systems. 06/2013; JAIS(14(6)):283-311.
Nonlinear Dynamics, Psychology, and Life Sciences, Vol. 1, No. 1, 1997
A Complex Adaptive Systems Model of
Kevin J. Dooley1,2
The study of complex adaptive systems has yielded great insight into how
complex, organic-like structures can evolve order and purpose over time.
Business organizations, typified by semi-autonomous organizational members
interacting at many levels of cognition and action, can be portrayed by the
generic constructs and driving mechanisms of complex adaptive systems theory.
The purpose of this paper is to forge a unified description of complex adaptive
systems from several sources, and then investigate the issue of change in a
business organization via the framework of complex adaptive systems. The
theory of complex adaptive systems uses components from three paradigms of
management thought: systems theory, population ecology, and information
processing. Specific propositions regarding the nature of dynamical change will
be developed, driven by the complex adaptive systems model. Supporting
evidence for these propositions is then sought within the existing management
theory literature. In doing so, the complex adaptive systems approach to
understanding organization change will be better grounded in domain-specific
theory, and new insights and research areas will come to light.
KEY WORDS: organization development; management; agents; schema; organization learning.
The prevailing paradigm of a given era's management theories has his-
torically mimicked the prevailing paradigm of that era's scientific theories
(Ackoff & Emery, 1972; Hayles, 1991), For example, most of science lead-
1Department of Mechanical Engineering, University of Minnesota, Minneapolis, Minnesota
2Correspondence should be directed to Kevin J. Dooley, Department of Mechanical Engi-
neering, University of Minnesota, Minneapolis, Minnesota 55455; e-mail: kdooley@maroon.
1090-0578/97/0100-0069$12.50/0 C 1997 Human Sciences Press, Inc.
ing up to this century was heavily influenced by the scientific principles of
Newton, LaPlace, and Descartes (Capra, 1982). This paradigm held as prin-
ciple that, for example, the natural state of a system was equilibrium and
that departures from equilibrium would be damped out. Likewise, the ap-
proach by which a system was understood was reductionist and determi-
nistic: through understanding the component elements of a system and the
manner in which they interact, the future states of the system could (theo-
retically) be predicted (Bohm, 1957).
Management theories in the nineteenth and early twentieth centuries
also held reductionism, determinism, and equilibrium as core princi-
ples—indeed, all of social science was influenced by this paradigm (Hayles,
1991). Management theorists such as Fayol, Mooney, and Urwick invented
management control mechanisms that are based on the "organization as
machine" metaphor (Morgan, 1986). Organizational direction is embedded
in plans which are then deployed via planning, budgeting, and manage-
ment-by-objectives systems. A centralized, bureaucratic structure is key in
helping organization leaders determine proper actions and deploy instruc-
tions to the workforce. Control is explicit: monetary rewards and punish-
ments are a common form of motivation.
Reductionism gave rise to ideas such as division of labor, the idea of
task, interchangeability of parts, standard procedures, quality control, cost
accounting, time and motion study, and organizational charts (George,
1968). Frederick W. Taylor was responsible for integrating these ideas with
the concepts of the scientific method to design a coherent management
philosophy. His principles of scientific management have great influence
over management practice of today. Taylor brought analytical logic to man-
agement. Taylor's managerial principles (1911) were summarized into four
points: (a) develop a science for each person's work, (b) train and develop
the workperson, (c) heartily cooperate with others, and (d) divide work
and responsibility between labor and management.
laylor believed in a "social system determinism"—that management
of the organization could be predictable if we understood the science of
management (Dooley, Johnson, & Bush, 1995; Spencer, 1994). Taylor and
others focused on describing work by its elemental tasks, and managing
work as such. Under scientific management, work tasks are divided into
basic skills, and training and standardized methods helped eliminate dif-
ferences between peoples' performance. Organizational controls, such as
budget, performance review, audits, standards, etc., are used as negative
feedback mechanisms for maintaining equilibrium (Leifer, 1989; Spencer,
1994; Wheatley, 1992).
As science has progressed in the twentieth century and different para-
digms of scientific thought have emerged, so have different paradigms of man-
agement theory. While current management practice is still deeply rooted in
the mechanistic approach, it has also evolved and matured via concepts from
other paradigms of thought. The purpose of this paper is to explore how one
specific paradigm of modern science—complex adaptive systems—can be used
to understand organizations and the nature of their change. The next section
explores management theory from the perspectives of systems theory, popu-
lation ecology, and information processing. Section three will explore the his-
tory of complex adaptive systems and how management theorists have used it
to date. Section four forges a composite model of complex adaptive systems,
and specific propositions regarding the behavior of complex adaptive systems
are determined. These propositions are also translated into the domain-spe-
cific language of business systems, and supporting literature is cited. The con-
cluding sections summarize the nature of complex adaptive systems change, a
strategy for change, and design principles.
THREE PARADIGMS OF MANAGEMENT THEORY
Systems theory and cybernetics developed as an important field in the
1940s, via the work of, e.g., Ashby (1958), von Bertalanffy (1968), and Wie-
ner (1948). One of the key purposes of systems theory is to describe a
system in such a manner that automated control mechanisms can maintain
the system's behavior at some desired goal. Fluctuations from the outside
(environment) force the system to adjust in order to maintain equilibrium,
and negative feedback mechanisms dampen the effect of these fluctuations.
Likewise, positive feedback mechanisms, which accentuate fluctuations, can
be used to enhance and accelerate the effects of positive fluctuations.
Systems theory caused management theorists to see the organization
as an "organism" (Morgan, 1986). Fluctuations or contingencies from the
environment are adjusted to by organization change (Lawrence & Dyer,
1983). These environments coexist and to a large extent are created by the
organization—the organization does not play a passive observational role
(Weick, 1979). The nature of change can be strategic (e.g., offering new
product lines), tactical (e.g., developing closer relationships with suppliers),
or cultural (e.g., initiating employee motivation programs).
Contingency theory claims that the manner in which the organization
is organized and functions must correspond to the nature of the environ-
ment which it finds itself in (Lawrence & Lorsch, 1967). In a study of many
different types of firms, Burns and Stalker (1961) found that for organiza-
tions embedded in stable and certain markets (e.g., a textile mill), a mecha-
A CAS Model of Organization Change
nistic organizational form was more appropriate: standardized methods, hi-
erarchical structure, clearly defined lines of authority, communication and
decision making, and clearly bounded roles. For organizations embedded
in unstable and uncertain markets (e.g., an electronics firm), a organic or-
ganizational form was more appropriate: informal and changing lines of
authority, open and informal communication, distributed decision making,
and fluid role definitions.
Other contingency studies have found similar results (e.g.,, McKelvey
& Aldrich, 1983; Miles & Snow, 1978; Pfeffer & Salancik, 1978), showing
contingency effects along the dimensions of environment, industry, strategy,
structure, technology, employee commitment, key beneficiaries, and em-
pirical configurations (Morgan, 1986). For example, Mintzberg (1979) and
Miller and Friesen (1978) investigated structural contingency via defining
five structural forms along the spectrum of mechanistic to organic (machine
bureaucracy, divisionalized form, professional form, the simple form, and
the adhocracy), and found that the fit between structural form and envi-
ronment was key to organizational performance. A hybrid structural
form—the matrix organization, which distributes personnel both along
functional lines as well as within project-related clusters, has become a
popular way of structuring in lieu of uncertain environments (Mintzberg,
In the organic metaphor, organizations evolve according to contingen-
cies in the environment. Change typically enhances complexity via further
differentiation and specialization (being able to handle a wider variety of
tasks) and/or better integration (being able to handle tasks which are more
complex in a more effective manner; Lawrence & Lorsch, 1967).
Rather than viewing the organization as adapting to the environment
in a planned way, "population ecology" believes that organizational attrib-
utes are tested in a Darwinian landscape, where firms with inferior struc-
tures and/or practices die in a resource-constrained competition (Hannan
& Freeman, 1989). Variation in organizational attributes occur in random
and planned fashion, and variations which are beneficial are retained.
Variation is induced by institutionalized experimentation, direct and
indirect incentives, and playfulness; selection is a purposeful managerial
choice of action, based on pre-established goals, values, criteria, check-
points, or competition; useful variations are retained via standardization
and oversight via institutional controls, and internal interpretive schema
change accordingly. The cycle of variation-selection-retention continues on.
This broader view of change allows one to see an organization's change
trajectory in light of larger, systemic issues. For example, several studies
have shown that the organization and its corresponding technology co-
evolve, each affecting one another over time (Rosenkopf & Tushman, 1994;
Van de Yen & Garud, 1994).
The ecological view also generates a hereditary perspective on organ-
izational behavior. Organizational traits can be transmitted vertically for-
ward (e.g., old to young), vertically backward (e.g., young to old), or
horizontally (e.g., young to young) (Baum & Singh, 1994). While the or-
ganizational inheritance process is more complex than its biological coun-
terpart, positive (e.g., McKelvey, 1982) and negative (e.g., Csanyi, 1989;
DiMaggio & Powell, 1983) traits have been observed to linger on in an
organization well beyond any member's stay. Much transmission occurs
within short time frames, because of the ease of iteration and the short
expected lifetime (Cavalli-Sforza & Feldman, 1981).
Population ecologists have often modeled organization change as the
diffusion and adoption of both technical and managerial innovations. The
basic models describing the diffusion process can be characterized by S-
shaped logistics growth curve. These curves describe a rapid initial growth
and a subsequent saturation (or decline) of use. They are based on the
assumption of "substitution," namely, growth in a new technology is linked
to decline of the old, and technologies compete in the environment of a
single, identified need (Cavelli-Sforza & Feldman, 1981). These models are
based on the assumption that information spreads in an infectious manner
via personal contact (Kodama, 1995). Because of pressures toward isomor-
phism, personal contact may lead to imitation of others within and outside
the organization (Kanter, Stein, & Jick, 1992). In such a model, the like-
lihood of adoption increases as the number of other "users" increases, until
usage is "saturated."
Davies (1979), Stoneman (1983), and Kodama (1995) have proposed
that the dynamics of innovation adoptions can be better described by as-
suming that resistance to adoption is caused by institutional (inter- and
intra-organizational) factors rather than an individual's risk aversion or pro-
pensity to imitate. In this model, adoption of the innovation co-evolves with
the organizational environment needed to nurture the innovation (Van de
Ven & Garud, 1992).
The innovation is adopted by only a few individuals at first, because
it is economically advantageous to only a few individuals at first. The in-
dividual makes a behavioral decision of adoption when some "critical"
value has been surpassed (mathematically, a "probit model" emerges). In
some cases, this critical value may be the ratio between the expected "pay-
off period" and the desired pay-off period; the exogenous variable(s) be-
hind the "critical value" needs to be formulated in each particular case.
A CAS Model of Organization Change
Thus, the co-evolutionary model differs in that rather than modeling adop-
tion as an exponential growth over time, it models growth in terms of con-
sequent change in some exogenous factor(s) in the organizational
environment. Dynamically, it may be best described by the dynamics of
punctuated equilibrium (Tushman & Romanelli, 1985; Anderson & Tush-
Innovation diffusion and adoption can also be viewed via a neo-Dar-
winian evolutionary model, where innovations are treated as memes (cul-
tural genes, e.g., Dawkins, 1976), fitness is determined not only by monetary
influx but also by the extent to which the innovation competes for the con-
sciousness of society's members, and the dynamics of innovation diffusion
are best described by models of organic, self-organizing systems. Various
mathematical models have been developed for the vertical, oblique, and
horizontal transmission of cultural traits (e.g., Cavelli-Sforza & Feldman,
1981). These have been applied to study phenomena such as bias trans-
mission from leaders (e.g., Boyd & Richerson, 1985) and meme-gene co-
evolution (e.g., Lumdsen & Wilson, 1981). In the case of cultural
transmission between "parent" (vertical) and "teacher" (horizontal), the
(single) parent can be represented by a so-ranked organizational superior
(e.g., supervisor), while a teacher can be represented by a colleague (su-
perior, peer, or inferior) who has great influence over other individuals.
The mathematical model shows that "transmission through a social leader
or teacher results in greater homogeneity in a population" (Cavelli-Sforza
& Feldman, 1981, p. 339) than transmission through a parental figure. The
parental figure, however, can have a long-lasting effect: the culture created
by the organization's initial leaders forms a "genetic imprint" for the or-
ganization's ontogeny; it will be clung to until it becomes unworkable or
the group fails and breaks up (Donaldson & Lorsch, 1983; Schein, 1992).
At the same time that systems theory was being developed, science
was developing better understandings of the human brain and its associated
cognitive mechanisms. Simon (1947) and others (e.g., Gailbraith, 1974;
March & Simon, 1957) explored the parallel between human and organ-
izational decision making. Simon's concept of bounded rationality posits
that since individuals are limited in their information-processing capabili-
ties, so too are organizations: organizations act on incomplete information,
explore a limited number of alternatives, and do not necessarily develop
accurate cause and effect maps of reality; heuristics dominate organiza-
tional decision making and thus organizational change.
Gailbrath (1974) considered the manner in which organizations adapt
to complex, uncertain environments by changing their information-process-
ing capabilities, either by reducing the need for information (making tasks
and subgroups more autonomous, or creating slack resources) or by in-
creasing the capacity for information acquisition, storage, and retrieval (via
There are four steps by which the organization acquires knowledge
and learns: knowledge acquisition, information distribution, information in-
terpretation, and organizational memory (Huber, 1991). Through such
learning, organizational members, individually and collectively, perceive the
environment around them and take subsequent action. Learning and
change may often be considered dual. Argyris and Schon (1978) differen-
tiate two types of learning: first (or single-loop) and second (or double-
loop) order. An individual's mind stores patterns which govern
interpretation and action as schema; these schema are limited in their re-
alism in the same way that we understand decision-making to be ration-
ally-bounded (Kielser & Sproull, 1982). First-order learning involves
comparison between a perception and an expectation (via the schema); er-
rors are corrected in a simple, cybernetic way through negative feedback.
Most problem-solving hinges on first-order learning: a problem is perceived
as such because the current, observed organizational state does not match
the expected, desired state. Appropriate corrective action is taken.
Second-order learning involves active manipulation and change of the
interpretive schema. One barrier toward such learning is that high skill in
first-order learning—leading typically to rewarded performance and pro-
motion to managerial ranks—can actually detract from the ability to per-
form second-order learning (Argyris, 1991). The nature of schema change,
etc. will be discussed in much greater detail later.
Schon (1975) also identified deutero-learning as an important
skill—the ability to learn how to learn. At a general level, this could include
learning how to: make schema explicit, seek data to test schema, confront
the actual schema ("theory-in-use") with the publicly stated schema ("es-
poused theory"), and a focus on dilemmas (Schon, 1975). In a more specific
example, organizations have learned how to improve one of their key learn-
ing strategies—total quality management—through the use of structured
comparison to prescribed learning systems, e.g., via Malcolm Baldrige Na-
tional Quality Award; Dooley, Bush, Anderson, & Rungtusanantham, 1990;
Garvin, 1991) and the ISO 9000 criteria (Lamprecht, 1992) and benchmark-
ing other "learning systems" (Camp, 1989).
Three types of organization change models have been discussed: or-
ganic (systems theory), organismic (population ecology), and cognitive (in-
formation processing and organizational learning). Others obviously exist
A CAS Model of Organization Change
(e.g., cultural, political, psychic, dialectic, and domination metaphors; Mor-
gan, 1986). These three, however, supply the key elements of the complex
adaptive systems model of organization change: agents scan the environ-
ment and adapt accordingly (organic), using schema to interpret reality and
context, and trigger decisions and actions (cognitive), while competing and
cooperating with other agents for resources and information (organismic).
COMPLEX ADAPTIVE SYSTEMS
The complexity paradigm uses systemic inquiry to build fuzzy, multi-
valent, multilevel and multidisciplinary representations of reality. Systems
can be understood by looking for patterns within their complexity, patterns
that describe potential evolutions of the system. Descriptions are indeter-
minate and complimentary, and observer-dependent. Systems transition
naturally between equilibrium points through environmental adaptation
and self-organization; control and order are emergent rather than hierar-
chical (Dooley et al., 1995; Lewin, 1992; Waldrop, 1992).
Core epistemological concepts from the mechanistic era were seriously
challenged by Einstein's theory of relativity, and by the discovery of quan-
tum mechanics (Bohm, 1957; Gribbin, 1984). As Kuhn (1970) points out,
while paradigmatic revolutions are discontinuous change, they typically do
not occur at one instant in time. Instead, "change occurs through negotia-
tions at multiple sites among those who generate data, interpret them, theo-
rize about them, and extrapolate beyond them to broader cultural and
philosophical significance" (Hayles, 1991, p. 4). The language of the old
paradigm is likely to be used in the new paradigm, causing further confu-
sion (Hayles, 1991, p. 4; Kuhn, 1970, p. 149). In the Newtonian paradigm,
for example, equilibrium is considered the natural state of the system (Kel-
lert, 1993); in the complexity paradigm, equilibrium is just one of several
states possible—whether equilibrium is natural is situational.
A theory of complex adaptive systems was borne from the discovery
of chaotic dynamics in systems' behaviors. Chaos theory has developed
along two dimensions. Experimentalists (as popularized in Gleick, 1987)
found ways (primarily grounded in topology) to discover deep and complex
patterns in seemingly random or "chaotic" data. Prigogine and Stengers
(1984), among others, use chaos to describe how order can arise from com-
plexity through the process of self-organization. Here is a summary of some
of the main characteristics of systems described by chaos theory (Dooley
et al., 1995): (a) seemingly random behavior may be the result of simple
nonlinear systems (or feedback-coupled linear systems), (b) chaotic behav-
ior can be discovered via various topological mappings, (c) nonlinear sys-
tems can be subject to sensitive dependence to initial conditions—this
sensitivity forces a re-examination of causality—which now must be con-
sidered multilevel and multideterminate (Abraham et al., 1990), (d) systems
that are pushed far-from-equilibrium (at the edge of chaos) can spontane-
ously self-organize into new structures, and (e) changes in the essential
nature of a system take place when a control parameter passes a critical
threshold—a bifurcation point.
While the concepts of chaos and self-organization have evolved from
the physical sciences, the notion of complex adaptive systems (CAS) has
its roots in the biological sciences (Gell-Mann, 1994). Whereas chaos theory
relates to a particular behavior of complex systems, complex adaptive sys-
tems theory allows one to analyze the organizational system from a more
holistic point of view. A CAS is both self-organizing and learning; examples
of CAS include social systems, ecologies, economies, cultures, politics, tech-
nologies, traffic, weather, etc.
The first managerial applications of complexity theory were driven by
Maturana and Varela's work on autopoiesis. In the theory of autopoiesis
(Maturana & Varela, 1992), structural change occurs through self-renewal
(replication, copy, and reproduction). This self-production produces a his-
tory of change: "Ontogeny is the history of structural change in a unity
without loss of organization in that unity. This ongoing structural change
occurs . . . either as a change triggered by interactions coming from the
environment in which it exists or as a result of its internal dynamics . .
. . The overall result is that the ontogenic transformation on a unity only
ceases with its disintegration..." (p. 75). The ontogeny of the autopoietic
system is contained in its internal components and their interactions; the
environment only triggers change; the possible futures are encoded inside
the system at the time of trigger (Maturana & Varela, 1992). This is in
marked departure from previous organic models of organizations, which
modeled the interface between organization and environment as more
Wheatley (1992) recognizes the bridge between the companies focused
on core competencies (Prahalad & Hamel, 1990) and autopoietic organi-
zations. She states that by structuring around skills instead of business units,
new business opportunities can be more rapidly responded to; such organi-
zations exemplify the nature of the autopoietic system to be highly self-re-
flective (but not ego-maniacal).
A CAS Model of Organization Change
A second avenue for studying organizations as complex systems was
via the work of Forrester (1961) in system dynamics. The purpose of system
dynamics was to use simple differential equations—as used in the cyber-
netic control of electromechanical systems—to model large-scale systems,
including economies (Roberts, 1963), social organizations (Simon, 1952),
and culture (Levin, Hirsch, & Roberts, 1972). Complexity, in the form of
"surprising behavior," was found in many of these studies. Most users of
system dynamics models still do not incorporate nonlinear behavior into
their models; as mentioned before, CAS are typified by nonlinear flows of
information and resources. As such, these linear dynamical models typically
fall short of discovering truly complex behavior. There have been a number
of researchers (e.g., Allen & Sanglier, 1981) who have used cellular auto-
mata-type simulations instead of system dynamics, and have discovered the
complex clustering phenomena found in CAS.
Prigogine and Stenger's work on dissipative systems (1984) inspired a
number of management writers to posit how the theory of far-from-equi-
librium systems could be used to describe, and perhaps prescribe, organi-
zation change (Guastello, Dooley, & Goldstein, 1995). Leifer (1989),
among others, recognized that far-from-equilibrium conditions can be gen-
erated by trigger events such as crises (e.g., impending competitive death
of Xerox) or leader-declared revolutions (e.g., GE's Jack Welch) (Tichy &
Ulrich, 1984). Unlike first-order adaptations, these crises overwhelm the
organization's normal capacity for change. At times, the leader may even
invoke artificial crises, such as through creation of a challenging vision, in
order to move the system away from equilibrium (Nonaka, 1988).
Far-from-equilibrium conditions tend to create a dynamic stability
where paradox abounds. The paradoxical nature of CAS—between ran-
domness and order, freedom and control, learning and unlearning, adap-
tation to the environment and construction of the environment—has not
gone unrecognized. Wheatley (1992) and Stacey (1992) advocate purposeful
surfacing of these tensions, in order to generate far-from-equilibrium con-
ditions which will hopefully lead to creative new solutions. Stacey goes so
far as to say that a manager's job is to allow political forces to challenge
one's own source of power and control—the tension between the mecha-
nistic management of the status quo and the organic management of the
learning organization allow the system to thrive. Dooley et al. (1995) have
pointed out the paradoxical nature of total quality management programs,
with the simultaneous focus on control and learning.
Goldstein (1990, 1994) has developed specific techniques for moving
a group "far-from-equilibrium"; once there, the potential for re-organizing
around a new mode of behavior can be realized. Of course, one cannot
control the end result of the self-organization; it can only be influenced by
shared value and purpose (Dooley et al., 1995). Goldstein advocated the
following ways of generating far-from-equilibrium conditions in business or-
ganizations: (a) work with (define, discuss, change) organizational bounda-
ries; (b) connect the system to its environment (customers, suppliers,
competition); (c) difference questioning (seek divergence in group discus-
sion; method based on similar approaches first developed in family systems
therapy (Palazzoli, Boscolo, Cecchin, & Prata, 1980); (d) purpose contrast-
ing (heightening awareness of the state gap; Kiesler & Sproull, 1982); (e)
challenge self-fulfilling prophecies; (f) challenge assumptions creatively
(e.g., Argyris & Schon, 1978); (g) develop nonverbal representations of the
system, such as Morgan's (1986) "imaginization"; and (h) take advantage
of chance (e.g., statistical methods which generate knowledge via outliers).
Eoyang (1993) has developed tools which help accentuate differences
in systems, particularly across organizational boundaries. Group communi-
cation patterns are diagnosed into a 2 x 2 matrix corresponding to the
dimensions of high or low transfer of information, and high or low differ-
entiation; self-diagnosis is employed. Others have employed organizational
"games" to develop (at least) self-reflection, if not far-from-equilibrium
conditions (e.g., Guastello, 1995).
A number of writers have focused on the "chaotic" nature of a change
event. Tushman, Newman, and Romanelli (1986) noted that successful,
large-scale organizational change was typically followed by prolonged pe-
riods of stasis, or incremental change. In CAS terms, this phenomena is
called "punctuated equilibrium" and it has roots in the evolutionary studies
of ecologies (Gould, 1989). Similarly, Gershick (1988) found that work
teams go through similar developmental stages: inertia and revolution in
attitudes and behaviors alternated, often triggered by real and perceived
deadlines. The key theme in punctuated equilibrium is that second- (or in
some cases, third) order change does not need to be invoked by cata-
strophic events. Change occurs when the system has evolved far-from-equi-
librium, which could come from an accumulation of small perturbances or
the cascading, compounding effect of a small disturbance while the system
A CAS Model of Organization Change
is hypersensitive to such disturbances (e.g., operating in a chaotic regime
The unpredictability of CAS systems, especially when they are in cha-
otic regimes of behavior, has been noted by strategic management writers
(Cartwright, 1991; Kiel, 1994; Levy, 1994; Mintzberg, 1994; Stacey, 1992;
Wheatley, 1992; Zimmerman, 1993). Stacey advocated an open-ended plan-
ning process, whereby strategic issues are formulated and then exposed to
divergent thinking and group dialogue.
A number of researchers have explored organizational change in a
more quantitative fashion by applying the statistical modeling techniques
of nonlinear dynamical systems (e.g., fractal dimension, Lyapunov expo-
nent; Peitgen, Jurgens, & Saupe, 1992) to organizational performance data.
The goals of such analyses are to: (a) determine whether apparently ran-
dom data has embedded structure, (b) determine the complexity of the
system (the fractal dimension of the data's attractor corresponds to the
(sub)system's degrees of freedom), (c) potentially develop short-term pre-
dictive models, and (d) explore nonlinear causalities. A number of studies
in different areas have been done: Guastello (1995) has examined a mul-
titude of organizational phenomena, including decision making, stress and
human performance, accidents, creativity, and group development. Pries-
meyer (1992) used phase plane analysis to look at applications in market-
ing, finance, and production. Kiel (1994) looked at nonlinear relationships
(in public management) between work effort and service requests. Gresov,
Haveman, and Oliva (1993) used catastrophe models (Guastello, 1995) to
study the nature of strategic competitive response. Numerous studies have
looked at chaotic dynamics in economics (see Guastello, 1995).
In order to demonstrate the type of analyses being attempted, works
investigating the dynamics of production systems will be detailed here.
Chase, Serrano, and Ramadge (1993) have found possible evidence for
nonlinear, chaotic conditions in a simple, deterministic, switched-flow
model. Huberman and Hogg (1988) have found point, periodic, and chaotic
attractors in scenarios relating to distributed production control systems.
Lin and Solberg (1987) observed chaos in the production rate of an as-
sembly system when a delay is increased.
Beaumariage and Kempf (1994) have done a series of experiments using
a simple, deterministic manufacturing system model. It consists of four ma-
chines and four part types, and some simple part arrival, batching, and sched-
uling rules. They have found eight and ten distinct attractors (looking at
interdeparture times) in two different models. The attractors are periodic,
ranging in periodic length from six to 1278. By changing some simple pa-
rameters, e.g., the release size, they observed changes in both period length
and shape. Small changes in queue content, at certain times, were shown to
Fig. 1. Chaotic behavior in a manufacturing system.
create great changes in output performance, and indeed even shift the system
from one attractor to another. It is not clear whether this sensitivity to initial
conditions is in fact exponential in nature, which would be a prerequisite to
indicate chaos (as opposed to complex periodic behavior).
Deshmukh (1993) observed part interdeparture times from a multima-
chine workcenter under three different sequencing rules, shortest process-
ing time (SPT), longest processing time (LPT), and first in first out (FIFO).
An analysis of return maps, indicates that the FIFO rule induces chaos in
the output process while LPT results in periodicity. Periodicity was found
to be particularly useful in reducing system performance variance.
For general queuing systems, Erramili and Forys (1991) have found
periodic and chaotic attractors in a model of a telephone switching system.
Leland, Taqqu, Willinger, and Wilson (1993) claim to find fractal-like struc-
ture in ethernet traffic.
Dooley (in press) simulated a simple manufacturing system and dis-
covered low dimensional chaos in the throughput times of the simulated
production facility. Chaos appeared more likely in congested systems whose
scheduling schemes do not alter queue rank. The distribution of throughput
times resembles the dual-mode distribution generated via the mushroom
catastrophe dynamics (Guastello, 1995). A return map of the data, shown
in Fig. 1, shows the strange attractor indicative of chaotic behavior.
The area of organizational innovation has also been one studied in
some detail, in terms of change dynamics. The process of innovation has
been characterized as "being inherently uncertain, dynamic, and a ... ran-
dom process" (Cheng & Van de Ven, 1994; cf. Jelinek & Schoonhoven,
1990). Individual, chance events play a large role in the discovery process.
A CAS Model of Organization Change
History is littered with inventions brought about by a researcher's "mistake"
(e.g., post-it notes). Yet there is evidence that some degree of order may
exist, implicit to the observed randomness, suggesting that such systems are
operating at the edge of chaos (i.e., they have chaotic dynamics of low-di-
Cheng and Van de Ven (1994) and Koput (1992) have found numerical
evidence for chaos at the beginning of the innovation process. Leonard-
Barton (1988) found evidence for a "nonlinear process involving complex
recursive cycles of adaptation" (Jayanthi & Sinha 1994) in a series of case
studies. The studies of Tyre and Orlikowski (1994) suggest that random
events in the innovation process trigger adaptive cycles. Jayanthi and Sinha
(1994) have also found numerical evidence of chaos in the activities bridg-
ing innovation and production.
CAS AND ORGANIZATION CHANGE: THEORETICAL
An important analysis of chaos and organization theory was recently
performed by Thietart and Forgues (1995). Many of the ideas presented
in this paper can also be found in that paper; the authors developed six
1. Organizations are potentially chaotic.
2. Organizations move from one dynamic state to the other through
a discrete bifurcation process (second-order change).
3. Forecasting is impossible, especially at a global scale and in the long
4. When in a chaotic state, organizations are attracted to an identifi-
able configuration (order out of randomness).
5. When in a chaotic state, similar structure patterns are found at or-
ganizational, unit, group, and individual levels (fractal nature of cha-
6. Similar actions taken by organizations in a chaotic state will never
lead to the same result.
The authors state that the organization may enter a chaotic state when
system variables are highly coupled, system variables follow different peri-
odic patterns, and the system is composed of a large number of counter-
acting forces. One subtle yet important distinction that must be made is
that chaotic behavior is not an attribute of the system per se; rather, chaos
is an attribute of some aspect of the system's behavior (Kellert, 1993). The
state of the system can be described by numerous performance charac-
teristics. The chaotic behavior of one characteristic (e.g., the firm's daily
stock value) is not an indication that other characteristics will behave simi-
larly (e.g., the firm's productivity levels). In fact, chaotic behavior is not
even guaranteed to remain constant across scale (Mandelbrot, 1983); while
daily stock prices may vary chaotically, weekly prices may vary periodically
(Peters, 1991). That said, propositions 2-6 remain true in the more general
case of a CAS. The general state of far-from-equilibrium conditions present
in living systems is enough to engage these propositions.
Before business organization-specific propositions can be developed,
a generic model of CAS must be presented. Such a model does not exist.
A number of people have put forth particular models of complex adaptive
systems: Gell-Mann (1994), Holland (1995), Jantsch (1980), Maturana and
Varela (1992), and Prigogine and Stengers (1984). The essential principles
of CAS has been taken from each of these works and synthesized into a
A CAS behaves/evolves according to two key principles: order is emer-
gent as opposed to predetermined, and the state of the system is irre-
versible and often unpredictable. One way to demonstrate this property of
emergence is to visually study cellular automata (CA) (Prigogine & Steng-
ers, 1984). CA is the computer-simulated embodiment of a complex (al-
though not necessarily adaptive) system; this demonstration will also help
introduce some of the basic behavioral engines in CAS.
Two-dimensional CAs are represented by a grid of cells (in the com-
puter each cell is defined to correspond a particular number of pixels in
size), and the state of each cell, in the simplest case, is binary. In a com-
puter simulation, the state of the cell in the succeeding moment is deter-
mined by the state of the cell in the previous moment, as well as the state
of neighboring cells in the previous moment. On the computer, each binary
state (e.g., on or off) can be represented by a color (e.g., white=on,
black=off), and the state of all cells is updated during each iteration.
Figure 2 shows a CA whereby the middle cell behaves according to
the rule: if three of less of my neighbors are "on," turn "off'; otherwise,
turn "on." In this case, since only three of the neighbors are "on," the
middle cell turns "off." To demonstrate the emergence of order in a com-
plex system, a large grid is generated and filled randomly: 50% of the cells
are "on," and an equal number "off," although in no spatial pattern what-
soever. The grid on the right of Fig. 3 represents this (these figures were
generated using the software program CASim; Karakatsios, 1990). Next,
each cell is allowed to evolve according to a simple behavioral rule, called
vote (Karakatsios, 1990). If three or less of your neighbors (including your-
self) are on, turn off; if six or more are on, turn on. This is simple majority
rule. The twist is near the tie: If four neighbors are on, turn on; but if five
A CAS Model of Organization Change