118
‘TENSIONS AND INTERFACES
IS SYSTEM DYNAMICS THEORY COMPLETE?.
Xenyon B. De Greene
University of Southern California*
ABSTRACT
System dynamics consists of a body of theory, philosophy,
metkodology, policy-related applications, and experience. . Basic to
system dynamics is the theory of the semi-closed, fully closed-loop
system in which the feedback loop is the principal construct. In the
20 years of its existence, major emphasis has been placed on the me~
noéology of model-building, on applications, and on philosophical
detates involving alternative approaches, particularly the static
econometric approach. Experience has produced improvements in the
original theory. However, feedback loops are not the only constructs
for dynamic theory-building, and cybernetic, self-regulating systems
are not the only kinds of living systems, nor is the cybernetic per~
spective invariably the only or most appropriate perspective over the
life hiatory of a particular system. The processes of self-organi-
zation and the emergence of new structure deserve equal attention
in the evolution of systems.
This payer briefly reviews the history of system dynamics.
An analysis is then made of present system dynamics theory. This is
followed by summaries of three field theories—-of critical phenomena,
catastrophe theory, and dissipative structures--and attempts at syn-
“correspondence and requests for reprints should be directed
to 4345 Chaunont Road, Woodland Hille, California 91364. Or tele~
phone (213) 340-5199.
-2-
thesizing these theories and system dynamics. Then ways of enrichen-
ing existing system dynamics models with fuller use of knowledge
from behavioral/social science and aociotechnical systems, with par-
ticular relevance to the National Model, are discussed. The paper
concludes with an identification of three immediate next steps in
research.
INTRODUCTION.
It is now 20 years since the appearance in 1961 of Jay W,
Forrester's Industrial Dynamics [7]. System dynamics theory is the
outgrowth of a marriage among cybernetic (information-feedback) theory,
practical experience with the management of large organizations, and
the computational power of large digital computers. The early use
of system dynamics models, based on the original theory, in turn ge-~
nerated further theory about the behavior of complex systema and or-
ganizations, especially in the context of policymaking. This second
arily-derived theory includes the well-known counterintuitive behavior
of complex systems, resistance of behavior to most parameter changes,
unexpected sensitivity of behavior to some parameter changes, etc.
It would appedr, however, that advances in the theoretical under-
pinnings of system dynamics modeling have been few in recent years.
Nevertheless, over the years a number of real or imagined
limitations have been ameliorated. For example, statistical methods
like GYPSIE have been developed for system dynamics models. Proba=
bilistic system dynamics is an interesting blend of a system dynamics
model with a FORTRAN-based event-on-trend, event-on-event, and trené-
on event cross-impact model.
Much of the debate expressed in the large literature, though,
3+
seems to be framed as follows: given the basic theory of system dy~
namics, let us concentrate on the methodology. The fundamental ques—
tion as to the completeness of system dynamics (or alternative) the-
ory remains unanswered. Little argument can be presented against
the pervasiveness of feedback processes in both living and nonliving
systems. Feedback processes, however, are not the only processes
immanent in these systems. Thus, the basic cybernetic theory of
which system dynamics is one important part is excellent for des-
cribing how systems behaved the way they did in the past, behave the
way they do now, and will behave in the future--given the same kinds
of historic or ongoing processes that change only quantitatively.
System dynamics cannot handle emergence, that is, qualitative recon-
figurations or non-preprogrammed changes in patterns, Neither can
any other large computer-simulation methodology. The theory of emer-
gence is just now being developed.
Tke present author hypothesizes that the evolution of phy-
sical, biological, and social systems is characterized by stages in
which different kinds of processes may be predominant. These stages
differ by: (1) the rates of change, for example, quasilinear, ex-
ponential, or nyperbolic; (2) the creation or destruction of order
or form (anabolic and catabolic morphogenesis); (3) the relative im-
portance of deterministic versus stochastic factors; (4) the kind of
dynamic process that is dominant; and (5) the qualitative nature of
the emergent pattern, In short: these systems are both self-regu-
lating systems and self-organizing systems, and dominance of a given
kind of process differs according to evolutionary stage. System
dynamics at present does not appear to be equipped to deal with the
different stages of evolution even though Forrester and others
119
-4-
(including the present author) believe that the world has entered a
atage of major transition or transformation. Thus, system dynamics
societal systems modele must be geared to a fuller range of proces-
ses if they are to have the desired effectiveness in futures research,
long-range planning, and policymaking.
A great deal of effort haa in particular gone into the com-
parison of system dynamics with econometrice/regression analysis.
The present author accepts the superiority of system dynamics but
ake: given the emphasia of the dynamic over the static, whither now?
The next section discusses a number of apparent limitations
of system dynamics theory. In the third section, particular empha~
sis is placed on critical phenomena, catastrophe theory, and the the-
ory of self organization (theory of dissipative structures) under
conditions far from equilibrium. Considerable attention is paid
in the fourth section to the need for better understanding of be-
havioral/social science and sociotechnical processes and constructs.
Ways in which these theories can enrichen system dynamics, as well
as possible irreconcilable differences between system dynamics and
alternative dynamic theories, are considered throughout the paper.
The fifth and concluding section presents suggestions for further
research.
This article represents a continuation of a long-term re-
search effort directed toward societal and organizational systems
theory~building and toward the improvement of computer simulation
modeling used for long-range planning and policymaking. Several re-
ferences present earlier thinking and a detailed céverage of the
original literature (1], [2], [3], [4], [5], [6].
—
ANALYSIS OF PRESENT SYSTEM _D" ICS THEORY
This section discusses areas that have not yet received much
eritically analytic focus and in particular areas that are directly
related to the new field theories to be emphasized in the following
sections
Exogerously versus Endogenously Induced Behavior
System dynamics stresses the closed-system or closed~loop~
system characteristics of its models. Unfortunately, the thermody-
namic term "closed" and the communications-theoretic term "closed
loop" are often used interchangeably. System dynamics models can-
not be closed, nor can any other model of complex systems because
matter and energy cross the system boundary. Consider, for example,
the cloud symbols representing sources and sinks. The external en-
vironzent itself is ever changing; for instance, the flow of manpow-
er changes qualitatively over time as a function of forces outside
‘stem boundary. Most system dynamics models are, however,
closed-loop systezs in that no feedback loops are allowed to cross
the system boundary, Exogenous variables, if used at all, are consi-
dered to be temporary pending further understanding of the real world.
This theoretic approach does not appear, however, to circum
vent a more serious question concerning the origin, external or in-
ternal, of behavior in realworld systems. All systems can be affeo-
ted by external perturbations or stimuli. But one class of systems,
living systems, is capable of self-regulating, self-maintaining,
and self-organizing behavior. System dynamics models apply to liv-
ing systems and describe aptly most features of self-regulation and
self-naintenance. Once the system boundary is determined and the
-6-
minimum number of relevant variables and feedback loops incorporated
within that boundary, the system strives to regulate and control the
interactions of its variables within tolerable limits and to main-
tain itself as a viable system in spite of external perturbations.
Sometimes rates of change are so great or frequencies or periods of
oscillations such that limits to the capability for regulation and
control are exceeded, and one or more levels essentially collapse.
But system structure does not change, nor do new behaviors arise
spontaneously within the system. Thus, system dynamics does not
capture the self-organizing properties of realworld living systems.
Self-organization appears to be especially related to processes of
fluctuation, threshold, and discontinuity.
Fluctuation
The term "fluctuation" is widely used in the literature of
dynamic systems including the eystem dynamics literature. In system
dynamics the term is applied both to the high-amplitude, low-frequen-
cy oscillations so characteristic of system dynamics models and to
the exogenous random-noise inputs. But these applications capture
neither the spontaneity nor the changes in likelihood in realworld
systems, especially antecedent to reconfiguration. Contraat two
realworld situations. In one case, frequencies of internal cycles
“in organisms are sensitive to and even become entrained with the fre~
quencies of environmental stimuli, The deleterious physiological
and performance responses of human operators to the frequency ranges
of vibration and the many forme of biorhythms provide examples.
This situation appears consistent with the system dynamics explana-
tion of the effects of random noise on the system. The other case
-T-
involves the appearance of genetic mutations and of human discove~
ries, innovations, and "great men." Consideration of the probabil~
ty of occurrence, and the subsequent amplification or damping, of
such fluctuations is of utmost importance to the evolutionary theory
of organisas, organizations, and societies. Probabilities of oc-
currence do not appear to be constant over time or place and appear
to speed up near critical points or thresholds. The fluctuations
can be expressed in terms of much greater variety in sizes, struc-
tures, and behaviors. In addition, probabilistic behavior, once a
critical threshold is exceeded, triggers deterministic behavior.
Discontinuity
Enough examples are now recognized from the behavior of non-
living systems and from organic evolution, human history, and current
human individual human and societal behavior to characterize discon-
tinuity as an inherent feature of system behavior. The importance
of discontinuity to modeling theory appears to have been lost in the
controversy between continuoue and discrete schools of simulation
modeling. System dynamics is a continuous modeling methodology uti-
lizing differential equations approximated by difference equations
so as to meet the requirements for discreteness in digital computerse
System dynamics handles discontinuities in two ways: (1) by step-
function inputs and (2) by CLIP functions internal to the model.
But these means do not capture the poorly anticipated suddenness of
Trealworld discontinuities. The step inputs mask antecedent changes,
and the CLIP functions require preprogramming. Thus, while system
Gynamice realistically descrives many aspects of past behavior, its
usefulness as a predictive modeling methodology is still severely
2d.
constrained.
The Continuity of Evolutioy and torical Processe:
The concepte of stochastic evolutionary and historic stages
intercalated with determinietic stages, discontinuity, fluctuation,
and self-organization in open living syateme suggest that, although
there is continuity of basic processes across time in general, dif-
ferent periods exhibit a dominance of different processes. In other
words, feedback dynamics cannot portray or substitute for the entire
repertory of evolutionary/historic processes. Several observers
have commented on the importance of recognizing stages of transition
or transformation. System dynamics describes these stages well un~
der certain circumstances, for example, in a world described by a
logistic curve. A stage of slow, then rapid positive-feedback-based
exponential growth ie followed by a quasilinear stage in which the
forces of growth conflict with the forces of competition, exhaustion,
or negative-feedback-based regulation and control, which is in turn
followed by a stage of growth the rate of which through regulation
decreases toward some asymptotic value. Repeated competition be-
tween growth and regulation, involving two or more levels, of course
produces the familier oscillations of system dynamics models.
However, exponential, logistic, and sinusoidal constructs,
for several reasons, do not accurately represent all reality. First,
analysis of person-artifactsand person~innovatione from the earliest
Paleolithic to the present indicates that the rate of growth is hy-
per-exponential or hyperbolic rather than exponential. Second, the
envelope curve of successive logistic stages shoots through the
local ceilings or limite producing an acceleration of evolution and
-9-
history, Third, the local stages of such intense interaction between
society and technology are separated by. platforms characterized by
slight change and high degree of stability, end these platforms
separate qualitatively different forms. These platforms therefore
often represent discontinuities following the exhaustion of potential
for further innovation. On the surface it may be quite difficult to
differentiate between a period around the inflection point within a
local logistic stage and a period of slow change between successive
local stages. It may be likewise difficult to distinguish between
a local limit to growth and an absolute limit described by the en-
velope curve.
Eouilibrium
The concept of equilibrium is widely employed in the scien-
ces in both the static and dynamic senses. In economics the concept
has had a particularly invidious impact, for example, that supply
will adjuet or can be made to adjust to demand and vice versa. John
Maynard Keynes stressed stimulating the growth of the economy (and
the zoney supply) through increasing demand via government inter-
ventions and thereby reducing unemployment. Recently, he has be~
come known as the "father of modern inflation." The present Reagan
Administration speake of supply-side-economics. Presumably, remo~
val of government regulations coupled with tax cuts, reductions of
spending for supposedly unproductive purposes, and tight control of
noney supply will provide the incentive for private business and
industry to increase production so as both to increase the ratio of
goods and services to money and to create enough new jobe signifi-
cantly to reduce unemployment. Although these alternative policies
ave had and will continue to have a profound effect on the world
10-
economy, they are based on limited and increasingly obsolete and in-
appropriate thinking, and theory.
System dynamics deals with both equilibrium and disequili-
brium conditions. Equilibrium is defined in terms of the equality
of value of level variables, which is usually an initial condition
or a condition when exploding oscillations and asynchrony among vari-~
ables have been brought under control via the proper policies.
Even the aystem dynamics view may be limited, however. A
system may display two or more levels or surfaces of equilibrium and
may suddenly jump, drop, or flip from one level to another. There
may be no transition into equilibrium. In some systems these changes
are essentially irreversible, for example, in the case of apecies
hovering around the threshold of extinction. Systems also can func-
tion far from equilibrium where the results of perturbations and
fluctuations can be the emergence of qualitatively new structure.
At-a time of apparent societal transformation, socioeconomic poli~
cies intended to reestablish equilibrium between supply and demand
may instead trigger massive and irreversible unemployment, destruc~
tion of natural resources, civil disorder, or war.
Complete equilibrium and stability are very likely not even
desirable policy goals. They reduce ayatem resiliency and therefore
capability to learn and adapt [9]. Just how much slack to leave in
the system hence becomes an important policy consideration.
The Time Frame of Models
Syatem dynamics provides great flexibility in the choice of
solution intervals and of total time simulated. Short intervals of
seconds or subseconda can be applied to the simulation of metabolic
lie
and other physiological processes. However, social system models
have typically involved total times of hundreds of years. These long
run-times may be unrealistic for two reasons discussed above: (1)
the question of qualitative continuity of evolutionary-historic pro-
cesses and (2) the increasing evidence of world society's being in
a stage of major transformation. If these points be granted, then
it seems unlikely that the same variables will persist and that the
same feedback loops will extend beyond the imminent discontinuities.
One can think of many examples of recent or imminent system change
where purely system dynamic models would be inappropriate--the Iranian
revolution, the Polish worker strikes and the emergence of Solidarity
and governmental reforms, the 1980 Miami race riots, the 1977 New
York City power failure followed by blackout looting, the summer 1981
civil disorders in British cities, and the situation in American pri-
eons and with the “criminal justice system" in general.
Given the immense turbulence of the world today, it would ap-
pear that the simulation run-time of the National Model should end
around the year 2000. Deserving particular attention is the question
as to whether the present world transformation, if that be accepted,
can be portrayed solely by the concept of a downturn in a Kondrati-
eff cycle. One must remember that the last downturn~-the 1929 stock
market crash and the Great Depression—-was accompanied not only by
an excess of capital, too rapid growth, a high degree of speculation,
and other economic factors but also by dramatic social and political
changes in the United States, Germany, Japan, Italy, and elsewhere.
Some authors argue that only the quick, positive, and supportive ac-
tiona of Franklin D. Roosevelt (a fluctuation of the "great man"
ind) and his administration prevented a revolution in the United States.
123
12-
Feedforward
Realworld systems show the capability to anticipate and to
learn; hence, in many cases involving the future, feedforward may be
a better construct than is feedback. Rational expectations theory
in economics provides one example. People, once disappointed by go=
vernment policies, anticipate and take counteractions egainst a re-
pitition of these policies.
NEW THEORIES AND CONSTRUCTS
In the last several years theoretical advances in physics,
physical chemistry, and topology have provided new insights as to how
aynanic systens change structure and organization qualitatively. These
advances stem from the study of: (1) critical phenomena in physics,
(2) sudden changes beyond discontinuities (catastrophes), and (3) self-
organization (dissipative structures) through fluctuations in open
physicochemical systems far from equilibrium. Collectively, these
developments represent advances in field theory, that is, the theory
of how structure and behavior depend on the interplay of forces in
@ field rather than upon the specific properties of the elements of
the system. Importantly, a number of theoretical interpretations
of ecosystem and societal aystem evolution and behavior have been
based on these theories and constructs.
Critical Phenomena
The critical phenomena are the qualitatively different orga-
nizations and behaviors on either side of a critical point or thresh-
old, say, of a temperature. Examples include liquid-gas phase tran-
sitions, ferromagnetism, metel alloys, and miscible-immiscible fluids.
Quite dissimilar substances dieplay strikingly similar behaviors.
-13-
The most important construct that can be “applied to complex biolo-
gical and social systems is the critical point, usually a catastro-
phic point, and the nature and frequency of fluctuations in the vici-
nity of this point. Far from the critical point there ia randomness,
closer to the point short-range order emerges, and near the point oc-
cur a great number and variety of fluctuations, for example, in the
scale of organization. Within a given scale of organization are in-
cluded smaller scales, and in turn a given scale is incorporated in-
to even greater scales. local or short-range forces have produced
long-ranze order (correlation), often through neighbor-to-neighbor
impacts. This type of system change does not appear to involve feed~
back. In the case of a ferromagnet, all electron spine may be align-
ed so as to favor magnetism, but the system remains unmagnetized
until after the Curie point has been passed from above. Thus, struc-
tural change is antecedent to behavioral change. The system is set
to reconfigure but needs one last push.
, Of course no serious systems theorist would propose a reduc-
‘tion of the behaviors of complex biological and social systems to
laws governing only the particles of physics. Nevertheless, there
does appear to be increasing evidence for the universal applicability
of several natural laws to different hierarchical levels of organi-
zation of matter, energy, and information. But even if the construct
of critical thresholds of qualitative reconfiguration turns out to
be no more than a guiding metaphor when applied to biological and
social change, its present heuristic value should be evident. Thus,
at the phenomenological level, there do appear to be both a higher
frequency, greater amplitude, and greater variety of fluctuations
and of exergent forms now than wae the case in the two decades or
124
othe
so following the end of World War II, Empirical evidence for this
increased intensity, pace, and variety of change can be seen in human
life styles, religious beliefs, cults, drug use, book publishing,
television, crime and violence, economic indicators, government chan=
ges, insurrections, land use, and species brought to the brink of
extinction. A simple list of course saya nothing about the causal
nature of each phenomenon or event. One could construct all sorts
of diagrams showing mutual causality among these and other phenomena,
but the diagrams could misrepresent the real world greatly because
the causal pathways are open-loop and irreversible or because the
listing gives symptoms rather than underlying causes, Some quite se-
rious world changes are in these categories. Land use and abuse
exemplifies the former.: Land use produces habitat destruction, which
yields species extinction with no feedback loops involved. “In the
second cage, the underlying force may be cultural old age and ex-
haustion, with the fluctuations’ representing the erosion and break-
ing of restrictive bonds in the decaying culture and both last-gasp
efforts to preserve the old and the germe or nuclei around which is
built the new configuration.
Catastrophe Theory
Catastrophe theory is a means of showing how slight incremental
changes in one or more continuous variables (independent variables,
causal factors, control factors) produce sudden discontinuous jumps
or drops in one or more behavior (dependent) variables. A number of
different catastrophes have been identified. The "elementary" ca-
tastrophes are classified by the number of control and behavior fac-
tora each possesses. The fold (one control and ons behavior factor),
the cusp (two control factors and one behavior factor), and the but-
-15-
terfly (four control factors and one behavior factor) have been used
most, especially in the behavioral/social and biological sciences.
Fige. 1 and 2 further illustrate some principles of catastrophe
T ee
aw
prvdsheld 26
Se,
PeeH,,
he: 1
: ch
—airerector Bren
Control Factor
Fig |. Fold Catastrophe
Fig. 1 shows the simplest catastrophe, the fold. ‘The manifold
Jomp
wi fh
Aystevesis|
Behavior Factor
Fig 2. Cusp Catastrophe
represents equilibrium points of maxima or minima (the attractor)
separated by points of minima or maxima, respectively (the repellor
or inaccessible region). The fold (and also the cusp) can be shown
to evolve from the logistic curve. Positive feedback may contribute
to the growth shown on the lower limb and to the decline shown on
the upper limb, but regulatory negative feedback is involved insig-
nificantly if at all. When the magnitude of the behavior variable
reaches the singularity on either limb, behavior jumps or drops dis~
continuously to the other limb. Further, the eysten dynamics practice
of averaging (as well as the econometric practice of finding the line
of best fit by regression analysis) would be highly inappropriate in
such situations. The average or best-fit line could lie in the
repellor region, the least likely behavior.
The cusp catastrophe is a particularly useful construct be=
cause it includes a family of sub-constructs. If one observes one
of the latter, a clue is provided for search for the others. The
125
~16-
five sub-constructs are: (1) bimodality of behavior in part of the
range, with sudden jumps from one mode to the other; (2) hysteresis
or path non-reversibility in jumps between modes or sheets; (3) an
inaccessible or repellor region on the behavior axis, representing
least likely behavior; (4) divergent behavior on either side of the
fold on the behavior surface, so that a small perturbation in the
initial state of the system may produce a large difference in the
final state; ana'(5) the catastrophic jump iteelf,
The cusp catastrophe provides further insights as to how a
modified system dynamics might better mimic the real world. Sudden
fads, religious conversions, and political switches, social behaviors
quite important to policymaking, appear better depicted by the bi-
modality of behavior with sudden jumps from one mode to another than
by the continuous oscillations of system dynamica levels and rates.
' The last-minute switch from Carter to Reagan in the 1980 presidential
elections may have represented such a catastrophic jump. Hysteresis
can be viewed as a delay or inertia in the system, but hysteresis
does not appear to be isomorphic to any system dynamics delay because
the jump and fall pathways are qualitatively and quantitatively dif-
ferent. In some cusp models the two sheets of the behavior surface
represent polarizations or qualitatively different system atates, for
example, dove and hawk attitudes. In other models, the two sheets
represent different scales of the same variable, for example, num-
bers of spruce-budvorm larvae in Canadian forests. The last case
would apparently require inserting a delay within a level.
Some of the most powerful sub-constructs of the cusp catas-
trophe are the concepte of attractor, repellor, divergence, and
splitting. These concepts can aid the modeling of competition, con-
-17-
flict, polarization, and evolution of new forms, which are almost
inherent in most behavioral/social and ecological situations. For
example, fear and rage may be conflicting emotions influencing aggre-
sive behavior. Divergence and splitting may lead to new syatem forms
as in speciation.
In Zeeman's model of a stock market crash [13], the normal
factor is excess demand for stock (which could be a system dynamics
level variable), the behavior factor is the rate of change of the price
index (which could be a system dynamics rate variable), and the
splitting factor is the proportion of the market held by speculators
as opposed to long-term investors (which could be a level variable,
but wich would be difficult or impossible to enter causally into a
system dynamics model). Increase in the splitting factor causes
greater and greater divergence between the top (bull market) and bot~
tom (bear market) sheets; that is, the larger the eplitting factor,
the more severe the crash. The slow smooth recovery involves posi~
tive feedback loops in which the behavior factor affects the control
factors. Once again, it appears that catastrophes occur in the ab-
sence or exhaustion of negative-feedback regulation and control.
When equilibrium breaks down, catastrophes follow.
System dynamics and catastrophe theory have two important
properties in common: (1) determiniam and (2) equilibrium. These
features could start the melding of the two basic constructs.
Dissipative Structures
Dissipative structures are self-organizations arising
through the occurrence and amplification of certain fluctuations in
systems operating far from equilibrium. The basic dynamics are as
follows:
126
18+
Nonequilibrium Threshold Instability through structural
L fluctuation
‘Increased auaietioaed
Instability, triggered by nonequilibrium conditions, maintains
a continuous energy dissipation (measurable by entropy production in
physicochemical systems), which further increases the level of dis-
sipation, leading to further instabilities. Prigogine [10] calls
these processes evolutionary feedback. Nonequilibrium conditions lead
to exceeding a threshold, which now increases instability by means
of structural fluctuation, which in turn produces increased dissi~
pation. ‘The last, in a feedback loop, modifies the threshold,
leading to evolution through a succession of transitions.
Most of the theory of, and experimential substantiation for,
dissipative structures comes from physical chemistry. However,
Prigogine and his associates, incorporating research ideas from wor-
kers in several fields, have extended the theory to ecosystems and
societal systems. One example is the oscillation between “autocra—
tic" and "democratic" social structures among the Kachin tribes of
northern Burma. When the prestige of a new chief is larger than dis—
satisfaction with his ascension, the autocratic regime remaina sta~
ble. But if dissatisfaction is greater than prestige, entry into
the system of a few rebellious persons drives the system to a revolu-
tionary state as the number of rebels increases explosively.
Cybernetic theories like system dynamics and control theory
do very well in describing self-regulation and why ecosystems and
societies stay the same, but they do not describe vell how these
syetema change into new forms. Once sgain, in evolution equilibria
“19
are not maintained but are destroyed. A system, subjected to a field
of external and internal forces that collectively surpass some thresh~
old, ruptures and/or yields emergent new forms. As long as the boun-
gary of stability is not passed, the system will return to essential-
ly its equilibrium state if the external perturbations and internal
stresses and strains abate. System equilibrium, stability, instabil-
ity, collapse, and emergent organization thus are functions of the
tensity and persistence of the given field of forces. There are
many realworld examples of systems that do not return to equilibrium
following removal of a perturbation. Holling [9], for example, dis~
cusses the extinction of several species of commercial fish in all
five Great Lakes. Even when fishing pressure was removed, the fish
did not return. It is the belief of the present author that the same
dynamics apply to modern Western society, producing new structural
forms like the chronically unemployed, underemployed, and hopeless.
System dynamics modele do not exploit sufficiently the con-
cepts of domain and boundary of stability, that is, behaviors at the
extremes of oscilletions or fluctuations beyond which the system
cannot return to its original condition. The adaptive capabilities
of systems are sorely taxed under conditions near stability boun-
daries as established functions and policies fail to perform their
corrective actions. Although programs and policies, for example,
a training program, a birth control program, or a technological break-
through, may be introduced exogenously at later times in a system
dynamics simulation run, these policies operate through the pre-set
system structure and do not capture the breakdown of system struc-
ture and emergence of new structure under conditions far from equi-
librium. For exemple, the analysis of the Kondratieff cycle in the
127
20
National Model should focus not just on apparent causal factors such
as an excess of capital, on the quantitative features of oscillations,
and on presently apparent corrective policies, but also on the emer~
gence of new kinds of structure. A retrospective and retrodictive
qualitative analysis of new forms emergent during the late downswing-
early upswing phase of past Kondratieff cycles could contribute great~
ly both to theory-building and to policymaking. Evolution in general
proceeds in the direction of greater complexity, with each stage re-
presenting a successive reequilibration to changing forces. As em-
phasized throughout this paper, it ie the critical intervals between
stages, such as the present one, that should receive our greatest
attention.
Toward a Melding of System Dynamics and Complementary Perspectives
As a point of focus, consider the anthropological study of
the New Guinea Tsembaga tribespeople by Roy A. Rappaport. Rappa-
port emphasized the self-regulating function of ritual in thia so-
ciety. Other, evolution-oriented anthropologists, however, have cri-
ticized Rappaport's interpretations on the basis that they explain
social stasis but not social change. Systems ecologists have inter
preted the study in‘terms of a society's maintaining fluctuations
80 that the stability boundaries do not contract and thereby reduce
the capability to respond to unexpected perturbations and in tum
reduce survivability.
Shantzis and Behrens [12] designed a system dynamics simu-
lation model based mainly on Rappaport's Pigs for the Ancestors. The
model included critical levels or thresholds for pig-person compe~
tition (social temperature or conflict to other authors) and for the
amount of human labor required to tend a given number of pigs. These
2t-
critical levels, via a CLIP or FIFGE function, triggered festival
and warfare behavior in which many pigs and some people were killed,
thus restoring human population and pig population to equilibrium le-
vels consistent with land carrying-capacity.
This model therefore simulates some features of a cates~
trophic jump. Indeed, one could envision an alternative catastrophe
theory model with pig numbers and human numbers as control factors
and human aggression as a behavior factor.
IEHAVIORAL, AL/SOCLOTECHNICAL THEORY
ENRI!
Within the scope of present system dynamics theory, great
strides can be made toward greater fidelity through the better un~
derstanding and incorporation of theories and findings from psychol-
ogy, sociology, cultural anthropology, and the study of sociotechnical
systems. Table 1 gives some representative constructs.
Table 1. Representative Behavioral/Social/Sociotechnical Constructs
Percevtual Motivational, el
‘Avsention ‘Aspiration Competition
Eias Conflict Conflict
Gestalt Drive Contagion
Saturation Frustration Diffusion
Stimulus strength, duration Gap Movement
Hierarchy Role
Incentive Social comparison
Need Social temperature
Attribution tonal tara
B Emotion: Cul:
Ceeearison ‘Aggression Established practice
Creativity Alienation Bthnie difference
ision Helplessness lores
Dissonends Hostility Political philosophy
Expectancy Nonrationality Religious difference
Inagination Sex difference
Intelligence Activity
Judament “Achievement Sociotechnical
Acquaition ‘Autonomy
Consumption Innovation
Participation
Work system
Discovery, invention
Technological impact
128
“22
Most of the areas have been studied by disciplinary special-
ists and apply to several hierarchical levels of living systema. Un-
fortunately, the specialists have provided few if any grand theories
recently and few if any "off-the-shelf" models for ready incorporation
into system dynamics theory and models. This does not, however, jus-
tify the practice of some system dynamicists of aggregating and mask-
ing these factors in averages, delays, or purely economic variables.
Table 1 is meant to be both a source of new insights and constructs
and a source of caution that system dynamica theory and practice ap-
plied to socioeconomic systems will fail to represent the real world
in the absence of these factors. In many cases using these factors
is straightforward; for example, level of aspiration, level of expec-
tancy, and social comparison between these levels and actual achieve-
ment lend themselves directly to system dynamics modeling.
Because of the criticality of societal problems at the na-
tional and international levels, and because of the potential help-
fulness of the system dynamics National Model in the solution of these
problems, the remainder of this section will be devoted to work, em-
ployment, nonemployment, and productivity. Here motivational and
sociotechnical theory and findings oan most meaningfully be contri-
buted. In keeping with theories discussed earlier, it is collective
rather than individual behavior that most demands our attention.
A at the Nat el.
Consider first the basic structure of the labor and produc-
tion sectors of the National Model. Unless recent modifications have
been made of which the present author is unaware, the basic structure
of the labor sector is as follows [8], [11]. The sector poasesses a
level or pool of the generel nonemployed, a level or pool of the
-23-
nonemployed for each industrial-production sector, a level or pool
of the employed for each production sector, a level of wages, and five
rates, namely, departures from the sector and arrivals in the sector
connecting the first two levels, separation rate and hiring rate
connecting the second and third levels, and change in wages accumu-
lated as wages.
A number of ways in which the model might be improved by con-
sidering behavioral/social and sociotechnical factors can be summari-
zed as follows:
We
2.
Shere is no simple relationship between wages and work performance.
Assembly-line workers, among the most highly paid blue-collar
workers, are also among the most alienated. Work fulfills other
functions beyond earning money. Money has a symbolic meaning
traneeending purchasing pover and consumption. People make trade~
offe among the various positive and negative incentives ("valences")
of a job and anong work and non-work factors. Values and attitudes
toward work have changed greatly, partly as a function of the
time and environment characterizing a person's early life. What
were once considered privileges are now considered entitlements.
This attitude change extends from the shop floor to the executive
suite. There is increasing demand for a high quality of work-life,
which in turn influences work-system design.
Fajor attention should be paid to chronic unemployment and under
employment.
catastrophic flips to new levele of equilibrium.
These developments may represent nearly irreversible
Unemployment
and underemployment are not simply levels or pools into and from
which people flow mechanistically. Rather they represent the
results of evolutionary processes whereby the individuals consti~
tuting the level and the level itself have changed qualitatively.
129
Be
=24-
A paradox is created with regard to education and skill levels,
which are difficult to assess and involve counterveiling forces.
Qn the one hand, political and psychological pressures and grade
inflation, coupled with mass education, appear to have reduced
the quality of given level of education. On the other hand,
companies require higher educational and skill levels, aggravating
‘the tendency to chronic unemployment of those with modest or ob-
solescent education and skill levels. At the same time there is
a disparity between the educational and skill levels employers
demand for many jobs and the abilities and skills actually re-
This @ieparity exacerbates the tendency
Work is a primary psychosocial need,
quired for these jobs.
to chronic underemployment.
not just a means of earning money and enabling consumption. Work
provides meaning, dignity, a feeling of self-worth, and an op-
portunity for self-fulfillment. No work, poorly designed work,
and exploitive work produce disaffection and alienation, which
not only reduce the productive capacity of the sociotechnical vork
system but also spill over into the family, the community, and
leisure-time activities. Workers displaced for economic and
technological reasons are not necessarily easily transferred
among industrial sectors, thus exacerbating chronic conditions.
4nd chronic conditions breed hopelessness and decrease incentives.
The effects of automation and technical change may exceed a cri~
tical threshold. After assuming the role of a nonproblem fol~
lowing the 1966 release of the voluminous reports of the Nation-
al Commission on Technology, Automation, and Economic Progress,
automation and technological change may be poised to trigger a
further catastrophic reconfiguration in the already weakened
4
Se
=25=
enploynent/nonemployment system. Rapid advances in computer and
coamunications technologies, and in the perceptualmotor and sim-
ple decisional aspects of artificial intelligence and applications
to hierarchical industrial control and to industrial robotics,
seem likely to have imminent effects on the nature and availabil-
ity of work.
The effects of disaffection and alienation are already profound.
The decade-long productivity slump is at least partly due to these
factors. Alienated workers express their dissatisfaction on the
job by increases in job turnover, absenteeisu, foot-dragging,
pilferage, vandalism, sabotage, crime, strikes, sick leave, al~
cohol and drug abuse, and psychosomatic reactions. In society
at large, disaffection and alienation underlie the rapid increases _
in delinquency and crime, particularly random attacks and large~
scale, attention-getting crimes like mass murder, assassination,
arson, and skyjacking. The costs of delinquency and crime are
huge, not only economically but also in terms of fear, societal
breakdown, and the compensatory retaliatory measures taken.
roductivity are complex and 10% be ct ti
The limitations apply
Producti
simple or uniform production functions.
not only to simple functions like Cobb-Douglas (used in the Me-
sarovic-Pestel and Bariloche world models) but to more elaborate
Production functions ignore, distort, or mask
Perhaps
ones as well.
the effects of natural resources, technology, and labor.
their greatest limitation, however, is the assumption that if
only enough and the right mix of capital and labor were applied
to production, the results would be positive, increasing, and pre-
dictable. This assumption ignores the processes of saturation
130
-26-
and exhaustion.
productivity were usually one-time-only occurrences, for example,
Historically, large increases in production and
the mechanization of agriculture, the switch from agriculture to
manufacturing, the switch from industrial electromechanical con-
trol to electronic control, and the switch from physicians’ house
calls to physicians’ seeing many patients at a central location.
In many areas further improvements may be spurious or may displace
additional large numbers of workers. Further, the decline in rate
of growth of US productivity has often been attributed to declines
in funding for R&D. But this explanation overlooks the insidious
decline in creativity, rate and quality of discovery and innovation,
and psychosocial climate in aging, ingrown funding and research
organizations. In addition, there are many inter~induetry and
inter-organizational differences. ‘
model" has already influenced thinking in many companies so that
ideas like layoff rate are obsolete. Also, it is quite difficult
For example, the "Japanese
to predice labor requirements for emerging jobs in the future
world where the National Model will partially operate.
Bas: ici \d_ 1.960:
Yield dangerous outcomes.
The ‘Phillips eurve tradeoff between inflation and un-
n_ideas that arose. the
Much economic theory seems hardly
current.
employment provides a salient exemple. A given level or percentage
of unemployment might be accepted by the conetituents, given a
particular configuration of the enveloping field of forces, until
one small increment more results in surpassing a critical thresh-.
old of temporary stability and the system erupts in violence,
This is what is happening in English cities
Prime Minister Margaret Thatcher's
looting, and arson.
at the time of thie writing.
~27-
monetarist tradeoff between inflation and unemployment, apparent-
ly unsuccessful on other grounds as well, seems to have been
singularly poor policy.
In short: causal diagrams involving the factors just dis-
cussed could be quite different from those representing the Nation-
al Model.
CONCLUDING REMARKS
The present author considers system dynamics to be an inte-
gral part, perhaps the core part, of a unified dynamic theory of the
evolution of complex living systems. This paper has discussed theo-
ries and constructs that appear to extend or enrichen feedback-vased
system dynamics in building the larger theory. But how compatible
are the theories and constructs when one approaches the tasks of
actually constructing theories testable against the real world and
models useful in policymaking?
Three categories of immediate further research are indicated:
1. Improvements of present policy-significant system dynamics mo-
dels like the National Nodel to reflect better the behavioral/
social and sociotechnical forces, findings, and factors.
2. Construction of models to interact with syetem dynamics models.
System dynamics models could be variously entered; left, modi~
fied, and reentered. Probabilistic system dynamics was mentioned
as an example of this approach. It was also suggested thata °
cascade of models might be designed involving alternating system
dynamics and catastrophe theory or digsipative structure models.
Considerable thought would have to be directed toward the eitu-
ations that trigger exit from one model and entry into another.
It would be relatively straightforward to "fudge" such situations.
131
=28-
However, model design should be based on a better understanding
of the relevant realworld thresholds, fluctuations, discontinui-
ties, ete.
3. Construction of a “unified field-theoretic" model which endoge~
nizes all constructs. Thie task might yield major incompati~
bilities stemming from the basic mathematical model of system
dynamics and the DYNAMO compiler. For example, full mathemati-~
cal treatments of critical phenomena, catastrophe theory, and
dissipative structures variously involve sets of nonlinear equa-
tions, partial differential equations, higher-order ordinary
differential equations, and stochastic considerations. It might
turn out that a simulation language like GASP IV vould be more
appropriate for this formidable task.
Continuation in the observation, collection, and interpre-
tation of realworld phenomena and processes that can be used to test
and substantiate the basic theory must of course at least keep pace
with model-building and policymaking.
REFERENCES:
[1] De Greene, Kenyon B., 1973. Sociotechnical Systems: Factors
in Analysis, Design, and Management. Englewood Cliffs, New
Jersey: Prentice-Hall.
{2] De Greene, Kenyon B., 197, "Force Fields and Emergent Phenome-
na in Sociotechnical Macrosyatems: Theories and Models.”
Behavioral Science, 23 (1), 1-14.
[3] De Greene, Kenyon B., 1978b. "Field Theory as a Framework for
the Computer Simulation Modeling of Complex Societal Systems."
In Proceedings of the International Conference on Cybernetics
and Society (Tokyo-Kyoto, Japan, 3-7 November 1978), Vol. I,
[4]
{5}
{6}
{7]
{9}
(10)
1]
[12]
[13]
-29-
Pps 335-345. New York: Institute of Blectrical and Electronics
Engineers.
De Greene, Kenyon B., 19818, “Limite to Societal Systems Adap—
tability." Behavioral Science, 26 (2), 103-113.
De Greene, Kenyon, B., 1981b. The Adaptive Or, nt -
ation and Management of Crisis. New York: Wiley.
De Greene, Kenyon B., in press, "Anticipating Critical Recon~
figurations in World Societal Fields." IEEE Transactions on
Systems, Man, and Cybernetics.
Forrester, Jay W., 1961. Industrial Dynamics. Cambridge,
Massachusetts: M.I.T. Press.
Forrester, Jay W., Nathanial J. Nass, and Charles P. Ryan, 1976.
"The System Dynamics National Model: Understanding Socio-Econo-
mic Behavior and Policy Alternatives." Technological Forecast-
ing and Social Change, 9 (1 & 2), 51-68.
Holling, Crawford S. (Ed.), 1978. Adaptive Environmental Assess~
men: Management. New York: Wiley.
Nicolis, Gregoire and Ilya Prigogine, 1977. Self-Organization
Nonequilibrium Syster yn_Di. tive ture Or-
der Through Fluctuations. New York: Wiley.
Runge, Dale, 1976, "“Labor-Market Dynamics." In Jgrgen Randers
and Leif K. Ervik (Eds.), The System Dynamics Method, Proceed—
ings of the International Conference on System Dynamics, pp. 97-
126. Oslo: Resource Policy Group.
Shantzis, Steven B, and William W. Behrens, III, 1973. "Popu~
ation Control Mechaniems in a Primitive Agricultural Society.”
In Dennis L. Meadows and Donella H. Meadows (Eds.), Toward Glo-
bal Equilibrium: Collected Papers, pp. 257-288. Cambridge,
Massachusettes: Wright-Allen Presa.
Zeeman, E. Christopher, 1976, “Catastrophe Theory," Scientific
American, 234 (4), 65-83.
-132
