THE DYNAMICS OF SYSTEM DYNAMICS
Willard Fey
School of Industrial and Systems Engineering
Georgia Institute of Technology
Atlanta, GA 30332
Presented at
The IEEE 1981 International Conference
on Cybernetics and Society
Atlanta, Georgia
October 28, 1981
Copyricht © 1981 by
Institute of Electrical and Electronics Engineers, Inc.
345 East 47th Street, New York, New York 10017
All rights reserved. No part of this paper may be reproduced
in any form or by any means, electronic or mechanical, including
photocopying, recording, or by any information storage and
retrieval system, without permission in writing from IEEE.
i
i
‘
THE DYNAMICS OF SYSTEM DYNAMICS
Willard Fey
School of Industrial and Systems Engineering
Georgia Institute of Technology
Atlanta, Georgia
ABSTRACT
After twenty five years of development and
some notable achievements the field of System
Dynamics is not as large, well-known, respected
and influential as it should be based on the
breadth and power of its principles and the need
of industry and society for dynamic analysis of
this kind. It is suggested that System Dynamics’
methods be used to analyze the growth of the field
and improve its development. This paper initiates
the self-analysis by presenting a review of per-
formance and a preliminary model structure for the
field to encourage constructive criticism and to
facilitate understanding and cooperative revita-
lization. The model structure may be general
enough to apply to cther fields as well.
INTRODUCTION
In the fall of 1956 Jay Forrester founded the
Industrial Dynezics Group at the Sloan School,
H.I.T. In the ensueing twenty five years many
people have leazned Systen Dynamics methods and
used them to study a wide variety of systems. On
the occasion of s silver anniversary it is
appropriate to acknowledge our many accomplish-
ments as well as to reflect on our present condi-
tion and future aspirations as a professional
field of knowledge and practice. Perhaps, as a
result of that reflection some of our weaknesses
may be recognized and our research, teaching and
professional practice may be revitalized, coordi-
nated and redirected in ways that will produce an
even better future.
System Dynamics is the only science-based
methodology that is sufficiently logical (causal
based), comprehensive, flexible and quantitative
that it can serve as the basis for realistic anal-
ses and substantial improvements of complex, dy-
namic, nonlinear, nonstationary, noisy human sys-
tems at the managerial levels of aggregation where
the major long run behavior patterns are controlled.
These are the systems (the world, countries, social
agencies, industries, companies, cities, .-.) and
problems (inflation, escallating antagonisms and
debt, oscillating profits and exchange rates,
increasing hunger and crime, ...) wpon’which rest
the survival of our civilization. Considering
the vast number of critical dynamic problems now
facing these systems and considering the enormous
potential contribution SD could make to the solu-
tion of these problems, one would expect that after
25 years of development SD would be well-known,
widely used and extensively taught. Most of the
SD practitioners with whom I have talked are dis-
appointed with our progress in these areas.
The purpose of this paper is to review our
condition and to suggest procedures for identifying
strategies that might improve the field and our
contribution to the world community and its organ-
izations at all levels. In order to improve the
dynamic performance of the field, it is suggested
that SD.methods be used to perform an analysis of
the growth of the field. This requires identifi-
cation of performance goals, time patterns of
important variables and dominant feedback loops;
model building; analysis of model: validity and
operation; evaluation of the capabilities and
attitudes of SD practitioners, clients, students
and potential system participants; creative syn-
thesis of a more effective system structure; and
continuing collaborative implementation and modi-
fication of the recommendations by the field's
practitioners.
This paper initiates the self-analysis pro-
cess by presenting a preliminary review of goals,
past performance, system structure, several pos—
sible improvement strategies to be tested, and a
suggested procedure to carry out the self-study.
This review will be presented at two conferences
in October, 1981 to provide the greatest possible
involvement of SD practitioners in the process.
While the best improvement strategies are not yet
known, it is certain that greater levels of com—
mitment, direction, understanding, communication,
collaboration and cooperation will be required of
SD practitioners to achiéve faster growth and high
quality practice.
PERFORMANCE OF THE FIELD g
The SD field is defined to include all world-
wide accumulated SD knowledge, wisdom and infor-
mation both written and mentally stored; the
people who have been trained in SD methods whether
or not they are now practicing SD and the SD work
they have done or are doing; the client individuals
and organizations who have used or are using the
methods and/or have supported or are supporting
their development, use or teaching; SD educational
programs and students; and the combined images of
the field in the mind's of potential clients, the
academic community, potential students, publishing/
media, and the general public. While some infor-
stint
mation about the time histories of these variables
does exist, particularly at M.I.T., it is not ex-
tensive. Therefore, no definitive statement can be
made about the past history or present condition.
However, my perception of the time histories and
present state of SD variables roughly matches the
perceptions of several colleagues with whom I have
talked.
This perception suggests that SD has not de-
veloped as clear and broad a base of theoretical
knowledge as 25 years of work should produce or is
needed to create a clear image of professional com
petence and to support a teaching effort which must
transform normal college graduate students into
capable practitioners of a difficult science-aided
art. Reports of the theoretical research that has
been done are widely scattered and not readily
available. Careful, objective assessments of po-
tential contributions from existing statistical and
mathematical methods exist, but are not sufficient.
The applications literature is considerably
broader than the theoretical; but it is inadequate
in convincing, understandable, practical successes.
Most applications epply either to hypothetical
situations or to unimplemented studies of real
systems. The nature of the field makes clear, prac-
tical successes ¢izficult to achieve and document,
but that is an obstacle we must overcone.
The number cf trained practitioners seems to
be smaller than it "should" be after 25 years of
teaching and much smaller than it must be to do the
theoretical research to develop the field's know-
ledge base, the teaching to increase the quantity
and quality of practitioners and the analyses to
study the many systems that could greatly benefit
from exposure to the method. Furthermore, the
quality of the work is not uniformly high. This
may be due to low capability of a few analysts or
special circumstances of some studies. Ia any case
practitioner quantity and average quality both
appear to need i=provement. .
Past and present clients and supporting organ-
zations such as universities at which SD programs
are taught, research funding organizations, indus-
try and government do not appear to, be exceptionally
enthusiastic about the field. Certainly, a few are
enthusiastic, but there are only a few universities
worldwide that I know about that offer SD as a
field of specialization at the Ph.D. level in man-
agement, engineering, the physical or social sci-~
ences. The hundreds of universities that do not
have such programs do not seem to be actively seck~
ing skilled SD people to start advanced SD programs
in their schools. The large government funding
agencies in the United States such as NSF and DOD
do not seem to be actively soliciting SD studies.
There are some, of course, but not a number even
remotely commensurate with the age of the field
and the potential benefits from the use of the
method. There are some SD staff groups in industry
and some industrial consulting, but it is not ex-
tensive. There is some academic, industrial and
government support for SD programs and projects
abroad, but it also seems not to be commensurate
with the promise.
Finally, the perception suggests that the SD
image is not clear, widespread or very positive for
people outside the field. Antagonism and/or lack
of respect for the field is fairly extensive and
well-known in the academic community. Since SD is
not widely taught, many academicians know little
about its principles. A cursory observation does
not reveal the depth of knowledge and judgment re-
quired to do a proper SD study. A way must be found
to communicate the true nature of the field. SD
does not seem to be well-known in industry either,
particularly in medium and small business, or in
government, particularly at the operating levels.
SD is not well-known in publishing and the news
media and the general public knows virtually noth-
ing about SD. Those who have heard of SD usually
have not heard glowing reports of great successes.
Our reviews typically are mixed at best. °
In summary the apparent condition of SD as of
mid 1981 is that it 4s not growing very rapidly, is
not very large for its age, does not have extensive
professional acceptance, has not produced many clear
successes, and is not widely known or highly re-
garded. But it has the theoretical potential to
transform the prospects for mankind. While the
details of this perception vary somewhat from per-
son to person, I have never heard anyone suggest
that SD is a large, healthy, rapidly growing, well-
known, widely respected field that is producing
substantial numbers of high quality practitioners,
important theories and successful practical results
(my goals for SD). The above perceptions involve
both an awareness of actual conditions and judg-
ments about goals (what is desired). I hope that
I am either misinformed about conditions or overly
demanding in setting goals. But if I am correct,
a@ great deal of study and effort is needed to im-
prove the field's performance.
A SUGGESTED SELF-ANALYSIS
The SD methodology teaches that human systems
are complex combinations of coupled feedback loops
which function through time to create the funda-
mental patterns of variation (trends and cycles).
observed in the important variables. Improved
patterns are achieved by altering feedback struc~
tures in appropriate (effective and possible) ways.
In any particular system the identification of
effective and possible changes requires a clear
understanding of the existing feedback loop struc-
ture, a thorough understanding of the human char-
acteristics of the system's participants who will
4nfluence the changes, and a creative synthesis
that discovers effective modifications within the
realities of structure and human capabilities and
attitudes.
I suggest that the field of System Dynamics is
a human feedback system as described above which
exhibits unacceptably low growth rates for its
important variables. It would seem that if we are
to increase the growth rates substantially, the
feedback structure should be redesigned. In order
to redesign the structure we must understand the
shown in Figures 2-4. In a few cases variables not
shown in Figure 1 are added to complete some loops.
The following examples are all partial representa-
tions, so most of the variables also are influenced
by variables not shown in Figures 2-4.
The knowledge base is increased by research
effort (Figure 2). Research effort is influenced
by practitioners' desire to do research and the
availability of support for it. The existing know-
ledge base supports the scientific credibility of
the field, influences applications quality, and helps
to define the areas of insufficient knowledge. Cre-
dibility positively enhances the ability to obtain
support; awareness of insufficient knowledge sti-
mulates the desire to do research. After a percep-
tion delay applications’ quality positively affects
credibility. The error between desired applications
success and applications quality stimulates desire
to do research. The latter negative loop will serve
to limit growth only if the success goal is too low
or applications are highly successful.
A second area involves teaching and analysis
by practitioners (Figure 3). After a time delay
extra teaching effort (more SD programs) results in
an increase in practitioners, some of whom teach.
Total teaching effort depends on the number of prac-
titioners and the fraction of time spend teaching.
As teaching effort increases, analysis effort by
the practitioners must decrease and vice versa.
Increased analysis effort (given reasonable quality)
leads to greater client awareness of the field and
its benefits. This in turn generates greater anal-
ysis demand. Whether increased analysis demand
leads to a larger or smaller analysis fraction de-
pends on whether short or long run priorities dom
inate the decision. A+ sign represents the choice.
There are two positive loops and one negative in
this sector regardless of the sign at analysis frac~
tion,
A last exexple involves education quality after
a perception delay influencing students and their
quality. Students influence the student/teaching
effort ratio that is one input to education quality.
Education quality also influences practitioner skill
which contributes to applications’ quality. Appli-
cations quality is one foundation for the field's
total demand which partly motivates support for
teaching. Student quality influences education
quality and practitioners’ skill which in turn posi-
tively affects education quality. The + sign at
education quality reflects the shape of the inset
curve and ensures four positive and one negative
loop.
PROCEDURE FOR SELF-ANALYSIS
Many positive loops exist in this complex feed-
back system that {4s worthy of our best analytic
skills. The analysis should indicate why the loops
are so weak. The synthesis should point to new
strategies to improve quality and growth. Modeling,
analysis and synthesis for a complex system are time
consuming. The fall meetings should serve to iden-
tify goals, problems, some loop structure, new stra-
tegies to test, and those who want to participate
in the study. Equation writing, model validation
and analysis, and synthesis are tasks a group of
individuals could share and communicate to the oth-
ers. If enough interest exists, several groups
could do independent studies to see if common con-
clusions emerge.
A general conference in the fall of 1982 could
include a review of the analysis and discussions of
new strategies. Consensus will be difficult to
achieve, but if we can agree on new strategies, im
plementation can begin then. (1)
REFERENCES
(1) Fey, Willard, 1980, "The Philosophy of the
System Dynamics Method,” In 1980 Proceedings
of the International Conference on Cybernetics
and Society (New York, Institute of Electrical
and Electronics Engineers, 1980).
4 Error ~+— Applications —t accumulation
Desired # Quality and Delay
Applications Fan + |
Success {FG g
Needed | Knowledge Scientific
Research Base * Credibility
+
* a » t ») te
Desire to? Research Research
do Research Effort & Support
Figure 2. Some research effort loops.
Teaching + Teaching
Effort Fraction
+ =
\s t
Delay (— Analysis = Analysis
G + Fraction Demand
+
/. te +
Working Analysis Client
Practitioners + Effort + Awareness
Figure 3. Some teaching effort loops.
Student Long
Quality Teaching 5 Delay
+ Quality +} |
t CG + { ig Gr a
Perceived Education Practitioners
Education + Quality % = «Skill
Quality :
| * yA
+ Students Applications
Students —»— Teaching |“T\ — Quality
+ Effort |+tt-
=} ; |
+
Total
Teaching
Demand
Demand +
Figure 4. Some education quality loops.
