System Dynamics and Systems Thinking:
It Takes All Kinds
Alan K. Graham, Ph. D. and Sharon A. Els
Pugh-Roberts Associates
41 Wm. Linskey Way
Cambridge, Mass. 02142
617-864-8880
Alan.K.Graham @ PA-Consulting.com
Sharon.Els@PA-Consulting.com
1. Abstract
Traditional System Dynamics modelers (“Complex Modelers”) view with distrust the
often simple but participatively-developed results of “Insight Modelers”. Both types
of quantitative modelers often regard “Systems Thinking” as dangerously incomplete,
and irresponsibly advocated. Yet nearly all modelers will sometimes be dissatisfied
with the substantial experience and resource commitments required to create an
adequate quantitative model.
This paper suggests that Systems Thinking, Insight Modeling and Complex Modeling
(and consultants and academics) each have their place within a larger system. This
view arises from examination of the sales process, the technology diffusion process,
and finally from analogy to the medical services delivery system.
A number of practices likely are hindering growth and quality of the field, broadly: 1)
Ignorance of the broad range of practices and skills now a part of modern System
Dynamics practice, 2) Blanket disparagement of other approaches and methods
within System Dynamics, and 3) Hype—advocacy of one’s own practice, without
demonstration of effectiveness. Finally, the paper suggests that, following its own
suggestions, quantitative modeling would be helpful in furthering the discussion on
the growth of the field and balancing the various roles and practices within it.
2. Introduction
Many System Dynamics practitioners feel that their discipline should become a
standard part of 21st century curricula and business life. Yet most practitioners are
aware of the dynamics of fads. Many will remember the rise and decline of popularity
of World Dynamics. Many will be fearing that Systems Thinking will follow the
same trajectory, at some point falling quickly to oblivion, and in the worst case, taking
much of System Dynamics with it. By contrast, we should discuss what it takes to
cause rapid, uninterrupted growth in teaching, skill, and use of System Dynamics as
well as Systems Thinking.
The related debate within our field tends to be simplistic: "approach A is correct;
approaches B and C are flawed." Traditional system dynamicists, who build what we
can here call “Complex modelers” often regard both Systems Thinking and “Insight
Modeling" (which uses modestly-sized models and relies heavily on group learning in
workshops) as potentially misleading. Systems thinking practitioners can fault
quantitative modeling for its inaccessibility by normal managers. Insight modelers
question the complexity of traditional System Dynamics models, again because of
System Dynamics and Systems Thinking: It Takes All Kinds
accessibility. Consultants can point to academic's weakness in real-world modeling
and implementation; academics can point to consultant's relative weaknesses in
analytical skills and breadth of knowledge of cases and applications.
Ironically, such "debates" are usually not systematic, and seldom consider the
dynamics of the field as a whole. Models of product or technology diffusion, such as
(Bass 1969), (Homer 1987), and (Maier 1998) generally show that the success of
products sold through their reputation requires both extensive word of mouth
exposure (for which Systems Thinking is well-adapted), and continued tangible
demonstration of product effectiveness (for which qualitative modeling is necessary).
There is synergy here. In short, for maximum success in System Dynamics as in most
ecosystems, it may be that "it takes all kinds."
The expression “it takes all kinds” is usually muttered upon seeing behavior or values
quite different from one’s own. Indeed, the words are usually delivered whilst
shaking the head with disbelief. But look at the words closely: they remind the
speaker to remember that in order to make almost any organization work, be it an
army, a church, a company, or a country, it takes all kinds of people.
In the realm of government, democracies value discussion and disagreement. But
peaceful resolution should be valued perhaps even higher; Western democracies have
the concept of the “loyal opposition.” It means that newly-elevated rulers, be they
kings or prime ministers or presidents, no longer need to exile or liquidate their
predecessor to assure prosperity and stability in the realm. It means that political
factions no longer feel compelled by fear to wage war (literally) on one another.
Perhaps it’s time to recognize the importance of differences within our own ranks as
well, and the importance of mutually respectful, collegial discussion, which enhances
the good of the realm.
3. Three Kinds, Two Differences
Let us first examine two of the divisions within our field. (There are doubtless more,
not perhaps as visible; the continuing friction between academics and consultants
comes to mind.)
Insight Models v Complex Models
A style of model and modeling which has gained popularity over the past 15 years is
one we can call “Insight Modeling” even though it could equally be called “group
modeling,” described in, e.g. (Vennix 1996) and (Vennix, et al. 1997). Most
characteristically of the process, much of the model conceptualization and structuring
is done in groups, with people from a client organization being fully active
participants. The mutual learning that comes out of such efforts is admirable and
desirable. As we know, even modestly-complex models can still yield interesting and
counterintuitive insights, insights probably not attainable without System Dynamics
modeling.
Insight Modeling, however, is limited to models of modest complexity. (Graham and
Walker 1998) explore some of the limits of this approach: constraints on speed of
modeling, number of people who can be involved, scope of issues addressed, and
depth of validation. At a broader level, there is the more fundamental limitation of
appropriateness for very important questions: Should the analytical basis for hundred
System Dynamics and Systems Thinking: It Takes All Kinds
million-dollar decisions or billion-dollar decisions be limited to the number of
feedback loops that can be conceived and understood by a small group of upper-
middle-level executives? If your answer is “no,” we could characterize your position
as belonging to the “Complex Model” school.
Contrary to common perception, what makes most Complex Models complex is not
disaggregated detail, but the breadth of scope, richness, and robustness of feedback
interactions. These require a prior understanding of the feedback dynamics to be
modeled adequately—it’s not possible to do a good job modeling going directly from
nothing to a final elaborate model. The path to a realistically complex model analysis
often in fact starts with 1) a Systems Thinking-type diagramming exercise, then
progresses to 2) a modestly-sized Insight-type model. Only then are the issues and the
complexities clear enough for 3) construction of a reliable and realistically complex
model. (Lyneis 1998b) describes how this sequence works in practice.
So Complex Modelers by and large are familiar techniques of Insight Modelers and
Systems Thinkers, but the reverse is unfortunately rarely true.
Qualitative vs Quantitative models
An even more fundamental debate takes place over Systems Thinking approaches to
problem-solving, which don’t even quantify relationships before drawing the
conclusions, which are based on relatively imprecise diagramming. There is now an
established body of literature and indeed, a conference circuit for Systems Thinking.
It is extremely popular and probably better-recognized by the general public than
System Dynamics. And there is remarkably little cognizance by Systems Thinking
aficionados that related quantitative modeling is even possible, let alone routinely
practiced. (This is in distinct contrast to many leaders in the field, who were trained in
Complex Modeling or Insight Modeling at MIT and elsewhere.)
The thought of drawing actionable conclusions from diagrams alone is profoundly
disturbing to many System Dynamics modelers of both the Insight Modeling and
Complex Modeling persuasions. Time and again, System Dynamics modelers see
their understanding of policy implications shift and evolve in parallel to the modeling
work. So the pitfalls of never checking one’s thinking through quantitative modeling
are palpable and compelling to those with experience in building realistic quantitative
models. Discussions of qualitative vs quantitative feedback modeling in the System
Dynamics field predate The Fifth Discipline, e.g., from (Wolstenholme and Coyle
1983) to (Wolstenholme 1999), but these have not engendered the attention and
discussion the issue deserves.
Finally, System Dynamicists and Systems Thinkers alike are aware of the dynamics of
fads. An appealing idea can spread very quickly, especially within the business
community. Inappropriate and unskilled applications of the idea can follow nearly as
quickly. As news of failures in practice spreads, the early popularity turns into
common knowledge that the ideas don’t work—regardless of their intrinsic merit.
This has been the fate, for example, of TQM, despite clear evidence that it does work
well when used well: “Everybody knows” that it doesn’t work. So there is the
apparent danger of Systems Thinking doing major damage to the reputation of System
Dynamics.
System Dynamics and Systems Thinking: It Takes All Kinds
That being said, most problems are not billion dollar problems, and cannot come close
to justifying a full-scale Complex Modeling effort, or even Insight Modeling. Is it not
equally inappropriate, a Systems Thinking advocate might well ask, to simply walk
away from problems that won’t (for any of several reasons) ever get decent
quantitative modeling? Indeed, the Systems Thinking work at MIT was started as a
continuing experiment in ways and means of bringing some of the fruits of System
Dynamics (and other learning disciplines) to be accessible by individuals, without
extensive training and degrees. How can the benefits of System Dynamics be
thoroughly spread into society if a Ph.D. and years of experience are required?
4. Differences Have Serious Implications
Are there documented cases of Systems Thinking clearly misleading the people using
it? The good news and the bad news is that applications of System Dynamics and
Systems Thinking are diverse enough that little overlap has occurred, at least to the
authors’ knowledge. The only complete case we know of where a Systems Thinking
exercise has been followed by intense quantitative modeling was Callaghan and Park
(1998). In that case, the Systems Thinkers were extremely experienced System
Dynamics modelers, and the conclusions were drawn both systematically and
conservatively. The qualitative conclusions weren’t contradicted by the subsequent
quantitative modeling, although it yielded considerably broader and more precise
conclusions.
But the potential for badly erroneous conclusions exists. Consider two case studies,
the first a Systems Thinking approach to implementing change in the US Navy
(Systems Thinker 1998):
Recognizing that success begets success, the...reform team sought to create-
and then promote—early successes. The “Communications Success Loop”
shows the importance of carefully selecting initiatives that have a high
probability of success; an initial success can trigger a “virtuous” reinforcing
loop, while a failure can cause the loop to run in a “vicious” direction.
Now consider a second study of implementing change in a semiconductor company,
this time a relatively well-validated quantitative model (Sterman et al. 1997):
Early results are widely advocated to demonstrate the validity of a program,
kick-start diffusion and boost the virtuous cycle of commitment and effort
(Shiba et al. 1993). On the other hand [the model analysis shows that] a focus
on quick results biases decisions against innovations with long time delays and
leads to myopic resource allocation. Focusing on early results may lead to
excess capacity, financial stress, downsizing and the collapse of commitment
to the program. Improvement programs can fail not in spite, but precisely
because of their early success.
Now it is clear that there are important differences in the contexts for the two
studies—the US Navy will not be surrounded by the same feedback loops as a
semiconductor company. But it is equally clear that if the qualitative study had been
performed on the semiconductor company, it almost certainly would have missed the
structural elements and feedback loops that turned out to be critical to the behavior
and policy implications.
System Dynamics and Systems Thinking: It Takes All Kinds
Indeed, it is relatively common to discover key structure and feedback loops even late
in the process of model calibration. For example, (Lyneis 1998a) describes modeling
to forecast cycles in aircraft orders (to the extent theoretically possible). In that effort,
they discovered through calibration that a previously-neglected factor—the rise of
aircraft leasing—was a major factor in creating an unprecedented (and difficult to
forecast) surge in aircraft orders.
So at least within the Complex Modeling community, it is commonly held that it is
extremely unwise and potentially damaging to embark on a Systems Thinking exercise
for an important problems without the intention of immediately embarking on
quantitative modeling to gain a certainty commensurate with the stakes involved.
5. It Takes All Kinds (To Get To Market)
For an individual practitioner, we can perhaps conclude that the only true
methodological flaw is failure to ensure that appropriate techniques are brought to
bear on a problem. But there is a second level of conclusion as well, in the arena of
System Dynamics and Systems Thinking as part of a larger system—the production
and consumption of systems-based ideas, insights, and results. We need to return to
the theme of “it takes all kinds.”
Consider the simple microcosm of someone within an organization deciding to use
System Dynamics or Systems Thinking. What in the environment will help that
process along or retard it? Salespeople sometimes use an acronym, AIDA, to describe
the sales process. The acronym stands for Awareness, Interest, Decision, and Action.
The relevant point for the present discussion is that different kinds of information are
needed at different points in this process, as Figure 1 below illustrates.
In the Awareness phase, publicity counts for a great deal. We shouldn’t need to
remind anyone of the tremendous surge in publicity created by Peter Senge’s The Fifth
Discipline for Systems Thinking and indeed System Dynamics. This kind of publicity
is very difficult for, e.g. academics or complex modelers to generate, due to a very
different orientation, skill set, and body of experience. It takes all kinds.
In the Interest phase, we’ve found it helpful time and again that someone inside an
organization says “yes, we studied that in school; it’s real.” They have personal
knowledge, however sketchy of System Dynamics or Systems Thinking, and that
internal legitimization moves the whole process of convincing stakeholders forward.
Who creates that personal knowledge? It’s not the publicity efforts—their contact
isn’t at all personal, or is it distinguishable from a fad. It’s academics and workshop
leaders. They end up teaching and giving confidence to vastly more people per annum
than a complex modeler. It takes all kinds.
System Dynamics and Systems Thinking: It Takes All Kinds
Awareness __ Interest Decision Action
Figure 1. “Ingredients” required at different phases of the sales cycle.
The existence of precedents is also extremely helpful---the demonstration that this
stuff not only works in the abstract, but works in your industry, on problems like
yours. The modeling schools that create the largest volume of success stories are
going to be academics and others doing Insight Modeling, simply by virtue of sheer
volume. Someone who does an Insight Modeling workshop every two weeks can
create twenty five times as many cases as a complex modeler doing year-long projects.
But at the same time, at least some of the success stories need to be clear-cut and
unequivocal demonstrations that not only did someone build a model and succeed,
someone had to build a model to achieve the kind of success they did, and that takes
the kind of in-depth analysis that only comes from realistic Complex Models. It takes
all kinds.
Finally, when it comes to deciding and acting, it’s important to have people available
that can do the Systems Thinking or the System Dynamics in practice. Academic
training isn’t enough, no matter how thorough-going. To do an engagement reliably
takes several years of experience and exposure. Such experience is usually found only
in consulting companies. Moreover, as modelers who have experienced both Insight
Modeling and Complex Modeling, we suggest that becoming truly skilled at Insight
Modeling or Systems Thinking requires experience at Complex Modeling as well.
Note that many of the most illustrious practitioners of Insight Modeling and Systems
Thinking in fact trained and practiced in Complex Modeling at places like MIT and
Pugh-Roberts. It takes all kinds.
6. It Takes All Kinds (To Build an Infrastructure)
To move away from the anecdotal and more toward the analytical, consider the causal
diagram below in Figure 2 that relates the different resources and activities within the
“System Dynamics system.” If the AIDA model in Figure 1 shows how the state of
the field influences the sale of new modeling activities, the causal diagram in Figure 2
“completes the loops” by showing the variety of ways that modeling activities build
the infrastructure of the field.
System Dynamics and Systems Thinking: It Takes All Kinds
Figure 2. Causal diagram of resources influencing society’s use of System
Dynamics and Systems Thinking
Two observations: The “current modeling activities” has many inputs, and the nature
of those inputs is probably nearly multiplicative. For example, with zero public
acceptance and zero personal knowledge about System Dynamics or Systems
Thinking present, the current modeling activities will be ceteris paribus zero or not far
from it. This contrasts with the case where there are a balanced set of inputs we talked
about in the AIDA selling model. So the flow diagram again formalizes the
proposition that “it takes all kinds.”
The second observation: Most of the loops in Figure 2 are positive, or self-
reinforcing. They are all capable of working in either direction. There is only one
“opposite” effect on the diagram, where more modeling activities, with the same level
of experienced practitioners, reduces the quality of work. The diagram suggests that
poor quality work can turn the positive loops into vicious circles, creating the sort of
fad dynamics that have nearly done in TQM, for example.
7. It Takes All Kinds (To Respond to Many Needs)
Consider the collection of professions shown in Figure 3 below.
First, we observe in Figure 3 an orderly progression of patient care, indicated by thick
arrows around the outside. The simplest injuries are dealt with by First Aid, which
millions of people are trained (or have learned) to do. Quickly-developing, life-
threatening situations are dealt with first by Emergency Medical Technicians in
ambulances. Stabilized but challenging conditions are dealt with by physicians,
specifically internists or general practitioners. The most challenges cases then go to
surgeons and other specialists.
System Dynamics and Systems Thinking: It Takes All Kinds
Figure 3 also shows systematic interaction with universities, where medical activities
of all kinds are studied, and best practices and new findings are continually fed back
to the various professions.
Figure 3. “It takes all kinds” to run a health care system.
We can suggest approximate analogies to the various “systems” disciplines.
e The high-stakes, high skill, resource-intensive world of the surgeon corresponds to
Complex Modeling.
e The more frequent activity is for lower stakes and less resource-intensive care,
which corresponds to Insight Modeling.
e For very quick action, there are the focused skills of emergency medical
technicians—the ambulance crew, just as there are skilled facilitators who use
Systems Thinking to quickly defuse dysfunctional executive interactions.
e Finally, there is self-help or treatment by non-professionals, which was and
perhaps still is the goal of Systems Thinking as used by executives and other
corporate employees.
8. What All Kinds Don’t Need
Rather than pontificate specific requests of specific groups, let us, hopefully with a
modicum of humor and charity, suggest a top-ten list of things the field would be
better off without:
1. Systems Thinkers ignorant of the existence of System Dynamics. They exist, and
in large numbers, blissfully plying their trade ignorant of both the pitfalls of
exclusively qualitative modeling, and a cure for them. (As noted above, the situation
is quite different for many of the leaders of the Systems Thinking movement, who are
“classically trained” in qualitative modeling.)
System Dynamics and Systems Thinking: It Takes All Kinds
2. Systems Thinkers failing to point out System Dynamics as an available follow-on.
This is roughly equivalent to the ambulance driver saying “what hospital? We’ ve got
everything you need right here.”
3. Hype--Failure to distinguish between experimental methods and proven
practices. Consultants of all persuasions will continue to be in bad odor with
academics until they are clear about the credentials of what they’re practicing,
specifically whether there are clearly documented successes. The business community
tends to accept new ideas even without proof at first, but then rejects them if they fail
to live up to the hype. At some point, advocacy without testing the hypotheses (e.g.
“the Systems Thinking approach reliably creates tangible benefits”) moves from “new
experimental thinking” to hype.
4. Professional Systems Thinkers without training in System Dynamics. Facilitative
skills, causal diagramming, and archetypes can move a group of decision-makers to
the point of agreeing that a set of issues has been captured. No System Dynamics
background is needed thus far. But taking action still requires drawing conclusions
from a complex set of interactions. Conclusions are far more likely to be correct if the
facilitator has been exposed to similar systems for which conclusions have been
rigorously tested. That is, the facilitator is much more likely to elicit correct
conclusions with training in qualitative System Dynamics modeling.
5. System Dynamics curricula without explicit training in facilitated model
conceptualization and construction. Facilitation expands the simplest modeling
skills into two important areas: general consultative and leadership skills, and
conscious modeling process. The alternative to training in facilitated modeling is
many years of trial and error to acquire a primarily unconscious modeling skill. This
is not only inefficient for individuals, but creates a permanent quality problem for the
field, as younger practitioners will not have fully developed the needed modeling and
leadership skills.
6. Ph.D.s in System Dynamics with no experience in Complex Modeling.
Specifically, Ph.D.s should have an experience of constructing and thoroughly
validating a realistically complex model, operating in a team, and packaging results
for non-dynamicists. How can one master a subject experiencing no more than the
students?
7. Insight modelers denigrating Complex Modeling in general. At least until
someone shows us a five-level model on whose policy predictions they’re willing to
bet several hundred million dollars on.
8. Quantitative modelers denigrating Systems Thinking in general. Disparaging
something, even when using it as part of one’s toolkit seems inconsistent, bad
manners, and bad Public Relations for the field.
9. Regarding delivery of model analysis results as the final stage in helping an
organization. Modelers who only “throw results over the wall” often fail to actually
produce improvements, and fail to maintain a reputation for practicality and value. A
higher ideal of modeling practice is direct and seamless follow-through to
implementation, preferably supplying the additional expertise required in this separate
realm. This means that modelers, be they consultants or academics, need to engage
some organizational change and consulting skills as part and parcel of the modeling
System Dynamics and Systems Thinking: It Takes All Kinds
process. Only thus can the modeling be counted on to do some good, and only then
can the field build a reputation based on tangible accomplishments.
10. Erecting barriers between consulting and teaching. Academics are often blind
to the management and consulting skills that surround a consultant’s modeling skills.
Consultants often fail to appreciate both the grasp of the SD literature and other fields
of knowledge, and the critical thinking skills required of successful academics. If
these blind spots are allowed to cut off continuing contact and learning, there are great
difficulties incorporating real-world priorities and experience into curricula. There are
also great difficulties introducing truly rigorous thinking and institutionalized learning
into consultative activities. The object of academic activities shouldn’t be solely to
create more academics. There needs to be some contact with actually being able to
make a difference in “the real world.” The object of consulting shouldn’t be just to
make money; there’s a learning infrastructure that needs continual maintenance. In
the long run, “it takes all kinds.”
9. Is This Systems Thinking Exercise Satisfactory?
One final note. Up to this point, we’ve discussed some issues, drawn three diagrams,
and drawn ten conclusions. Among those conclusions is the proposition that, for
sufficiently important problems, simply drawing some diagrams and drawing
conclusions isn’t an appropriate point at which to stop the analysis.
Complex modelers may have noted already that the causal diagram above has a
multitude of positive loops and only one negative loop. Often, that’s a sign that the
modeler hasn’t yet recognized many of the constraints and negative feedbacks that
determine strategic success and failure. The conclusions from the diagram are almost
certainly incomplete, even if the limited conclusions are true.
It therefor would be particularly appropriate to go further, to model the development
of our field quantitatively. In April, Jack Pugh, acting as President of the System
Dynamics Society, sent out a general announcement that began:
The System Dynamics Society has been in existence 13 years. We have
achieved most of our initial priorities. Now it is time for the society to review
the purposes for which it was organized and perhaps to set new priorities.
Participation by the whole society necessary if we are to generate as many new
ideas as possible and for the membership to "own" the new goals that come
from this process.
Developing a quantitative model would serve two purposes:
e Create an explicit dialog among the multiple stakeholders in the field, many of
whom are not well represented in the System Dynamics Society and _ its
conferences
e Create an explicit process for the Society to consider strategies and tactics going
forward
System Dynamics and Systems Thinking: It Takes All Kinds
10. References
Bass, F. M. (1969). A new product growth model for consumer durables.
Management Science 15, 215-227.
Callaghan, Martin and Rick Park (1998). Gaining strategic leverage at London
Underground. Proceedings of the 1998 Systems Thinking in Action Conference, San
Francisco, CA. Pegasus Publications, Waltham, Mass.
Graham, Alan K. and Robert J. Walker (1998). Strategy modeling for top
management: Going beyond modeling orthodoxy at Bell Canada. Proceedings of the
16" International Conference of the System Dynamics Society, Quebec, Canada, July
20-23.
Homer, Jack B. (1987). A diffusion model with application to evolving medical
technologies. Technological Forecasting and Social Change 31(3) 197-218.
Lyneis, James M. (1998a). System dynamics in business forecasting: A case study of
the commercial jet aircraft industry. Proceedings of the 1 6" International Conference
of the System Dynamics Society, Quebec, Canada, July 20-23.
Lyneis, James M. (1998b). System dynamics for business strategy: A successful
approach. Proceedings of the 1 6" International Conference of the System Dynamics
Society, Quebec, Canada, July 20-23.
Maier, Frank H. (1998). New product diffusion models in innovation management—a
system dynamics perspective. System Dynamics Review 14(4) 285-308.
Shiba, Shoji, Alan Graham, and David Walden (1993). A New American TQM: Four
Practical Revolutions in Management. Productivity Press, Portland OR
Sterman, J. D., Repenning, N., & Kofman, F. (1997). Unanticipated side effects of
successful quality improvement programs: Exploring a paradox of organizational
improvement. Management Science 42(4), 503-521.
Systems Thinker (1998). A systems view of communicating change: The Navy
acquisition reform team. The Systems Thinker 9(9), 9. Pegasus Publications,
Waltham, Mass.
Vennix, Jac A.M. (1996). Group Model Building: Facilitating Team Learning Using
System Dynamics. John Wiley & Sons, Chichester.
Vennix, Jac A. M., David F. Andersen, and George P. Richardson (1997). Special
issue on group modeling. System Dynamics Review 13(2).
Wolstenholme, E. F. (1999). Qualitative vs quantitative modelling: The evolving
balance. Journal of the Operational Research Society 50(4), 442-429.
Wolstenholme, E. F. and R. G. Coyle (1983). The development of system dynamics
as a methodology for system description and qualitative analysis. Journal of the
Operational Research Society 34(7), 569-81.
System Dynamics and Systems Thinking: It Takes All Kinds
CC BY-NC-SA 4.0