Table of Contents
Go Back
Applying system dynamics modeling for learning
to a messy problem in the public sector
Brendan Miller
31 Fairmount Ave.
Somerville, MA, 02144
Phone: 617-543-4058
bam@sloan.mit.edu
Extended Abstract:
The use of indicators and measures to help communities monitor their progress towards a
sustainable future has been spreading rapidly.' In the Boston, Massachusetts metropolitan
region, the Boston Foundation’s Indicators Project has been spearheading this effort.
Through a broad community process, indicators of quality of life and sustainability were
articulated and synthesized into a collection of 70 indicators group into ten categories
published in 2000 to local and national acclaim.
But important questions remain. Perhaps most fundamental is how these indicators drive
change, especially given the complexity of the social, economic, natural and political
systems that must be influenced. Attempting to improve the individual indicators in
isolation is likely to fail given the interdependencies. What is needed is a model of
problem that can be used for learning and to develop robust policy strategies.
If we consider the problem being modeled to be uneven progress in improving quality of
life and sustainability as measured by the indicators, how can we understand the system
that generates this behavior? Clearly, to make progress on improving all of the indicators
would require understanding the system well enough to find points of leverage. But this
is a quintessentially “messy problem” for group modeling.’ In particular, several
characteristics are important to note:
1. All participants are volunteers. The process itself must remain interesting and
work around their commitments and limited availability.
2. No small group of individuals can be found whose knowledge of the system
would be sufficient to model it. Furthermore, many conflicting explanations for
the structure driving certain behavior exist.
3. The power distance between participants may be large. Experts and average
citizens will need to work side by side if all perspectives are to be represented.
4. The modeling process will take a very long time to complete. Participants are not
likely to participate consistently over the entire process.
5. Many participants are not technologically or scientifically savvy, creating
feelings of alienation from, and distrust of, computer modeling. Likewise,
participants have no inherent interest in the techniques of systems dynamics and
| See the Internation Institute for Sustainable Development's website for more information.
http://iisd1 iisd.ca/measure/default.htm.
? See Vennix, Group model-building: tackling messy problems. System Dynamics Review, 15, pp. 379-
401, (1999).
will not suffer through significant upfront training in system dynamics’ tools
willingly. The process must be very accessible, jargon-free, and its value
immediately recognizable.
This paper describes a community modeling process for the Greater Boston region
developed jointly by the author and the Boston Foundation that emphasized the
modeling-as-leaming approach and addressed the unique characteristics of this public
sector situation. The process was successfully piloted this summer with 150 diverse
stakeholders from the Greater Boston region, drawn from all sectors and demographic
segments, in a series of eight three-hour workshops.
Although our initial plan was to use the standard model-building method, we quickly
found the need to revise the process significantly to address the unique challenges of this
situation. Through iterative revision based on participant feedback, we eventually settled
on a process consisting of the following five interrelated activities. These activities
essentially engage participants in Phase One, “Business Structure Analysis,” of Lyneis’
four-step approach” to modeling for strategy development, adapted for use under the
public sector conditions described above. Attention was paid to what Marjolein van
Asselt called narrowing the “metaphor gap” and communicating in plain, everyday
language so that all participants would feel engaged and understand the process.’
1) Provide context, set tone, and provide ground rules. The long-term change
process was outlined and participants were informed of their role within it,
providing a sense of context and continuity. This was important given that many
participants were there for the first time, or had not been engaged in several
months. Significant attention was paid to creating a space for inquiry and open
dialogue® to make double-loop learning possible in hopes of reconciling the
various mental models that may be held by the participants. This was
accomplished primarily by inspiring people with the possibility of creating a
sustainable future and the use of concrete ground rules.
2) Articulate a shared vision for the future. Participants were asked to share their
vision of what a sustainable Boston with high quality of life would look like in the
year 2030 in as concrete and vivid terms as possible. This activity helped
participants to gain some distance from their current circumstances to think more
boldly and question what is possible. Articulating an ambitious vision of the
possible creates something for the region to “grow into.” Empirically, we have
found that there is little conflict between the visions articulated. By encouraging
participants to think “both/and” rather than “either/or” a shared vision emerges,
which generates a sense of optimism, cooperation and commitment that
corresponds to a shift in the participants’ mental models from win/lose to
win/win.
3 Lyneis, System dynamics for business strategy: a phased approach. System Dynamics Review, 15, 37-70,
1999.
* Marjolein van Asselt, Globally Integrated Assessment Models as Policy Support Tools. Doctoral
dissertation at the University of Twente, Enschede. (DATE? DETAILS?)
° See William Isaacs, Dialogue, Doubleday, New Y ork, 1999.
3) Brainstorm important current trends that will shape the community’ s future.
Participants were encouraged to think not about what data currently exists, but
instead what trends are most important, even if they are very qualitative. These
trends were labeled either “positive” or “negative,” meaning that they were
currently moving towards or away from the vision articulated earlier. These trends
are prioritized and three selected for further inquiry by small groups. This activity
sometimes identified new indicators that would need to be monitored in the
future. At this point in the process, the current reality and trend projections were
contrasted with the shared vision from above, introducing a “creative tension” that
motivates participants to take action and stay engaged in the process.
4) Understand the drivers of the selected trends, looking at the entire Greater
Boston region as an integrated system. Facilitators guided this dialogue towards
the places in the system where leverage is likely to reside. Although one can
never know where leverage resides until a model can be simulated, Dana
Meadows “Leverage Points: Places to Intervene in a System” provides a very
valuable heuristic guide. Using such a heuristic was essential in order to keep
participants interested and to accelerate discussion given the limited time
available for each workshop. Participants were reminded that the leverage points
identified were merely hypotheses that would be need to be tested in order to keep
the ongoing inquiry open.
5) Investigate possible high-leverage strategies for change. Assumptions about
what is likely to generate change were openly examined and challenged in a spirit
of dialogue. These strategies were developed as possible ways to influence the
drivers of the selected trends to ensure they will quickly and effectively move the
community closer to the shared vision articulated earlier in the workshop.
By repeating the workshop multiple times, we were able to develop greater confidence in
the vision, trends, possible drivers, and recommended strategies as certain items would
arise again and again. Formal feedback forms and informal questioning indicated that
participants left the workshops energized by their new insight and hopeful that change
was possible. They will be likely to participate in the future. And by creating an open
dialogue, much of problems associated with perceived power imbalances and incomplete
exploration of the system’s dynamics could be overcome because all voices were
explicitly equally valued, supported by vigilant process facilitation.
Over time, a process that involves more and more stakeholders in workshops like these,
will begin to develop a consensus on a framework for action that will focus efforts where
they will have the greatest leverage. Stakeholders will also be committed to the long-term
change process because they will have a greatly enriched understanding of their role in
the context of the dynamics of the entire system.
But this is just the beginning of long process that can be thought of as an expanding
pyramid. The first participants in the process have new skills and understanding that
ready them to engage in further, deeper investigation. A gain referencing Lyneis’ process,
® The term “trend” used in the community indicators process corresponds exactly with the notion of a
“reference mode” in the systems dynamics modeling literature.
these participants, or some portion of them, will proceed to Phase Two. Meanwhile, more
and more people can be involved in Phase One as the process is refined and more people
are trained to give the workshops. In this way, the modeling process simultaneously
probes deeper and deeper while engaging more and more people in learning about the
systems that define the problem of sustainability and quality of life, generating latent
support and political will for the strategy recommendations that will emerge from the
process.
Stage of the strategy
Time process achieved
IV
Ill
I
#peonle enaaced
I
Subsequent stages of the process have yet to be developed but they will build on the work
already done. For example, small groups of representative stakeholders could be engaged
in causal mapping for various trends, which could then be useful in beginning to map out
the causal relationships between the trends of greatest concern. One of the virtues of
causal mapping as a tool for reconciling mental models is that multiple dynamic
hypotheses can be considered simultaneously. Thus, the process can hold the complexity
and multiple perspectives and explanations that will invariably surface as people
articulate their mental models.
Computer modeling will eventually be necessary and valuable, but it need not be
introduced for a quite a while, probably only in Phase Three, and then only for some
representative subgroup of the total participants. There is much still to learn simply from
engaging the various mental models represented in the process using the heuristics like
the one discussed earlier. And the emphasis on leaming-by-doing rather than learning-by-
lecture or -presentation helps to ensure the process remains interesting and relevant for
participants. Pushing computer modeling too early will alienate participants causing them
to disengage, compromising the credibility of any model that is made as it reflects fewer
and fewer of the diverse stakeholders’ mental models.
Until a single indicator can be articulated that synthesizes all of the other indicators, this
may be the closest we can come to modeling the “problem.” In a very real way, the
modeling process might even serve to assist in defining the problem itself more clearly as
the systemic interrelations of the key trends are better understood and we gain confidence
in our chosen model boundary.
The author:
Brendan Miller is a dual-degree student at the MIT Sloan School of Management and the
Harvard Kennedy School of Government and the focusing on the application of systems
dynamics and organizational learning to the development of socially, environmentally
and economically sustainable communities.
Back to the Top
CC BY-NC-SA 4.0