Dynamic Systems Modeling in Educational System Design & Policy
Dynamic Systems Modeling in Educational System Design & Policy
Vol. 2. No. 2. July 2013 pp. 72–81 ISSN: 2254-7399 DOI: 10.7821/naer.2.2.72-81
ORIGINAL
Figure 1. Example of system dynamics analysis in policy design (Source: Forrester, 1969; Forrester, 1991).
The other root is from the human tendency towards analysis— in educational system change—ultimately propelling the 1983
breaking things down to examine each piece, which is where release of A Nation at Risk, spurring the accountability
much of our conventional wisdom comes from and is generally movement of the 1980s and the introduction of large-scale
how we are educated (Meadows, 1991). As a result of our daily governmental regulation through mandated curricula and
interactions combined with our formal education, we all walk competency testing.
around with mental models of how the world works, which are For many, the publication of A Nation at Risk marks a turning
abstractions based on this world experience. Our mental models point in the discourse and emphasis on education reform;
and abstractions drive our behavior, from the very simple however, the outcomes post-1983 generally mirrored the level of
shutting of a door, to the attempts to create solutions to the impact and challenges encountered before. The 1980s became a
world's most complex challenges. However, mental models have time of strategy towards decentralization, where schools had
serious shortcomings. We are often unaware of assumptions increased ownership of their decision-making, allowing more
about how things work because our mental models are flexibility to meet the needs of the specific context of each
incomplete—"limited, internally inconsistent, and unreliable" school became the mode of emphasis. Yet with many of these
(Sterman, 2001, p. 10). Forrester explains, efforts operating on emphasis and strict adherence to
implementation guides, school were not left with the flexibility
the human mind is not adapted to understanding correctly the
to adopt and adapt a program to meet their needs (Fullan, 1991).
consequences implied by a mental model. A mental model may be
Michael Fullan, Professor Emeritus of the Ontario Institute for
correct in structure and assumptions but, even so, the human
Studies in Education of the University of Toronto, arguably has
mind—either individually or as a group consensus—is apt to draw
the strongest track-record of effective strategies in large-scale
the wrong implications for the future (1971, p. 5).
reform concluded that "restructuring reforms that devolved
How can we be so sure? Over the last 50 years the field of decision making to schools may have altered governance
system dynamics has been leveraging the power of dynamic procedures but did not affect the teaching-learning core of
systems modeling and computer applications to demonstrate the schools" (1995, p. 230). Countless case studies of districts
complexity of our world and how understanding that complexity attempting restructuring initiatives have demonstrated that the
can help us to predict the behavior of complex systems and the core goal of improving teaching-learning practices to increase
appropriate policies we should craft for those systems. student performance had not occurred. This sporadic progress
left many education stakeholders and reformers spurring towards
2 THE NEED FOR A NEW APPROACH: A large-scale, comprehensive reform that dominated the 1990s.
HISTORY OF MIRED PROGRESS AND In fact, Fullan has argued that the largest hindrance to reform
PERILS OF EDUCATION REFORM is the presence of too many ad hoc, uncoordinated innovations
and policies (1999). Whole School Reform Models emerged in
Since its appearance in the U.S. in the 1960s, education system response to this, where the entire school seeks to adopt a new
reform has a long and varied history with limited success. model (examples include Success for All, Expeditionary
During these decades we saw considerable investment of federal Learning Schools, CO-NECT Schools). These models have
funds towards various programs designed to produce better shown varied success depending on their implementation. On a
learning outcomes. However, according to a comprehensive review of 16 initiatives, (Datnow & Stringfield, 2000) found that
study conducted by the Rand Corporation, most of these whole-school reform models can have successful outcomes
programs – both large and small – came up short, explaining that when strong district and state support is present. Unfortunately,
many of these programs were adopted due to the incentives for in hindsight, many schools adopted a certain model without the
the federal funds rather than in pursuit of deeply changing consideration of how it will fit the school's goals, culture,
educational practice, with few programs being implemented teachers, or students, resulting in minimal positive outcomes and
properly in the first place and demonstrating little if any sustained improvement. This has ultimately pushed Fullan and
sustained growth (Berman & McLaughlin, 1978). With this lack others to believe the only effective method for systemically
of growth and disconnect, and only isolated pockets of improved improving large school systems is to not pursue a top-down nor
classroom instruction, the 1970s showed to be a stagnant period a bottom-up approach, but rather an approach that holistically
73
Groff, J. / New Approaches in Educational Research 2(2) 2013, 72-81
works with all levels together—what Fullan calls the Tri-Level Unfortunately, these tools have rarely been used in education
model that address the three critical levels of school systems: policy. Despite the fact that several notable researchers began
school & community, district and state (Fullan, 1994; Fullan, introducing these tools to the education reform discourse over
2001; Fullan, Rolheiser et al, 2001; Barber & Fullan, 2005). two decades ago, one can speculate that this lack of presence in
In reflecting on the journey of education reform, Fullan education reform work is largely due to people’s lack of
concludes: "as long as you have external models coming and familiarity with complex systems and systems thinking—as
going there will never be more than a small proportion of evidenced by the design of most of the reform work of the past
schools and districts involved, and any pockets of success will 50 years. Though sparse, these rare examples are worth noting
be short-lived (2001, p. 4);" therefore, "the primary goal is to here.
alter the capacity of the school to engage in improvement" [and] Wheat has illustrated unforeseen dynamics of student
second, “sustainable reform can only be achieved when working achievement when Standards of Learning were introduced to an
with whole systems" (2001, p. 5). Aligning the whole system is education system by generating an unintended side-effect that
critical, because no matter what capacity and gains a school undermined rather than supported the standards (2000). Wheat
makes, if the outer policies undermine those initiatives, the created a model that showed a “70-percent achievement rate” on
school improvement cannot and will be sustained. the assessments introduced to accompany the standards, which
Fullan’s whole-system approach shows promise, but it is clear to many seemed quite reasonable; however, as this goal is
to most education stakeholders that we are still far from having achieved over time, it will induce an expected increase in
the optimal education systems we desire. What truly is expectations that is unsustainable, impacting student/teacher
happening in these reforms efforts that is inhibiting the change motivation and “thereby reducing learning productivity and
we seek? What hidden effects and dynamics are in place that causing learning rates to be lower than they otherwise would be”
actually produce the outcomes we see rather than those we (2000, p. 7).
desire? How can this happen? Unfortunately, most legislation is
designed with linear, cause-effect thinking, and it is passed with
3 SYSTEM DYNAMICS AND EDUCATION little or no analysis of its system effects by either policymakers
POLICY or advocates. Jay Forrester, the father of SD, explains that
"governments pass laws after superficial experiments using a
Fullan’s whole-system approach actually preludes an approach country as a laboratory...including no dynamic modeling of the
and mind-set employed by many other societal systems and long-term effects" (1998 p. 6). Herein lies the greatest potential
structures known as System Dynamics modeling—a in SD tools—testing new policies and system changes, rather
methodology that helps us better design and respond to the than using the current education system as a test bed for policies
complex systems that make up our world. generated with human cause-effect thinking. Once a model is
In reality, all of our societal structures and complex systems built using SD tools, one can better plan for policies and changes
and what Fullan began to observe and response to is the nature to the system by forecasting their dynamics. Although the
these systems, which are complex, interconnected structures that method can be time-consuming and challenging, it offers the
are filled with feedback loops where behaviors and actions in best way to "gain experience with a system, because activity in
one part of the system impact (or are constrained) by other parts the real system is infeasible, costly, or impossible" (Banathy,
of the system. They are characterized by nonlinear, 1973; Garet, 1974). Forrester advocates for this type of
counterintuitive behavior, where not only is effect rarely 'designing the future'—"people try to cope with the failures of
proportional to cause, but often they are far apart in the system. systems, but seldom attempt to redesign systems to reduce
In other words, effects or changes to one part of the system often failure" (1998, p.1).
play out much differently than intended because the change
causes dynamic effects in the system as a whole. This 3.1 Dynamic Complexity
complexity makes all systems inherently difficult to understand What makes these systems so complex? And as a result,
intuitively and therefore policy resistant—because our decisions complicated and difficult to really see what’s happening?
often elicit unintended responses (Sterman, 2001). The As stated previously, we all walk around with mental models
education system is no different from other complex systems. of how perceive things to be operating. Yet, generally speaking,
There are many examples in current education policy where the these models are grossly insufficient. What makes our mental
outcomes are very different than what had been intended (Groff, models so ineffective – like those that likely were the cause in
2009; Wheat, 2000). An example of this is provided in Box 2. the scenario in Figure 1 – is not their lack of complexity or lack
System Dynamics (SD), a methodology and larger field of of understanding of the complexity of the system in terms of
study, was developed more than 50 years ago in an effort to cope number of possibilities in making a decision (also known as
with the complexity and difficulty in working with complex combinatorial complexity). Rather, it is usually due to dynamic
systems, and has subsequently been applied to numerous fields, complexity—“the counterintuitive behavior of complex systems
including businesses, medicine, economic behavior and even that arises from the interactions of the agents over time”
environmental change (Forrester, 1998). In essence, this (Sterman, 2001, p. 11). The challenging differences between
methodology helps us create models of the key dynamics in a simple and complex systems are many and often direct opposites
given system, by offering tools to: of one another. For example, in a complex system the actual
⎯ map the feedback structure of a system in order to cause to a behavior may originate from another part of the
understand why a system is behaving the way it is; system, often from a place that is distantly removed; in complex
⎯ test and plan for policies before implementing them; and systems, achieving a short-term goal can often mean undesirable
long-term consequences; and in complex systems the obvious
⎯ to increase the likelihood they produce the outcomes
decision often turns out to be an ineffective one (Forrester,
desired.
1997). And of course adding to the complexity of decision-
74
Dynamic Systems Modeling in Educational System Design & Policy
making is the very hierarchical nature of complex systems—the complexity of a system that we might otherwise not recognize;
goals of a subsystem can contradict or endanger the welfare of and it counteracts the tendency towards analysis by seeing the
the larger system. system as a whole—which makes the field inherently
An extended discussion of the dynamic complexity of systems interdisciplinary.
is described in Figure 2. It is this type of complexity, and the A system dynamics model is a representation of the structure
lack of understanding its nature, that leads to policy resistance of a system. Like all models, these models are never fully
and decisions that lead to unintended systemic effects. And it is accurate depictions of the behavior of our world. However, the
this complexity that requires to rely on more than our mental tools of system dynamics afford us a means of employing the
models for analyzing complex systems and creating policies and knowledge that underlies our mental models with a more
structures that govern the futures of these systems. Doing so has accurate representation of the complexity of the system. Systems
been achieved with considerable success in fields from medicine are the fundamental structure to our world. Simple systems are
to environmental change to regional economics using an nested within larger complex systems, which are nested within
approach known as system dynamics. in larger complex systems, and so on. Employing system
dynamics to construct a useful model requires determining the
Figure 2. The Dynamic Complexity Characteristics of Complex
scope of the system you will examine. Figure 3 diagrams the
Systems (adapted from Sterman, 2001, p. 12).
hierarchical complexity of the U.S. educational system. A
Dynamic Complexity Characteristics of Complex Systems system dynamics analysis can be conducted within any of these
⎯ Constantly challenging – Change in systems occurs at many time levels, or between one or more of these levels. Modeling these
scales, and these different scales sometimes interact. different dynamics will depend on defining the scope of the
⎯ Tightly coupled – The actors in a system interact strongly with problem, which is achieved through the four general tools in
one another and with the natural world; everything is system dynamics – described in Figure 4. People are generally
connected to everything else. most familiar with Behavior-Over-Time Graphs—they help
⎯ Governed by feedback – Our actions feed back on themselves, identify a problem or undesirable behavior and therefore a
giving rise to a new situation as a result of our actions. starting place for a system dynamics project. Once the "what"
⎯ Nonlinear – Effect is rarely proportional to cause, and what has been defined, Causal Loop Diagrams and Stock/Flow Maps
happens locally in a system often does not apply in distant will help identify the "why."
regions; it arises as multiple factors interact in decision-
making. Figure 3. Hierarchical Levels of Educational System Policy Analysis
⎯ History-dependent – Taking one road often precludes taking
others and determines where you end up; many actions are Federal
irreversible. State
⎯ Self-organizing – The dynamics of systems arise spontaneously District
from their internal structure, generating patterns in space and School
time creating path dependence. Classroom
⎯ Adaptive – The capabilities and decision rules of the agents in
Student Teacher
complex systems change over time. Adaption also occurs as
people learn from experience, especially as they learn new
ways to achieve their goals in the face of obstacles. Learning is
not always beneficial, however.
⎯ Characterized by trade-offs – Time delays in feedback channels
mean the long-run response of a system to an intervention is
often different from its short-run response. High leverage
policies often generate transitory improvement before the Constructing formal models exposes the assumptions and
problem grows worse. miscalculations in our mental models. While no model, formal
⎯ Counterintuitive – Cause and effect are distant in time and space or informal, will ever fully and accurately represent the real
while we tend to look for causes near the events we seek to world scenario, the usefulness and validity of a constructed
explain. model can be based on its ability to clarify our thinking and
⎯ Policy resistant – The complexity of the systems in which we are provide us with insights the system; therefore models can be
embedded overwhelms our ability to understand them, judged on (Forrester, 1968):
resulting in many seemingly obvious solutions to problems
⎯ their clarity of structure, particularly in comparison to the
that fail or actually worsen the problem..
clarity found in the verbal description;
⎯ whether or not the underlying assumptions are more clearly
3.2 System Dynamics exposed;
"The question is not to use or ignore models. The question is ⎯ the level of certainty with which they demonstrate the
only a choice among alternative models" (Jay Forrester, 1971, p. correct time-varying sequences; and
4) ⎯ the ease of communicating their dynamics compared to the
Understanding the structure and interconnections that create verbal description.
the behavior of a defined system is the goal of system dynamics.
Models are tools; therefore a useful and valid model is
System dynamics is a field that helps us to overcome policy
deemed so when it is determined to be the most useful and valid
resistance and the inherent human limitations described above.
tool for understanding the situation at hand. By using these tools
The tools of system dynamics demonstrate and unpack the
75
Groff, J. / New Approaches in Educational Research 2(2) 2013, 72-81
one begins to develop systems thinking, critical to overcoming for this purpose are not accurately capturing student
policy resistance. The next section will expand upon applying achievement and cognitive growth in a domain (Dawson &
system dynamics modeling in education system design; Stein, 2008); while discussing this is outside the scope of this
therefore a basic understanding of system dynamics modeling paper, it is worth highlighting that assessments used to capture
will be necessary. For a brief introduction to the concepts and the outcomes of a system are human-made, and potentially
building blocks of system dynamics models, please stop here seriously flawed—which can dramatically affect how the system
and review Appendix – An Example of System Dynamics in is perceived and therefore adjustments that are made to the
Play: A Generic Model of the Spread of Contagious Activity, or system in a damaging way. Yet even if the assessments
Infection. developed to measure and track the status of the "student
achievement" stock over time, there are countless variables
Figure 4. Tools of System Dynamics (adapted from Catalina Foothills within a system that affect this, and therefore their influence in
School District, 2003). the system must be explored. Just a few of the possible stocks
that could be measured and diagrammed include:
Behavior-Over-Time Graphs – Displays ⎯ teacher-student ratio
data of change in the system in a line graph ⎯ curriculum scope
format, where time is plotted on the X-a xis.
⎯ teacher attrition
⎯ student graduation rates
Causal Loop Diagrams - Understanding why ⎯ school accountability ratings
the behavior over time is occurring requires
understanding the dynamics of the system—
Some of these stocks, and many more, dramatically impact
which generated by causal and feedback loops student achievement and therefore are critical dimensions to a
within the system. Causal Loop Diagrams systems analysis of education. Because of the complexity of the
help us to map loops visually, showing how hierarchies and the variables within and amongst those
they may interact with one another. hierarchies, it may feel overwhelmingly complex to try to map
these dynamics. However, the good news is that any concept that
Stock/Flow Maps - "Stocks" are
can be clearly described in words can be incorporated into a
the accumulation of something in
the system, such as money,
computer model (Forrester, 1971). The large benefit to mental
people, etc. "Flows" are the rates models is that they represent a vast body of knowledge about the
of change of those stocks, such as dynamics of the system, which is used to construct more
savings or spending rate. accurate representations and simulations. Likewise, data is a
Feedback loops within a system are what control these flows. Through critical piece to constructing good computer simulations and
these three components, one can depict the dynamics of a given system. models. One of the greatest strengths of the field of education is
To ols for creating both Causal Loop Diagrams and Stock/Flow Maps the robust and vast array of research and data, which can be used
include STELLA, Vensim, and DYNAMO. for these purposes. Leveraging these two resources will allow us
Computer Simulation to construct powerful models of education—affording us the
Models - Once a system is ability to understand the current dynamics and the design
diagrammed, its accuracy stronger policies and interventions going forward.
can best be tested through Below is an example model based on educational research at
constructing a computer the national level, modeling dynamics occurring in the U.S.
simulation of that model. educational system.
While no one person
could simultaneously 4.1 Classroom Size and No Child Left Behind
calcu late the interdependent relationships of system of time that
produces the troublesome behavior, a computer model can. Numerous Class size and student-teacher ratio has long been considered a
tools have been developed to help achieve this, including StarLogo, and critical factor on student achievement. Although it's tangible
NetLogo. impact on student performance has been largely debated, the
Center for Public Education conducted a meta-analysis on this
literature and have several key findings, including: smaller
4 USING DYNAMIC SYSTEMS MODEL IN THE
classes, particularly in grades K-3, can boost student academic
EDUCATIONAL SYSTEM DESIGN achievement; a class size of no more than 18 students per
Given the magnitude and complexity of many national teacher is required to produce the greatest benefits; and minority
educational systems, system dynamics offers a set of tools and a and low-income students show even greater gains when placed
way of thinking that can not only help us unpack the in small classes in the primary grades (Center for Public
complexities and drivers of the existing systems, but also offer Education, 2007).
us the promising hope of strategically crafting future system From these findings, we can conclude that class size is a stock
policies and structures that better enable us to create more that should be monitored. For example, in the mid-1990s,
effective educational systems. The aforementioned hierarchies California's average elementary classroom size was 29 students
found in the U.S. educational system in Figure 4 show the (Bohrnstedt & Stecher, 2002). By the aforementioned findings,
various levels that systems analysis can occur within and this number is far beyond the scope of an acceptable ratio. The
between, on numerous variables (stocks). Of course, the most status of this stock could alert us to consider designing and
obvious – and the most monitored – system outcome or stock is implementing policies that help lower this ratio. In system
"student achievement," measured in standardized test scores. dynamics language, we might try to understand the current
However, some experts argue that most of the assessments used
76
Dynamic Systems Modeling in Educational System Design & Policy
dynamics of the system, in order to create new feedback loops models can also be used more locally as well. The next step
that bring this stock down. would be for a more localized school system to expand these
We can begin to understand causes of change in a stock such models, adding the dynamics that are relevant for their current
as this by identifying the influences or system elements effecting situation, and then generate a simulation of that model with
this change using causal loop diagrams. Often these diagrams computer modeling software. Such a process has been
are based on the stores of mental models and knowledge we demonstrated and leveraged powerfully in numerous other
hold. One causal loop we could identify is related to current domains to expose the unintended dynamics of system design,
policy structures of the No Child Left Behind Act (NCLB). This and therefore has much to offer the field of education. In the
law, enacted in 2001, sought to improve student performance of next section, we will demonstrate this process through a case
primary and secondary schools through increasing achievement study on Rhode Island and the effects of NCLB on curriculum
standards tied to federal funding. Under the law, schools that and instruction in that system.
consistently underperformed were subject to loss of funding and
potential takeover by the government. While the policymakers Figure 6. Causal Loop Diagram of reinforcing loop in the No Child Left
undoubtedly crafted this law with the intent to "push" educators Behind Act with reinforcement from state funding dynamics in 2002.
to improve their teaching to provide all students with an
adequate education, we can see an unintended reinforcing loop
–
that can ultimately have the directly opposite effect; Figure 5 is a Student
Causal Loop Diagram demonstrating these dynamics. If schools Archievement
77
Groff, J. / New Approaches in Educational Research 2(2) 2013, 72-81
⎯ Increased classroom time spent on tested subject. Study construct a computer simulation based on actual student
participants reported spending more instructional time on achievement data and other data that can be collected such as
English-language arts and math, which removed class frequency of various pedagogies and test prep in the classroom.
time formerly spent in other areas. Even in its current form, we begin to get an understanding of
the dynamics at play in the context of this case study. The model
From this data, we can begin to construct our model (see
suggests that classroom content and behaviors are increasingly
Figure 7). We begin by identifying our stocks; in this case, the
aligning to those that are directly related to test preparation. On
curriculum (or subjects) is our stock. Subjects taught and those
one level this may be easily dismissed as a given dynamic of
not taught represent the two stocks that embody this, and
having summative tests in education. Yet analyses such as those
therefore the rate of change between them we will call the
conducted by the CEP, followed by an SD model can be a
“content alignment” rate. Now that we have the focus of our
powerful way of shining light of unintended – and highly
model, the goal is to uncover the dynamics that affect this rate.
negative – consequences. For example, having the model with
Some of these dynamics are very straightforward. For example,
supporting data from participants describing the decreased time
as teachers begin to focus more classroom time on test material,
and attention to any curricular topics other than ELA and Math,
the “Subjects Not Taught/Tested” stock increases, which
as well as data emphasizing the amount of class time that is
increases the “class time available.” As “class time available”
spent in close-ended question discussion demonstrate two key
increases, this increases the “exposure to test content,” creating a
dynamics that fly in the face of learning theory and research on
reinforcing loop. “Exposure to test content” is also increased by
the design of high quality learning environments (OECD, 2010).
the decrease in the stock “Subjects Taught/Tested.” This
Whereas educators and other education stakeholders may have
variable, “exposure to test content,” has two reinforcements and
previously only voiced concerns and disagreements about the
therefore the loop must be completed.
dynamics and directions of schools and classrooms in light of
high-stakes testing, SD modeling offers concrete tools to take to
Figure 7. Stock/Flow analysis diagram of dynamic effects of the No policy makers in identifying detrimental dynamics and
Child Left Behind act in the state of Rhode Island in 2008, from the advocating for better policy and system designs.
research of the Center for Educational Policy (Srikantaiah, Zhang &
Swayhoover, 2008).
6 CONCLUSIONS
Content Alignment
Rate How can system dynamics modeling be used more proactively
8
Subjetcs Not
Subjetcs Taught/
Taught/Tested
to design educational systems of the future? Avoiding the
Tested
creation and implementation of policies that have dire effects in
the system is of critical importance as the complexity and
magnitude of our educational systems continue to expand and,
B R
Students most importantly, the trajectories of our children’s learning is
+ Proficient altered as a result.
exposure to test
The Organization for Economic Co-operation and
+
content Development (OECD) has been leading this consideration with
8
+ class time
avaliable their work on Think Scenarios—which uses scenario modeling
+
to inform educational leadership and policy-making (OECD,
Students Not
Proficient 2006). One approach within this work is the “possibility-space”
approach—a method for generating a larger set of future
scenarios, to get beyond the limitations of “trend-based”
From the data collected by the researchers at the CEP, study scenarios and “preference-based scenarios” (Miller, 2006). In
participants eluded to the fact that pressure to teach to the test this approach, Miller notes the parallels and tensions in the work
came from the need to meet minimum levels of student done by historians and futurists in scenario modeling (2006).
proficiency. This identifies the second stock that must be Both use “clues in the present and past in order to substantiate
included in our diagram, separated into “Students Proficient” their analyses of why and how life did not or might unfold, using
and “Students Not Proficient”—where “proficiency” is a term methods and theories that take into account multiple layers of
defined by NCLB in identifying those students who perform at a complex interaction and causality” (p. 96). Historian-derived
desirable level on the standardized test. The flow between these scenarios tend to be limited in vision, locked-in by existing
two stocks is affected by the variable “exposure to test content” trends—whereas futurist-based scenarios tend to lack accuracy
thereby completing the loop and our stock/flow diagram. and practicality. As a result, there are two fundamental problems
with these approaches: (1) risk of narrowness and lack of
imagination, and (2) lack of analytical precision.
5 DISCUSSION
System dynamics modeling can bridge these problems. By
The Rhode Island case study presented here offers a very basic leveraging this approach and marrying it to the “possibility-
application of SD modeling for educational policy in order to space” approach, educational stakeholders can strategically
demonstrate the SD process and approach. Yet in order to more design future scenarios for education that generate desirable
accurately represent the dynamics of this case study, we must outcomes and are supported by data and models of the past.
consider adding additional stock(s)—particularly those Very recently, researchers have made significant advances in
addressed in this case study such as the pedagogical approach of applying system dynamics and modeling tools to guide future
the teacher (closed questions versus open-ended student policy development based on analysis of past and current
inquiry). Next steps would be to engage study participants with dynamics in education, particularly as it relates to the
our model to elicit their feedback on its validity, and then development of science and math students. Developed by the
78
Dynamic Systems Modeling in Educational System Design & Policy
Raytheon company over the past three years, the model process ubiquitous, and most adults have not had the opportunity to be
data on variables such as class size, teacher turnover, gender exposed to experiences and instruction that help them cultivate
differences, teacher salaries, and data from scholarly research. the capacity of systems thinking.
What is most commendable about this initiative is the support Advancing our education systems forward requires that we
website for the tool, designed to allow anyone who registers to take the first step toward systems thinking, and that we begin to
freely download the model and contribute to the analysis and explore this mindset and the principles of complex systems so
discourse of it—this is where modeling is most powerful, when that we can make more informed decisions and utilize more
the tacit knowledge carried by stakeholders within in the system informed tools like SD modeling. The knowledge is there, with
is allowed to rub up against and help form the model, which countless examples of how it has worked for other domains, it’s
dramatically increases its validity. Brian Wells, Raytheon’s up to us to cultivate the will to use it.
chief systems engineer who helped develop the system, recently
explained in an Education Week article that “the model is not REFERENCES
meant to provide definitive solutions, but rather help Axelrod, R. (1976). The structure of decision: The cognitive maps of the
policymakers ‘think through the problem, [and] can help you political elites. Princeton, NJ: Princeton University Press.
discover unintended consequences’” (Cavanagh, 2009). This is a Banathy, B. (1973). Developing a Systems View: The Systems Models
dramatic and noteworthy step in the field of educational policy. Approach. Lear Siegler Fearon Publishers.
Barber, M., & Fullan, M. (2005, March 2). ‘Tri-level development’:
6.1 Designing the Future Putting systems thinking into action. Education Week, 24(25), 32-35.
Berman, P., & McLaughlin, M. (1978). Rethinking the Federal Role in
Moving forward in efforts to change our schools to be effective, Education. Santa Monica, Clif: Rand Corporation.
Bohrnstedt, G., & Stecher, B. (Eds.). (2002). What we have learned
sustainable institutions that meet the needs of all learners for the
about class size reduction. Sacramento, CA: California Department
21st century may feel like a nebulous, daunting task. However, of Education.
using the lessons we have gained from education reform, Catalina Foothills School District. (2003). Tips for using system
education research, and systems work we can build a plan for dynamics tools. A paper for the Creative Learning Exchange,
effectively designing these systems, rather than seeking to make accessed at
www.clexchange.org/ftp/documents/Implementation/IM2003-
partial changes that have little opportunity to be sustained by the 12TipsUsingSDTools.pdf on March 2, 2009.
system (Hargreaves, 2003). Transformation means strategically Cavanagh, S. (2009). Computer tool sizes up math, science policies.
working within the three systems levels (see Figure 3) to Education Week, 29(36). Retrieved from
redefine the teaching-learning core through disciplined www.edweek.org/ew/articles/2009/07/08/36simulate.h28.html.
Center for Public Education. (2007). Research review: Class size and
innovation and strategically designing and building an
student achievement. Retrieved from
infrastructure to systemically support it. http://www.centerforpubliceducation.org/site/c.kjJXJ5MPIwE/b.153
Of course, this is a tremendous task to say the least. However, 3647/k.3B7C/Class_size_and_student_achievement.htm
the dynamics and effects of the current system demonstrate the Datnow, A. & Stringfield, S. (2000). Working together for reliable
critical need for this type of strategic analysis and design. The school reform. Journal of Education for Students Placed at Risk,
5(1), 183-204. doi: 10.1080/10824669.2000.9671386
questions about the future of the educational system to meet the Dawson, T, L. & Stein, Z. (2008). Cycles of Research and Application in
needs of the 21st century, combined with our current state of a Science Education. Mind, Brain, and Education, 2(2), 90-103. doi:
tremendous array of educational research, makes this an 10.1111/j.1751-228X.2008.00037.x
opportune time for this type of strategic design. Leveraging Forrester, J. (1998). Designing the future. Address at Universidad de
Sevilla, Sevilla, Spain. December 15, 1998.
dynamic systems modeling to meet this goal can help us create
Forrester, J. (1997). System dynamics and K-12 teachers. Lecture given
an educational system that is effective and fair for all of our at the University of Virginia School of Education on May 30, 1996.
students. D-4665-4
Forrester, J. (1991). System Dynamics and the Lessons of 35 Years. A
6.2 Using SD to Move the Field Forward chapter written for De Greene, K. (eds.) The Systemic Basis of Policy
Making in the 1990s. D-4224-4 Accessed on March 28, 2009 at
System Dynamics modeling is a powerful methodology and tool http://sysdyn.clexchange.org/sdep/papers/D-4224-4.pdf
set that has made a dramatic impact in a variety of fields outside Forrester, J. (1971). Counterintuitive behavior of social systems.
education. It represents a transformative leap forward in how we Technology Review, January 1971, 52-68.
Forrester, J. (1969). Urban dynamics. MIT Press: Cambridge, MA.
view the world, and in turn respond to it and help design it. One
Forrester, J. (1968). Principles of systems. Waltham, MA: Pegasus
can only speculate as to why such an advancement in approach Communications.
to dealing with complex society systems has been leveraged so Fullan, M. (2001). Whole school reform: Problems and promises. Paper
effectively in so many other domains yet remains outside the commissioned by the Chicago Community Trust. Retrieved from
walls of education, but the answer is regardless of the need to http://www.extranet.ou.nl/inter-studie-alg-
o31411/Symposium4/materiaal/Whole%20school%20reform%20-
ignore it no longer. Given the state of many schools and %20Fullan.pdf.
educational institutions today, there is not only a clear need for Fullan, M. (1999). Change Forces: The Sequel. Philadelphia, PA:
educational to cross this threshold and being leveraging SD but Falmer Press, Taylor & Francis Inc.
there is also an urgency. At the heart of SD modeling lies Fullan, M. (1995). The school as a learning organization: Distant
dreams. Theory into Practice, 34(4), 230-235. doi:
systems thinking—a competency or capacity of an individual to
10.1080/00405849509543685
see the world as interrelated systems rather than linear and Fullan, M. (1994). Coordinating top-down and bottom-up strategies for
distinct parts (Kauffman, 1980). It’s an different frame of educational reform. In R. Anson (Ed.) Systemic Reform: Perspectives
mind—one that can and is taught all the time by the numerous on Personalizing Education, 7-24.
educators who see the value of such a mindset. Teaching Fullan, M. (1991). The new meaning of educational change. New York,
NY: Teachers College Press.
systems thinking is now commonplace in elite business schools Fullan, M., Rolheiser, C., Mascall, B., & Edge, K. (2001).
like the MIT Sloan School of Management and even in many Accomplishing large scale reform: A tri-level proposition. Prepared
elementary schools today (Meadows, 2008). Yet it is far from
79
Groff, J. / New Approaches in Educational Research 2(2) 2013, 72-81
ACKNOWLEDGMENT
Supporting Agencies: Nellie Mae Education Foundation;
Harvard University
80
Dynamic Systems Modeling in Educational System Design & Policy
the same basic structure. 10 2) Determining what flows are affecting the changes in
The behavior-over-time 5 these stocks
graph represents the 0
0 5 10 15 20 3) Defining the elements in the system that influence the
general number of affected rates of the flows (converters/connectors)
people over time. Example of a behavior-over-time.
4) Identifying feedback from the stocks that makes the
The graph demonstrates a characteristic s-shaped curve of system dynamic
infections over time, with the number of new infections
increasing then declining, cause a depression in the rate of For the dynamics of infection, constructing a stock/flow map
overall infection. What is the underlying system dynamics might look something like this:
causing this behavior? Other system dynamics tools can help us
In an infection epidemic, people are the stock, but this must be
model this system behavior.
noted in two separate stocks—healthy people and infected
Causal Loop Diagrams Since a system is composed of a
people:
collection of items that interact with one another, the parts of the
system have a circular relationship, or a “loop”—where output, Healthy Infected
or behavior, of a system is fed back into the system itself to People People
affect future behavior. Therefore, the central building block for
Causal Loop Diagrams are correctly diagramming the feedback Therefore, the flow affecting the quantities of these two stocks
loops within a system. The type of feedback loop is determined would look like this:
by the effect it has in the system, which includes: Healthy Infected
People People
⎯ Negative/Balancing – system elements seek to “negate” Getting Sick
any changes in the system, thereby pursuing balance and
As this flow has time to work, people from the “healthy” stock
stabilization in the system; this is the most common type
will be moving over to the “infected” stock. How does this
of feedback loop.
action affect the rate or flow of people getting sick? As the
⎯ Positive/Reinforcing – these are the drivers of growth number of people infected increases, so does the chance that a
within a system and amplify change. healthy person will be exposed to the sickness, thereby
Example of a causal loop diagram.
producing a reinforcing loop on the “getting sick” rate.
We can see both loops at work in infectious behavior. As the
Healthy Infected
number of people infected increases, the exposure rate is greater People People
Getting Sick
+
– + R
Healthy People B New Infections R Infected People
+ + Infected Contacts
with Healthy People
+
thereby increasing the amount of new infections. Since the
But how does the “healthy” population interaction with the
number of healthy people is declining, the actual number of new
“Infected Contacts with Healthy People” variable? The number
infections will lessen over time.
of “Infected Contacts with Healthy People” depends on the
Stock/Flow Maps Causal loop diagrams can begin to uncover
proportion, or fraction, of the population that was still healthy.
what is happening in a system, but they are only a rough sketch.
When a larger fraction of the population was healthy, the
Stock/flow maps can more accurately unpack the dynamics of
chance of an infected person meeting a healthy person was
the system by outlining the movement of items in the systems
higher. The “Fraction That Is Healthy” is the fraction of the
and the forces causing that moving. The central components to a
whole class that is still healthy. It represents the likelihood that
stock/flow map include:
an infected person will meet a healthy person.
⎯ Stock – an accumulation of something in the system, either
Healthy Infected
concrete (people, dollars) or abstract (anger); the People People
“nouns” of the system. Stocks can only be affected by Getting Sick
+
flows.
B R
⎯ Flow – action or force that moves things in a system (the +
Infected Contacts
“verbs”); this is always a rate, defined in terms of unit of Fraction That is with Healthy People
stock over time Healthy
+ +
⎯ Converter – policy or information that affects the rate of
flows in a system. *Adapted from Quaden, R., Ticotsky, A., & Lyneis, D. (2009). The
infection game. The Creative Learning Exchange, 18(1), p. 1-12.
81