Steve Peterson uses Stella® to Speed the System Dynamics Learning Curve

Having taught system dynamics to professionals and college students for 40 years, Steve Peterson attests that learning to think in a new way is challenging for most. “System dynamics requires mental discipline,” says Peterson. “The best way to learn how to do it is to do it.”

Peterson’s students get a big assist from Stella®. “isee systems designed a unique tool that helps users think through a system and investigate a question,” says Peterson. “Rather than formulating and correcting differential equations, they can focus on how a system works.”

This is the 40th year of Peterson’s career in system dynamics, a milestone that coincides with isee systems’ own 40th anniversary year. In fact, Peterson began his career at isee systems where, for 18 years, he wrote documentation, provided technical support to users, and served as a consultant on user projects.

He got his first taste of teaching in 1985 when he accompanied Barry Richmond, isee systems’ founder and Managing Director, to a workshop for high school educators. “Barry got food poisoning, so I had to step in as workshop leader,” says Peterson. “Participants in the workshop were new to system dynamics and the combination of Stella and an Apple computer was a revelation to them. Jumping into a workshop teaching position was a little nerve-racking but by the end I thought, ‘I can do this!’”

He also learned that he liked teaching system dynamics using Stella. “I was leading a workshop in the late 2000s, and, on the drive home, I was trying to figure out what I wanted to do next in my career. On top of leading workshops, I had taught Barry’s Dartmouth class a couple of times and found it really fun. I thought that if the opportunity arose, I’d love to teach at the college level.”

Going from a neophyte modeler to a conference presenter in 18 classes over 10 weeks is an impressively fast evolution. How does Peterson teach a whole new way of thinking in that short amount of time?

The Process

“I want students to get a sense of thinking in systems and a chance to practice it,” says Peterson. “The course is designed to get them to think through a system model, build it, and gain insights from that experience.”

By their third class, students are thinking about and building feedback loops. “I’m a stock and flow guy,” says Peterson, “but there is huge value in feedback thinking and I want students to develop their understanding of how feedback loops operate. Early in the course, I ask students to work with a simple simulator (below). The task is straightforward: After hitting the “run” button, students move a slider in an effort to get a blue curve to track a red curve on a graph. They find points where the lines meet and then part and are asked to explain the line relationship with a causal loop diagram. It’s an easy way to get students to think about feedback.”

Students get to practice their system dynamics skills by considering issues that are in the news. Every class begins with a short presentation in which 2 or 3 students apply a systems lens to provide insight into a news item. Three weeks in, students shift from thinking with a systems lens to building dynamic models using Stella.

“When I did technical support at isee systems, I got good at fixing problems with users’ models by just doing it,” says Peterson. “That’s why I try to maximize my students’ hands-on practice with modeling.”

Much of the students’ hands-on practice occurs in class. “Class is scheduled from 2:25 to 4:15 pm, which is not a good time to lecture,” says Peterson. “Instead, students complete readings and watch instructional or informational videos outside of class and use class time to work on model building exercises. My teaching assistants [TAs] and I are there to offer one-on-one assistance.”

One model building exercise uses the influenza process or SIR model (susceptible, infected, recovered) to investigate the impact of an outbreak. “Working in teams, students build a model that includes the financial implications of a flu outbreak, immunization availability and uptake, and disease replication,” says Peterson.

After gaining experience in building simple models, students practice auditing and debugging a model created by Peterson. Students review the model, identify problems with the construction of feedback loops, and fix or explain how to fix those problems. They also create a work plan that improves the overall model by refining or adding feedback loops.

Most modeling work is completed with Stella Professional, but students also have access to Stella Architect through the Dartmouth lab. “Stella makes it easy for new model builders to get started,” says Peterson. “They see the distinction between a static system map and a dynamic system model with information output. Using Stella’s output tools, they gain insights by comparing simulation runs that test assumptions and ideas.”

The Projects

Halfway through the course, students begin to work on their term project—a simple, straightforward model that helps them develop insight into a defined system problem. “Their first step is scoping out the project and that’s hard,” says Peterson. “TAs and I help students formulate their question and think about the bare minimum structure needed to detect system behavior and answer their questions.”

One of Peterson’s students had taken a college-sponsored spring break trip to Appalachia. During the trip, she saw how coal towns had been transformed by the shift away from fossil fuels. Once in Peterson’s class, she understood the challenges those towns face as issues that could benefit from system dynamics thinking and modeling.

In her culminating project for the course, she focused on the resource curse that can occur in coal country. A resource curse hits any locality that relies wholly on a single employer or industry that is in decline. In the case of this model, the resource was coal and the locality was a generic county. The project investigated how the economy could be developed during that decline.

The student focused on the feedbacks that connected the coal industry to the local economy.

Through a series of model runs, the student tested the economic impact of job automation, wage increases and marginal propensity to consume, and outside and inside capital investments.

Her model indicated that simply raising coal worker wages was not enough to boost the local economy. Reviving coal country would require replacing coal jobs through high quality local investment, augmented by outside investment from more stable industries and employers.

The Outcomes

New to system dynamics and Stella as they are, Peterson’s students do not shy away from big questions. Their models have investigated how rural hospitals might avoid closure, the intergenerational trauma of sexual abuse, the epidemiology of Ebola in West Africa, and how green energy adoption could cause a utility death spiral in the electric sector.

Conference presentations of student models are just one mark of Peterson’s teaching success. He’s just as interested in other measures of their growth as system dynamists and model builders.

“The ability to apply system dynamics gives students a competitive advantage in graduate school or the workplace,” says Peterson. “It teaches you to investigate a question, not just adopt someone else’s idea.”

U.S. military veterans embody the qualities of service, strength, and resilience. While most veterans are thriving today, still too many U.S. veterans die by suicide. According to the U.S. Department of Veterans Affairs, suicide is the second leading cause of death among veterans under 45 years old, and more than 125,000 veterans have died by suicide since 2001.

“Many students will start their term project wanting to prove something,” says Peterson. “I teach them to learn by trying to disprove something. You build confidence in a hypothesis or possible solution by disproving other possible solutions. That’s what the student studying the resource curse did by modeling counterfactual progressions.”

In another example, a student wanted to understand how climate and green space might impact vermin population dynamics in urban areas. Through modeling, he saw that while these factors had an incremental impact on vermin populations, the feedback from food availability was actually the more important constraining factor.

“When students present their prototype models in class and their peers ask, ‘Why didn’t you use a steady state?’ or ‘Could you have used a step function?’, I know they’re getting it,” says Peterson. “When they say, ‘I’ve learned a different way of thinking,’ I know they’ve gained a competitive advantage.”

SKIP Design pairs Stella® and AI to Tackle Complex Education Issues

There is nothing simple about public education. Public school systems serve whole communities. They include elementary, middle, high, and charter schools. Students, parents, teachers, administrators, neighbors, and taxpayers have competing demands, concerns, and worries. Within that complex environment, schools are mandated to educate students across socioeconomic levels and learning abilities. And, like all public services, they are expected to meet that mandate within their community’s budget.

A public school’s funding is linked to the number of students it serves, making decreasing enrollment a core challenge for many systems. With less funding, schools are forced to slash programs. As programs that enrich the school experience disappear, well-resourced families might opt to leave the system leaving classroom with higher concentrations of students with more complex academic or socio-economic needs. That shift exacerbates teacher burnout. Challenging classrooms, a beleaguered staff and fewer programs impact educational quality and student achievement. Decreases in school and student performance drive families away from public schools, completing the loop by causing a drop in enrollment.

Experienced system dynamists, the SKIP Designed team knew that understanding the St. Louis system, finding reasons for enrollment decline, and testing solutions would require collaborative model building and simulation. “Stella from isee system is our gold standard,” says O’Neill.

O’Neill was introduced to Stella by Dr. Saras Chung, her professor at Washington University in St. Louis and founder of SKIP Designed. Having applied system dynamics and Stella to their work in academia, O’Neill, Chung, and the SKIP Designed team now take a collaborative, system dynamics approach to helping clients define their problems, better understand their systems, and find solutions.

SKIP Designed clients, however, are not typically experienced in system dynamics. “We work to make concepts and technology products accessible to the stakeholder groups that need them,” says O’Neill. While working with stakeholders in St. Louis to tackle declining public school enrollment, SKIP Designed worked with isee systems to facilitate their stakeholders’ path from problem definition to solution identification.

AI-assisted system dynamics

“We started the St. Louis project with an in-depth process that put education leaders and administrators, teachers, and parents in a room to define their problem and map their system,” says O’Neill. “We were capturing great qualitative and quantitative data but are a team of just four people. It took a year just to get the data into a model. New to system dynamics concepts, it was difficult for our clients to understand and collaborate on the model and they had real urgency. We needed a way to make it easier for our clients to engage with system dynamics so they could inform the model we were building.”

The SKIP Designed team built CoModel, a simple tool that helps stakeholder groups create simple causal loops without system dynamics training. Users articulate their problem in CoModel which uses generative artificial intelligence (AI) to turn their descriptions and problem statements into causal loop diagrams.

“Having stakeholders create and share their own causal loops and mental maps was an important step,” says O’Neill. “It brought people together – and there was tension in that room. There were even lawsuits between a few of the parties. Causal loops, like one showing how staff turnover leads to teacher burnout, humanized all sides of an issue and neutralized politics. They saw they’re all part of one system.”

The AI used by CoModel is sd-ai, an isee systems/SKIP Designed collaboration. sd-ai is open source and acts as an intermediary between any modeling software (such as Stella and CoModel) and AI.

Once stakeholders had created and agreed upon their system’s key causal loops, the SKIP Designed team used that “rough draft” to create a Stella model. Using Stella Architect, SKIP Designed also provided an interface that allowed stakeholders to test solution ideas.

Using Stella to test assumptions and solutions

“When we had a functional, agreed upon model, we used it to understand why families, especially the most economically mobile families, were leaving the school system,” says O’Neill. “The best high school in Missouri is located in St. Louis. What was pushing people out of the city?”

One obvious explanation for decreasing school enrollments was the opening of new charter schools, which spread the already decreasing student population across more schools. With fewer students, schools lose resources, struggle to retain teachers, and fail to increase or even maintain educational quality and student performance. Perceiving a degradation in quality, families that are able to move away from the school system. Closing schools to concentrate resources in those that remained seemed like a good solution.

“Using the Stella model, stakeholders found that that 60 or more schools would have to close before remaining schools showed improvement in capabilities and capacity,” says O’Neill. “That wasn’t going to happen. They also discovered that opening more schools would not have the expected effect on increasing enrollment.”

In addition to engaging stakeholders in mental mapping and idea testing, the SKIP Designed team went into communities to gather data from parents. “Our surveys found that community safety was a key factor in school enrollment,” says O’Neill. “Parents were concerned about crime, especially violent crime. If a kid can’t walk to school safely, that’s a big problem.”

When crime drives people out of the city, they leave behind properties that often sit vacant. Vacant shops and houses decrease the tax base and school funding – and vacant real estate often attracts more crime. Meanwhile, as economically mobile families leave the city, schools are left with higher concentrations of students who need extra academic and behavioral support.

While having an important relationship to enrollment decline, community crime was outside the school system’s control, as were the types of students who were to leave and remain. Still, it was important for stakeholders to recognize and consider the impact of outside influences on their system.

Another factor in enrollment decline was embedded in the school system. “The St. Louis system of public and charter schools is complicated,” says O’Neill. “Some elementary schools are K-3, others are K-5. Parents have to choose the middle and high schools their student will attend next. There’s real decision fatigue, especially for families with multiple kids.”

There’s also decision stress. A student might apply to a charter school and not get accepted. They might lose in a lottery for their public school of choice. What happens then? Choice options and outcomes add uncertainty to a student’s educational path. For some families, moving to a district with assigned schools and proven quality is the more attractive educational strategy.

This project has uncovered several reasons for enrollment decline and is spurring progress towards improvements. “The group was able to make shifts in their mental model,” says O’Neill. “There is mutual understanding of the system and how it works and people are ready to collaborate on positive leverage points.”

For example, with a better, shared understanding of how school capabilities relate to student enrollment, schools have started to identify and share areas of strength. “Rather than going it alone, schools are connecting and learning from one another,” says O’Neill.

Stakeholders are continuing to use the Stella model to test ideas for system-wide improvements. Some St. Louis schools are customizing the model to answer questions specific to their own location.

Applying Stella and CoModel to new projects

Expanding on the funding and enrollment questions, SKIP Designed is also looking into teacher retention, an important factor in school performance improvement. “We want to understand what keeps teachers in the profession and St. Louis in particular,” says O’Neill. “Like the enrollment decline model, we’ll gather qualitative data from teachers, hold community listening sessions, and use CoModel to build a mental model that will inform Stella modeling and simulation.”

O’Neill and the SKIP Designed team are most enthusiastic about using CoModel to orient fifth grade students to thinking systemically. “Kids naturally think in systems, but schools don’t encourage it and, by high school, they’re focused on dates and facts,” says O’Neill. “We’re really excited to teach system dynamics to the next generation of problem solvers.”

Stella and CoModel are also at work in several other St. Louis projects. “We’re in the early days of a city-wide project that looks at factors in urban growth,” says O’Neill. “We’re taking an inverse approach to the enrollment decline project. First, we’re using Stella to create a generic city model that includes things like housing, innovation, and transportation. Next, we’ll collect qualitative information through interviews with community members and work with stakeholders and CoModel to build causal maps that inform and improve the Stella model.”

On the Road

In March, co-president Karim Chichakly traveled to Washington, DC to speak with New Hampshire’s congressional representatives. Goldman Sachs’ 10,000 Small Business program organized the trip, whose participants included 60 other small business owners from rural communities around the country. Karim discussed issues that impede small businesses in the granite state, such as uncertainty in the economy, the changing tax code—specifically changing R&D expenses to depreciation, and affordable housing and childcare.

The 43rd International Conference of the System Dynamics Society will be hosted this year in Boston, Massachusetts from August 3rd – 7th. isee systems will once again sponsor and attend this annual event. It is an exciting time for us to meet and talk with our customers and Systems Thinking practitioners. Stop by our booth to chat with us, see a demo of our new software, and get a sneak peek at what’s to come! We will be available throughout the conference to answer your modeling questions, learn about your project, and listen to what changes you would like to see in our software.

Software Update

Our latest software releases introduced many new features and improvements to Stella’s existing functionality. On the interface, we added a data manager widget and the ability to import runs as a button action, giving the interface the same data management abilities as the model layer’s data manager. The interface also picked up the custom shape tool, the ability to print variable values in text boxes and annotations, and sections, which allow users to organize interface pages into groups, much like stories, making it easier to navigate and collaborate during development. On the model layer, we introduced full likelihood computation during calibration, revamped date formatting, and the revolutionary AI-supported CLD development, where a Virtual Assistant lets you harness the power of AI to speed and improve the creation of CLDs. For a complete list of new features, please check out our feature updates.

Stella Users Network

Did you know there is a very active Stella users' group? This is an excellent resource, whether you are a novice, advanced user, or returning user. Ask questions about our software, modeling, or systems thinking and receive answers and comments directly from us, as well as from a community of experienced modelers.  To join, go  here  and click Join This Group at the bottom of the page. Should you have questions or issues about signing up, please email us.