[MUSIC] Hi, my name is Dan Taber, and today we're going to be talking about system dynamics. So any discussion of system dynamics needs to begin with the question, what is a complex system exactly? Because the word system gets thrown around a lot. Sometimes it's used to refer to a specific type of system such as a school system or the health care system or the immune system. But sometimes it's used more loosely without a clear definition. Some people use systems simply to refer to things that have a lot of pieces to them. For example, you might design a multi component intervention to reduce obesity, in which you have a lot of different activities to promote a healthy diet and promote physical activity. And it could be very well thought out and well designed. But just because it's well thought out and well designed doesn't make it complex. And it doesn't make it a systems intervention because a complex system isn't just about having a lot of pieces. And that's really the first key point that I want to emphasize in this video. As Donella Meadows says in her book on system dynamics, a system isn't just any old collection of things. Systems are not just about having a lot of parts. They're more about the relationships between those parts. And I think the best way to start to define a complex system, is to give examples of them. So the things that I've listed on this slide all represent conflict systems. Cars, cities, schools, basketball teams, highways, the human body. These are all conflict systems because they share certain core features. One common feature that they all have is that they all have goals. They're all designed for a very specific purpose. So a school's goal, for example, is to educate students. A basketball team's goal is to score points and win games. And the human body's goal is to live a long, healthy, happy, productive life. And a system's behavior is going to be driven by that goal that it's trying to achieve. So the immune system's behavior, for example, depends on whether a person's body is reaching its goal to be healthy. If the body's infected by a virus, then the immune system's going to adapt accordingly to restore the body's health. And that's how systems work. They adapt based on whether they're reaching their goals or not. Another common feature of these systems is that yes, they do have many pieces to them, but it's not just about the number of pieces, it's about the relationship between those pieces and how they're interconnected with each other. One of the defining features of a complex system is that it can't just be understood just by adding up the individual parts. You need to understand how those parts fit together with each other. A car is only a car if the pieces are assembled in a very specific way. To construct a car, you can't just throw together four tires and a steering wheel and all the pieces to the engine and call it a car. A car's function depends on how those parts fit together with each other. Systems are all about structure. That's one of the principle ideas of system dynamics. A system's behavior is caused by it's structure, not the individual parts that go into it. So so far we've talked about a couple of the features of complex systems, goals and interconnected parts. Adaptations are another key feature including how the immune system adapts when a person is sick. Other key features are being dynamic, as systems are constantly changing and evolving over time, are complex systems. They also include feedback groups which we'll be discussing a lot. And they often have non-linear patterns over time. And as I've already emphasized, their behavior depends largely on their structure. Now, listing out these characteristics, you can start to see why complex systems modeling is particularly relevant to obesity. Because a lot of these characteristics are things that we see all the time in obesity research. Weight loss is difficult to sustain, for example, due to metabolic adaptations that occur within the human body. And determinants of obesity are very difficult to study because our food supply is very dynamic. It's constantly changing. The foods that we're surrounded by today are very different from the foods that people were surrounded by 50 years ago. And the trends in obesity overtime have been highly non-linear. The prevalence in obesity was pretty stable for many years in the 1960s and 70s, but then it increased rapidly from 1980 to 2000. And now it's been stable again, relatively, but at a higher level. And non-linear trends such as this are very common in a complex systems. Now, going back to the key characteristics, I want to emphasize this idea of structure again, because that's really what the term endogenous is all about. So the term endogenous literally means arising from within. And it represents that idea that changes occur because of a systems structure, not because of any specific part. In an endogenous system, there's a huge emphasis on feedback loops. Feedback loops represent reciprocal relationships between variables. Basically, they represent the idea that causality is not always a one-way street. And this is a vastly different way of thinking from how we often talk about causality in obesity research and most public health research. I was trained as an epidemiologist, and in epidemiology we often talk about how a single exposure affects a single outcome. So let's say you might be interested in the effect that neighborhood walkability has on physical activity. You want to estimate that single effect of that one exposure on that one outcome. But we also worry about confounders in epidemiology, because a confounder can bias that single effect that we're interested in. So household income, for example, that might bias the effect that walkability has on physical activity. So epidemiology we talk about controlling for confounders such as household income. Because we want to estimate that single effect that that one exposure has on that one outcome. But causality is not always a one-way street. That's not how complex systems work. And it's very different from the approach that we take in system dynamics. In complex systems, causality goes both ways. The environment can affect our behavior, but our behavior can also affect the environment. because area goes in both directions. And when I'm explaining this idea I like to use bike share programs as an example. I love bike share programs, I use them all the time. For those who don't know bike share programs they allow people to rent bicycles on a short term basis from stations throughout a city. Bike shares, they've been very popular around the world for many years, but they've really exploded in the United States lately. I used to live in Chicago, where bike shares grew very popular very quickly. Chicago's bike share program, I think it literally started the day that I moved out of the city in 2013. But within a year or two, when I returned to Chicago, it was like bike shares had taken over the city. I mean, you look around and there's just stations everywhere. So this happened very quickly, which again is something you often see in complex systems, and I definitely notice more bikers on the road because of this environmental change. But what's interesting to me is how the environment has also changed in response to it. The cities adapted by installing more bike lanes to make Chicago more bike friendly, at least from my perspective. And that's what you often see in system dynamics. The bike shares were an environmental change that triggered the behavioral change, but then the behavioral change led to more environmental change. And that's what feedback is. So causality goes in both directions. And endogeneity represents the idea that change is driven by feedback such as this. Not by the environment alone or behavior alone. So feedback loops are one thing that distinguish complex system modeling from most public health models. Another distinguishing feature is the emphasis on side effects. Side effects represent the idea that changes in the system often lead to unintended consequences. "Side effects," is in quotes, because John Sterman, who's one of the leading figures in system dynamics, he would say that there are no side effects. There are just effects. But he also makes it a point to emphasize that complex systems do things that you might not expect. Promoting walkability, for example, it might achieve its goal to increase physical activity. But it might do other things, like drive up property values. And this is a problem that you see in cities a lot, when improving walkability or public transit. It makes the neighborhood a more appealing place to live in, and as a consequence wealthier populations move in, lower income populations are forced to move out. And that's another example of a system that's dynamic. Neighborhoods. Because neighborhoods are always changing, and shifting, as people move in and out. And when you study complex systems, such as neighborhoods, you often observe unintended consequences such as this. And these patterns are often counter intuitive, too. For example, a person can exercise more to burn more calories, and yet still gain more weight based on how they adapt for this increase in exercise. And counterintuitive results like this are very common in obesity research and they're common in complex systems. And that's why I also like this quote from Jay Forrester, who's the founder of system dynamics. Forrester said, it has become clear that complex systems are counterintuitive. Systems often produce patterns that you wouldn't expect based on individual parts. Like based on exercise alone. I also like this quote from John Miller and Scott Page, who say, the field of complex systems challenges the notion that by perfectly understanding the behavior of each component part of a system, that you'll understand the system as a whole. You can not understand a complex system by examining each individual piece and adding it all up. Just like you cannot understand obesity by looking at carbohydrate intake, and then looking at physical activity, and then looking at sedentary behavior, and just putting it all together. To understand obesity, you need to understand the structures that underly it. And because results are counterintuitive, they are hard to predict. That's what makes them counterintuitive. Especially because we're trained and to think in a very linear style of one exposure affecting one outcome. Just as humans, we're not particularly good at thinking about how feedback loops affect your outcome. But that's where computers become useful for modeling complex systems. And that's where we get into system dynamics. System dynamics is a computer simulation approach to analyzing policies. It's useful when simulating the impact that policies or other interventions have when you cannot study them with a traditional controlled experiment. Many policies cannot be randomized for practical or ethical reasons. So system dynamics uses simulations to get a sense of the impact that policies would have if they were to be implemented. System dynamics models are also useful for understanding why many interventions lead to counterintuitive results like we've talked about. And in the next section I'll start to discuss how you build a system dynamics model to explore some of these questions.
