Abson Causal Loop Diagrams

Good morning, or maybe not morning. I have no idea when you're watching this. Good late in the evening. Um, I'm Dave Abson, and I'm going to talk to you today about a particular method in the environmental sciences called causal loop modeling.

Before I talk in detail about the method itself, I want to talk about some of the kind of theoretical concepts related to kind of causal loop modeling, which are pretty important for you to understand where and when and why you might apply this particular concept. So, with that, let's begin.

Three learning outcomes, essentially, from this lecture, which I hope you go away with: One is to understand the importance of modeling in the environmental sciences, and I'll talk more about that in a minute. And then to have a basic understanding about the principles of causal loop diagrams, or causal loop modeling, and its relation to this notion of systems thinking, and how systems thinking is, um, slightly different from other approaches in science. And then thirdly, to gain some knowledge of the application of causal loop models, particularly to social-ecological systems, and I'll give a couple of examples, one of which I was involved with, about how you can apply this method in the environmental sciences.

So, before I talk anything about causal loop models, I want to talk about models more generally. And basically, we can think in science, so putting aside the kind of more classical understanding of a model, and using a reference from 2001, so probably none of you will get that joke. We can think about three basic sorts of models related to science, or in the world more generally.

So, the first is what you might call a mental model. So, mental models are essentially these deeply held images of thinking, acting, which inform the way we understand the world. So, we don't experience the world, or we don't kind of conceptualize the world directly, but rather we create these basic mental models about how the world works. And there are these kind of underpinning assumptions about, about the kind of systems in the world. So, we can think, for example, about a mental model being this notion that technology equals progress. So, these are sort of the underpinning thought preset processes on which other models are built.

And then, as we go more into the kind of scientific modeling, we can think about heuristic models. So, heuristic models are essentially a kind of almost like an advanced version of a mental model, where you're trying to provide a codified, shared understanding about how things work. These aren't models which kind of predict the future. They're not statistical models. They're not kind of formalized models in the scientific sense, but they're, um, ways of trying to understand complex reality.

So, one example which I have here, so you can see this figure, which maybe you haven't seen yet, but you're likely to see a fair amount over the next few years, is this notion of ecological resilience. And it's demonstrated from this heuristic model of the ball and cup. So, here we have these cups or basins, and here we have balls. And basically, the idea of ecological resilience is that systems have semi-stable states, but you can shock those systems until they change that system to such an extent that it flows over a threshold and changes into a new system state. So, for them, in other words, uh, you can have a forest, and then there's a perturbation to that forest, like a fire starts, and the forest burns, and it becomes open land, changes to a new system with new identity and new functions.

So, these heuristic models are super important in science in trying to understand simply the kind of way in which complex reality plays out when different components interact with each other.

And then we can think about formal predictive scientific scientific models. So, you can think about statistical models which are attempting to understand how a treatment affects, um, a particular kind of organism in the wild, for example, or kind of experimental models which you're trying to kind of predict how changes in systems occur. Or you can think about things like general circulation models in climate science, which are, um, attempts to actually predict what might happen in the future of increased carbon emissions or greenhouse gas emissions in the atmosphere.

So, all three of these types of models are particularly important in science. And what's key to me is that we need, basically, to have all three of these things. We need to have mental models, these kind of deeply held images of thinking, acting, how the world is. Then we need to have these more formalized heuristic models. And then we need to have these potentially formal predictive scientific models. Um, and what's interesting, I think, about causal loop modeling is you can use causal loop modeling to try and capture all of these three types of models in a kind of participatory, transdisciplinary way. So, you can get people to talk about their mental models, you can create new heuristic models, or you can even go as far as using causal loop modeling to provide kind of formal predictive scientific models.

And I guess you want to say something about this notion of, um, models. So, the way, the way I think about science is all science, at its core, is the kind of art of modeling. It's taking complex, uh, reality and simplifying it into codified models of the world. And there's a few things we need to think about when we think about what a good model is in, kind of, environmental science or sustainability science.

So, the first thing we need to be aware of, and this is a quote from Alfred Korzybski, who was a geographer: "The map is not the territory." So, whenever you're creating, um, a model, you're simplifying complex reality, and you have to be aware of the simplifications you make, and make sure that those simplifications are such that they're capturing the important aspects of reality that you're trying to understand.

Then there's a quote from George Box, who was a statistician: "All models are wrong, some models are useful." And this is basically the idea that when you abstract from complex reality, you have to leave things out. And the key thing is to make sure that you're leaving in the aspects of the model which are really important for you to understand what's happening. Now, come to that, come back to that notion a bit later when we talk about the difference between, uh, kind of causal loop modeling and some more traditional approaches in science.

And then the last, uh, kind of quote that I like about models comes from Albert Einstein: "Everything should be made as simple as possible, but no simpler." So, this is the idea that we shouldn't oversimplify models. They need to capture the things that we care about in order to understand the world. Um, so that's just a basic overview of how I think about models and the importance of models in science.

And now I just want to kind of relate that to this notion of sustainability. So, sustainability is a complex problem. It, uh, it covers lots of different facets of, uh, the world: so, economic aspects, ecological aspects, political aspects, aspects of culture. It occurs at multiple different scales. So, we can think about local, regional, national, global sustainability. We can think about inner and outer sustainability, so this kind of inner worlds and outer worlds. We can also think about where we're trying to understand sustainability. So, this is a silhouette of a city. Cities themselves are these, uh, highly complex systems of interacting components. You have people in them, you have infrastructure, you have buildings. All of these individual things can be sustainable or unsustainable, and they can interact with each other in ways which are sustainable or unsustainable. So, when we're trying to model sustainability, we have a kind of fundamental challenge of dealing with all these different sorts of complexity. And it's not that you just have ecological or, um, economic or interaction between different components within the system. We have these also occurring at all of these different scales, for example.

So, then, one of the key questions in science is, what kind of methods do we have for capturing some of this complexity, or creating useful models out of this complexity? And so, I would basically make the argument that science has two broad approaches to modeling such complexity. The first, I would, uh, kind of say, a jigsaw piece approach. And the second, I would term a jigsaw puzzle approach. And I'll talk a little bit more about each of those. Um, but essentially, when I say science taking a jigsaw piece approach, then you can kind of think about the complexity of the world as being a jigsaw puzzle, right? And one of the ways you can try and understand what's happening, um, in that complexity is to split the world up into lots of different component parts and try and understand those component parts in more and more detail. And basically, the idea is that you can understand the whole by splitting it into its component parts and really understanding how those component parts function. This is often referred to as reductionism.

And in science, we can think about methodological reductionism, which is essentially the attempt to explain the world in terms of smaller and smaller entities. So, we understand how atoms work, and then we understand how molecules work, and then we understand, uh, how chemical bonds work. And we can move up by understanding each of those lower things to understand how the system works as a whole. And there's this nice example, um, from the 19th century, uh, I think his name was, um, Jacques de Vaucanson, and this was basically the idea that you could take an organism like a duck and understand all of the component parts, so how its mouth works, how its guts work, how its legs work, and kind of mechanically recreate each part. And by doing so, essentially understand a duck. And this has traditionally been the way that modern science has, uh, has approached understanding complexity. You reduce the system to manageable parts, and you understand those individual jigsaw pieces, if you like. And this has been an extremely successful approach to understanding science, but it's not without its limits.

So, for example, we know that you cannot, um, understand a duck just as, kind of, these mechanical component parts which fit together. And, in fact, um, the, the automatic duck which was created didn't really work. So, they, they cheated to make it seem more like a living duck than it actually was. Um, and the other thing I would say about reductionism as an approach to science is that also tends to not just be about, um, taking small parts of bigger systems and understanding them and hoping that by understanding the small parts, all the small parts, you can then understand the whole, but it also tends to take this kind of snapshot approach. So, it tends to take, um, the world at one point in time and try and understand what's going on.

So, there's a few examples that I just have here. So, this is a snapshot image of carbon emissions per capita from different countries in the world, for example. Here's a snapshot image at one point in time about the relationship between different, uh, insect abundance and their occurrence in relation to the distance from the city center. And this is a land use map. So, all of this is providing useful information about complex reality, but at one particular point in time.

And then you can see three other figures here. So, this shows, uh, the pollinator loss in Britain since 1850. So, this seems like it's no longer a snapshot, right? Because it's showing changes, trends over time. This is greenhouse gas emissions. And this is the ecological footprint and biocapacity, uh, globally, maybe, but I'm not sure about that. And I would say this is actually still a kind of snapshot approach to understanding the world, in the fact that you're just measuring, um, these things you're interested in a single point in time, and then just stacking those points of time up to give you a trend. It's not really telling you anything about the dynamics which exist in the system. So, it's not telling you what's driving those changes, just providing a series of snapshots of how those changes occurred.

I don't want to dismiss the importance of this approach to science, this kind of snapshot, reductionist approach to science. Much of the scientific knowledge we have is based on that, but it's not the only way of thinking about or modeling the world. And what I'm going to focus on for the rest of this lecture is this notion of systems thinking as an alternative to reductionist science, and then causal loop modeling as a kind of means of, kind of, or method for doing some forms of systems thinking.

So, this is a second broad approach to science that we can think about: jigsaw puzzles. Um, and again, I would argue that this isn't necessarily an alternative to, uh, reductionist science, but rather a complement to it. And so, I'm going to start just by explaining what I mean by systems thinking. But broadly, we can think about this rather than thinking about the individual pieces of a jigsaw, but thinking about how the different pieces fit together, connect, influence each other. So, it's a, it's a more holistic approach to understanding complexity, not for looking at detail about the individual components, but looking particularly about how they interact and fit together.

And so, deeply tied to this notion of systems thinking is also system dynamics. So, it's not just about how things connect together, but how things change over time. So, how are changing one of those jigsaw pieces? And now the analogy becomes a bit lost because jigsaw pieces don't really change, but how changes in one component in their complex system influence other components in the system, and how that feeds back and makes, kind of, the system as a whole change.

So, I'd argue that system thinking is an important idea in sustainability science. Um, you can think about it as an analytical method. More fundamentally, I would say it's a way of seeing the world. So, rather than seeing the world as a, as a thing which can be split up into its component parts and understood, it's seeing the world as a thing that can only be understood by taking a step back and looking at the system as a whole. And I'd argue it's particularly useful for modeling system dynamics, or system changes.

And now, the way I'm going to talk about systems thinking here sort of implies that there's just one type of system thinking, but actually there are many different kind of views or perspectives, kind of, approaches to systems thinking, and which come from slightly different disciplines. So, I'll talk about one example, the distinction between hard and soft systems thinking. But you can see here some books, all of which take slightly different ways of understanding systems and come from slightly different, uh, kind of academic disciplines. So, systems thinking has kind of emerged from a kind of organization theory and, kind of, business theory, so what firms are systems, uh, kind of organizations of systems, has emerged from ecology as a kind of fundamental understanding of systems, and also from, kind of, other disciplines, almost as a discipline itself. So, systems thinking by people like, um, Jay Forrester would be one example.

So, one of the fundamental arguments about, um, systems thinking is that "the behavioral system cannot be known just by, uh, knowing the elements of which the system is made." This is a quote by, one of, I think, the most important system thinkers in the 20th century, um, Donella Meadows. And the analogy that you'll often see is that you have six blind people in a room with an elephant, and each is kind of trying to understand a small part of that elephant. So, one grabs the trunk and thinks it's a snake. One grabs the ears and thinks it's a rug. One holds the tusks and thinks it's a spear. Each of those, individual understandings don't lead up to an understanding of the system as a whole. This is a slightly, um, ableist, uh, example, I guess. In fact, I'm pretty certain that blind people could figure out that that's an elephant, but the idea still holds that by only looking at the individual pieces, you don't necessarily understand the system as a whole, because there are kind of emergent properties of systems which can't be understood just by breaking into its small parts.

So, largely in systems thinking, then, you can think about a system consisting of three things: system has elements, these are components that make up the system, you can think of these as subsystems. Then you can have the kind of interrelations between those elements. And then you can have, then you have the function of the system, so what the system does. So, you can have an, a classical example is a faucet or a tap, and a bathtub, and a drain. And these are three elements. Uh, water is the kind of thing which, uh, interrelates. The insulation between those things is the flow of water. And there's a function. So, you turn the tap on, it provides water into the bath. When the bath gets too full, you can open the drain, and their water will come away. So, this is a very simple kind of systems model of, uh, that particular system. So, you have an inflow of water controlled by a tap. The level in the water in the bath is another element. And then you have an outflow through the drain.

And when you're trying to think about a system, there's a couple of questions you can ask yourself: Can you identify the parts? Do the parts affect each other? Do the parts together produce an effect that is different from each individual part on its own? And does that kind of affect your behavior over time persist in a variety of circumstances? And you can use that to identify a system effectively.

And there's many different kind of types of systems we can think about. So, you can think about, kind of, ecological systems, ecosystems. This is a governance system. So, this happens to be a representation of the system of governance in the United States. You can think about a city as a system. You can think about a university as a system. You can think about the world as a system. So, there's multiple scales and multiple different ways of kind of identifying what systems are.

For me, one of the really important distinctions in systems thinking, and this comes from Peter Checkland, his 1999 book is a good place to start if you're interested in this distinction, is between hard systems thinking and soft systems thinking. So, as I said before, um, systems thinking is a way of, of, kind of, understanding the world. So, it's a way of thinking, if you like. And there's two broad approaches.

Hard systems thinking essentially starts from the perspective from this mental model, if you like, that a system is, systems are real-world tangible things, things that can be engineered. So, in other words, systems ontologies. So, ontology, essentially, this idea of how the world is. The world is made up of systems. They're real things which you can understand, grasp, and, uh, kind of, um, they are the things which make up the world.

And then there's a soft systems thinking approach, which essentially says the world is, uh, just, uh, complex, and systems is an approach for exploring that complexity. So, this sees systems as epistemologies. So, kind of, systems thinking, essentially, as a means of, um, understanding complexity. So, systems aren't necessarily real tangible things, but then heuristic models that we create to try and understand that complexity.

And the important distinction between those two is that from a soft system perspective, uh, there is no such thing as a single system. That where you draw a boundary around a system is always determined by, uh, the observer, by the, by the scientist who's trying to understand that complexity. So, for example, if we go back to this, uh, figure here, someone has looked at the world and said, okay, we can think of ecosystems as systems. They're flows of materials and energies within natural systems, and that's a particular system. But of course, humans interact with those systems. So, here we have a kind of social system of governance, and we've drawn a boundary around that. So, there's no environment in this system, and there's no people in this system. Of course, these two things can be combined, and you can think about a social-ecological system, which combines the interaction between the kind of ecological components and the, and their, kind of, anthropogenic or human components in the system.

And so, the key point about soft systems thinking is, um, how you bound your system, how you think about, uh, what's in your elements in your system, what aren't in your system, fundamentally changes the way you understand the world. So, I would generally consider myself to be more like a soft system thinker. And so, it's important to note that a system is bounded and defined by the subjective interests and, and, kind of, pre-analytical assumptions of the researchers. So, part of the reason the ecosystems were described without humans in them, largely when they were developed, because they were developed by ecologists, they were developed by people who are interested in natural systems and potentially less interested in human systems.

And so, this comes from the work of Rayson, but other people also talk about this, this notion of systems of interest. So, rather than think about an ecological system or social system or social-ecological system, it's useful to think about, this is my system of interest. And it acknowledges that the system, at least in part, are defined by the world views and concerns of the researchers. So, how you bound your system, it depends on what you're interested in. If you're interested in politics and governance, then you bound your system based on social rules. If you're interested in material energy flows, maybe you bound your system based on, kind of, ecological understandings.

And why that matters is because your initial understanding, your initial mental model, the world fundamentally changes the way you then understand the world. So, this is a nice figure of a rabbit. Now, the moment I've said that, probably half the people who are watching this video are going to say, "That's not a rabbit, it's a duck." And the moment I say that, half of the people who previously thought of rabbit can look again and say, "Well, actually, oh yeah, I can see that there's a duck there." And fundamentally, the way that you perceive this image changes your understanding of the world. And it, so, is the same with systems of interest.

And there's another example here. So, um, again, look at this image on the right. Now, do you see a woman, young woman with her head turned away from you, or do you see an old woman, um, in a big fur coat? Both of those things exist. So, you, you can see both the woman and the young woman and the old woman. But you often, depending on how you initially look at that image, or that system, shapes how you see it in the future. So, we have to be clear about what our underpinning assumptions are when we're doing systems thinking, because that fundamentally changes the way that we see and understand the world. Are we looking at rabbits or ducks? Young or old women?

And then one more distinction I want to make is between complex systems and complicated systems. So, in complicated systems, everything is tightly connected. It's largely predictable, and it's largely controllable. So, you can think about the mechanisms in a, kind of, non-digital watch, clockwork watch, as being complicated. There's lots of different parts which interact with each other, but you know exactly what they do, and they do the same thing all the time.

And then we can think about complex systems. So, these can have tight connections. So, some bits are tightly connected with others, and some bits are loosely connected. But they also have, um, this characteristic of being self-organizing. So, they decide, or they shape their own interactions and, uh, interlinkages, and they're unpredictable. But they have an ability to adapt to change and function. So, largely in the environmental sciences, particularly when we go beyond environmental sciences, think about sustainability science, we're talking about complex systems, systems which have agency, they're self-organizing, they're unpredictable, they're adaptable to change. And that's important to understand when we're thinking about systems too.

Okay, so I'm nearly finished with systems thinking. I will at some point get onto causal loop diagrams. So, one last thing about models. We can think about, uh, different types of models too. So, we can think about models of complexity, such as heuristic models of complexity. So, this first one here is sometimes called the iceberg model, and it's about understanding the relationship between events which happen, underlying patterns which drive those events, the structures which drive those patterns, and the mental models which allow us to understand those structures, patterns, and events.

Then we can have the kind of network models for understanding systems. So, how thing A is connected to thing B is connected to thing C. Then we can have, uh, dynamic models of system behavior. So, here you can think about things like, uh, predator-prey dynamics. So, you have a predator eats, uh, the prey. There's less prey, so the train numbers go down. That means the predator numbers go down. That leaves space for three numbers to go back up. That means the predator numbers can go up because there's more food for them to eat. So, you get this, [Music] and you can't see my finger now, you get this kind of, uh, sinusoidal relationship between predators and prey. So, this is a system dynamic model.

And I'm not going to talk about any of these first three, but I'm going to talk about this last one, which is called causal loop models, a useful way of understanding system dynamics. So, they're not just the networks of what's connected to what, but, um, how changes in one node in your system influences changes elsewhere. So, fundamentally, that's what the next 40 minutes will be focused on.

One last thing I just want to talk about before I do that is to make this distinction between stocks and flows, because it's often important in causal loop modeling. So, stocks are the kind of accumulation of material in or information that build up over time. So, going back to the bathtub example, the amount of water in the bathtub is a stock. And then there's a flow, which is the in and out flows of material, information, change, or energy that change those stocks. So, inflow of water, bath stock goes up. Outflow of water, bath stock goes down. And you can think about ecosystems as being something similar. So, there's a stock of energy or material in an ecosystem, which flows through, uh, through material exchanges, um, from trophic cascades and, uh, solar energy, and all this type of stuff. So, water, transpiration, evaporation, precipitation, all this type of stuff. Um, I'm not going to go into much more detail about this, but it's quite important when you're thinking about, um, causal loop diagrams. So, think about whether you're thinking about stocks and flows, and what the relationship between those two things are.

Okay, now, the, the last important thing to say about, um, systems thinking, I would say, is to make a distinction between lines and loops. Okay? What do I mean by that? So, two ways of seeing the world, two other ways of seeing the world are a world made up of linear causality, and a way, a world made up of causal loops.

So, linear causality basically says there's a causal chain of cause and effect. In other words, you do thing A, that leads to thing B, that leads to outcome C or D. Um, so, there's a direct linear relationship between what you do, what you do, and the outcome of those actions. I do thing A, that leads to thing B, outcome is thing C, and there's a root cause for all events, and the root cause is some action which occurs. I call you a name, you punch me in the nose, I have a broken nose. So, I'm the root cause of that event occurring.

Causal loop understanding is different in the sense that it doesn't see a linear causality, but rather it says it's the interaction between different things which happen in the system which lead to the kind of emergent behavior of that system. So, um, the root causes of that outcome that you're interested in isn't a single cause, but rather the system structures and feedbacks that define that system.

So, here we could think about the fact that, um, you're not paying attention to me in the class, that makes me angry, that leads me to call your name, that leads you to punching me in the nose. But this is not a, there isn't a single cause of that, because we could start by me being angry, calling you a name, that means you not to pay attention in class, and all these type of things. So, basically, it's the idea that to understand the behavior of a system, you need to understand these kind of, uh, dynamics in the system which drive that particular system property. So, there isn't a single root cause. The root cause is actually all the interactions in that system. So, it's a, it's a kind of a way of acknowledging complexity in a way which linear causality doesn't necessarily do.

I see now that that probably wasn't the best example. If I'd had more time, would have thought about a better one. But maybe when we, we go through some of these causal loops, we'll see some better examples of, uh, system behavior determined by interactions, not a single event. And again, I've already mentioned this slightly, but what I want to say about causal loop models is that they can be used to capture, uh, people's mental models. They can be used to create heuristic models, and they can be used in a more formal, predictive scientific model. So, you can quantify them in such a way that you can actually predict outcomes if something in that system changes.

So, now we actually get into the detail of causal loops. So, they have a strong focus on nodes as the state of an element or variable. So, something like the stock. And elements which link each other via feedbacks from changing states. And we can essentially think about, uh, two types of feedbacks: reinforcing or amplifying feedbacks, these lead to instability or expansion or contraction of the system, and balancing or damping feedbacks, which tend to produce self-correctional stabilization.

And I've got a very simple causal loop which I'm just going to briefly talk about here, which is, uh, how many people are occupying a, uh, restaurant, assuming we weren't in COVID. So, let's find a useful place to start. One of the things about causal loops is there's no, um, no predetermined starting point, because what determines the dynamics or the outcome or the behavior is a system as a whole. But let's start with new customers arriving at the restaurant. The customers arrive at the restaurant, and they see that their customers are occupying the restaurant. So, they're already people eating in the restaurant. So, you think, okay, it can't be a terrible restaurant. That increases the perceived quality of the restaurant and means that more people, more customers arrive. At some point, if there are lots of customers already at tables, then you might think, well, actually, now this is a bit of a pain because I have to wait half an hour before I get some food, and I'm hungry. So, that means that the length of the queue is increased, and that reduces the amount of new customers who are going to queue up to eat in that restaurant. And then there's another feedback loop between, if more people are eating there, there's less new additional people who can eat there, and so forth. So, that the behavior of this system, the amount of people who are eating it, is determined by, um, all these things which occur and the interactions between those things.

And here we can see the reinforcing feedback loop. So, having people at tables means that the perceived quality increases, which means you're likely to have more customers. But at some point, having too many people at tables, um, increases the length of the queue, and that decreases the amount of new customers who are likely to go there. So, this is a reinforcing feedback loop. And this is a balancing feedback loop, because, um, as less people arrive, the number of tables occupied decreases again, and that means the length of the queues goes down, and then more people are willing to queue up again and eat in the restaurant.

So, I guess to talk in a bit more detail about reinforcing feedback loops and balancing feedback loops. A few examples here. We can think about, uh, simplest one first, maybe sales of a product. As sales increase, you have more satisfied customers. Those more satisfied customers can tell other people about how good your product is, and that leads to more sales. So, you have this reinforcing feedback loop. The US arms race is a classic example, or the kind of Cold War arms race. So, the United States has weapons, that's seen as a threat to the Soviet Union, when it still existed. Um, that means the Soviet Union feels the need to build more weapons. That means the, the, the Soviet Union has more weapons, which was referred to the US. So, the US builds more weapons, and you have this continued reinforcing feedback loop.

Here's a more natural system, uh, feedback loop: Arctic sea ice. So, Arctic sea cover, it is decreasing. That increases the albedo, so the reflectance, um, essentially of ice or of the Arctic. So, now you have open water, which can absorb more sunlight, which leads to more energy being absorbed, which leads to more warming, which leads to less, um, ice cover, which leads to higher albedo, and so forth. So, you have this again, this reinforcing feedback loop.

So, when we're thinking about a reinforcing feedback loop, the way you can identify, uh, a reinforcing feedback loop is that it has either all pluses or an even number of pluses and minuses. Okay?

So, when, now we're starting to think about formally how this, how these system interactions and feedbacks are modeled. This is essentially a node here, the aggregate size of the local economy. And this is another node, the kind of maintenance and development of infrastructure in that local economy. And this is the feedback between those two things. So, um, a positive link means that, um, if this node, the stock of the local economy, is increasing, then the maintenance and development of infrastructure will also increase. So, there's a, a positive relationship between those two things. But similarly, if the aggregate local economy, so if your local economy is decreasing, so you have less economic activity, you also have less maintenance. In other words, they change in the same direction as each other. And similarly, if you have more infrastructure, that means that you can have a better local economy. So, there's also a positive relationship between this node and this node.

When you have a, a minus, so a negative relationship between two nodes, it means that if one increases, the other decreases, or vice versa. If this is decreasing, this is increasing. So, in other words, imagine that you have a society with lots of conflicts in it. So, the more conflicts there are, the less social capital you have. So, this goes up, this goes down. As social capital goes down, level of education goes down. So, that's a positive relationship, they change in the same direction. Um, and as the level of education goes, um, down, the amount of, um, poverty goes down. Let me see if that's right. Education, quality of education goes down, amount of poverty goes down. So, I actually think this is a, not a particularly good example. I might have actually done picked the wrong one here. And the amount of poverty goes up, then the amount of conflicts goes up. Um, the quality of education goes up, then the amount of poverty goes down. If that's right. So, in other words, this increases, this decreases. This decreases, uh, this, but this increases, this decreases, and this increases. So, these two things balance each other out in the end. That wasn't the best example. I'll try and find a better one, um, at some point.

Balancing feedback loops. So, um, here's a couple of examples. So, you have livestock on a field, and you have a herd size. As the herd size increases, the amount of available grass decreases because they're eating the grass. So, there's a negative relationship between herd size and grass. As the amount of grass decreases, production per head decreases. So, that's a positive relationship, or the amount of grass increases, the production increases. As amount of production per head increases, then the profits increase. As the profits increase, it encourages you to, um, have a larger herd. But overall, this one negative feedback, um, in this loop means that you have a balance. You can't keep increasing, um, the herd size, because at some point, having too many cows on the on the field means that you have less available grass, which means you have less production, which means you make less profit, and therefore it balances out your continued desire to increase the size of your herd.

And you can see even simpler example. I talked about this system dynamic before between number of predators and number of prey. So, the more prey you have, um, the more predators you get. But the more predators you have, the less prey you have, because they're eating the things that they're predating. And so, that's a balancing feedback loop between those two things. And there's other examples: increasing price and fewer customers. If you have more customers, that encourages you to increase the price because more people want your goods. But the more goods that, the more expensive your goods are, the less customers you're going to have. So, again, a balancing feedback loop.

To avoid balancing feedback loops tend to stabilize outcomes, and reinforcing feedbacks tend to amplify outcomes. So, uh, balancing feedback loops essentially have an odd number of minuses. They can be, uh, one minus, or three minuses, or five minuses, it doesn't really matter, but, and there's always an odd number. So, that, uh, the dynamics are affected by this negative relationship between, uh, an uneven number of components in your system.

And of course, you can have systems which have both, um, positive and negative feedbacks. So, more chickens you have, the more chickens that cross the road. This is balanced by the fact that then lots of them get run over, and therefore there's less chickens. The more chickens you have, the more eggs you have, and therefore the more chickens you have. So, reinforcing feedback loop and a balancing feedback loop. And just depending on the strength of those two dynamics, um, depends on the outcome, whether in the end you end up with more chickens or not.

Um, I just want to go back to this previous example, because there's something I, I maybe should point out here, that, um, these causal loop models are still making quite a lot of assumptions which aren't always true. So, for example, um, these relationships aren't always fixed. So, in other words, what actually affects production per head might not necessarily be the amount of available grass, but the available grass per cow. And if you've only got one cow on your field, adding one more cow, yes, it reduces the amount of available grass per capita, but if there's more than enough grass for, um, two cows, then it doesn't change your production per head. So, in other words, there's a threshold level at which that dynamic kicks in. You can keep increasing the herd size until the point that the available grass per cow is less than required to maximize your beef production, for example. So, one of the things which isn't always clear in causal loop diagrams is where thresholds are, and at which point some of these dynamics either come into existence or or change. It's just something I wanted to say.

Um, and so, going back to this simple example of the reinforcing feedback loop between how much you sell and word of mouth, meaning you're selling more stuff, you also have a balancing feedback loop. In fact, there's, uh, market saturation. So, for example, there are simply no more people available to buy your iPhones. And therefore, at some point, increases in here lead to increases in market saturation, which lead to a decrease in the amount of phones being sold.

So, give something briefly now on the conventions for drawing causal loop models. So, um, if you see this kind of R with a circle, that's identifying a positive feedback loop. A B is a balancing or negative feedback loop. Generally, you try and give names to each of these or numbers, and to show which direction the loop is directed, is traveling in. So, if it's going from this node to this node to this node, or this node to this node to this node. So, this is the direction of these arrows here. Arrows should only ever have one, um, head to them. So, the interaction goes one way. If you have two nodes which are connected to each other, so let's go back to a really simple example like here, then you don't put an arrow on both ends, but rather you show how this interacts with this, and then this interacts with this. So, you always have one arrow for one type of interaction.

Um, again, so a plus sign shows that if this node, this element of the system, is increasing, so is the one that it's interacting with, or if it's decreasing, then the other one is also decreasing. And the negative means that if one increases, the other decreases, and vice versa.

And then sometimes you'll see these, uh, two lines through one of the interactions or feedbacks, and that means that there's a time delay between a change in, um, this thing happening and, uh, the other thing happening. So, it might still be a positive or negative relationship, but they don't happen at exactly the same time. So, it might be that there's a substantial temporal delay between, um, one element in your system changing and the other one reacting to those changes. And that's quite important because you want to try and understand how rapidly those feedbacks in the system determine change in behavior.

So, causal loop diagrams can get quite complicated. This is an example about land use disputes and social cohesion in a social-ecological system in Polynesia, Solomon Islands, and it includes lots of things like how much money is going into the community, whether that means you're spending money on food or cash, how that affects community collaborations, how community collaborations then affect the amount of land in use, and all types of things like this. I won't go into detail on this. I'll show a detailed example which I'm more familiar with.

One of the key things I would say about causal loop models as an approach is they're largely based on developing these kind of more heuristic models about complex systems through expert elicitation. So, and those experts are often, uh, the people in, in the systems which you're trying to understand. So, for instance, in this example, a colleague of mine, Yuan Fazee, um, and, and his colleagues went to the Solomon Islands and essentially, in a kind of transdisciplinary approach, sat down with people in the island and said, "Okay, what's happening in your system?" So, um, what are the key elements of your system, and how do things interact with each other? So, in other words, one of the, the nice things about causal loop modeling is that it's something you can do with people. It's not just a set of scientists saying, "Okay, how do I understand, uh, this system?" but it's asking, uh, people in, in these complex systems, "How do you understand that system? How do you think about things interrelate with each other?" and building together these kind of heuristic models of, kind of, complex interactions.

Um, yeah, so maybe just showing, uh, a bit of, uh, some of those interactions, or a few things here. Let's just take a, a point here. You've got stress in, well, let's start here. You've got pressure on the ecological system, and there's a positive, and there's a, as the ecological system becomes, um, degraded, then yields from food in your gardens decrease. As yield from the gardens decrease, that decreases the amount of food available. As amount of food available decreases, stress in the community increases. And so, you have, kind of, a reinforcing feedback loop here. Or here, where's that loop finish? You have to look for the lines, goes all the way around here. So, in other words, you have to trace it through this way. So, as stress in the community increases, desire to make more money increases. Um, desire to make more money increases the amount of land under cash crops, and that increases their pressure on social-ecological systems.

So, one of the reasons for putting these, uh, kind of directions of the arrow is so you can see where that particular dynamic plays out in this system. One of the nice things about causal loop models is that it allows you to trace through these different dynamics and how they interact with each other. Because of course, there are other things which are affecting stress in the community. So, there's another feedback loop here, and there's, uh, elements within that feedback loop which are affected by things elsewhere in the system. So, you can start to, start to, kind of, unpick some of the complexity of, kind of, what leads to the emergent behavior in that system. And you can focus on which bits of, which elements of that system you're particularly interested in. So, for instance, if you're more interested in understanding how stress develops in these societies, then you can focus on this node, on this element. If you're interested in ecological pressure, you can focus under here. If you're interested in money flows, you can

focus over here so it allows you to understand more holistically how the different parts of the system fit together, but you can still focus on the jigsaw piece that you're particularly interested in.

This is another example which I won't talk about in detail, but you can kind of again see how complex some of these models can be. This is trying to understand stability in Afghanistan based on things like institutional interventions, political capacities, uh, kind of, uh, drugs, uh, popular support for, um, different factions in society, and then the infrastructure and kind of size of the economy, tribal governance, all types of things you can try and build into your models of how the world is.

And then you can also create multi-scale models. So this is going back to kind of sustainability as a multi-scale concept. So you can create your causal loop model largely based in a particular landscape in the similar way that you and Phase did here, but you can also then look at how dynamics from outside of that local scale affect it. So you can see how, kind of, regional droughts affect, um, behavior at the local scale, and you can see how climate change, which is caused on a global scale, affects local drought. So you can, you can layer these, um, these models. In fact, in practice, what you don't really do is necessarily create a new layer, but you just, um, add new causal loop elements out here which are at these broader scales. And again, you can zoom in and out of these to look at the kind of interactions between what's happening on the global, regional, local, uh, scales.

So just to reiterate a few things here. So development of causal loops is generally a qualitative, participatory approach led by experts, where those experts are just people who know the system. And often it's useful to get many different people. So if you want to understand what's happening in the Lunar Border Hider, for example, then you probably want to talk to tourists who use that particular land. You want to talk to local government, government people, you want to talk to farmers, you want to talk to business owners, you want to talk to some ecologists who understand the ecology, maybe some economists who understand what's happening economically in that system. So all of them provide their own insights about particular kind of loops within this bigger system, and you can bring those things together.

I'd argue that it's most useful for creating heuristic models, but useful for problem-solving, learning, and discovery. That's essentially the purpose of a heuristic model, and they employ kind of, um, practical methods. So they don't guarantee, um, the optimal or perfect outcome, but enough to kind of gain, gain sufficient understanding of the system. So they're not going to provide a perfect model of that system, but they are going to allow you to understand better than the system you did before. And what's, I think, particularly nice about them is that if you build these causal loops with people from government and people, farmers, and ecologists, is you can start to understand how other people see the same system in different ways and see what the linkages is between your understanding and their understanding. So you can build shared understanding of these systems.

And this comes back to some of the kind of transdisciplinary, uh, kind of characteristics of causal loop modeling, is that it can help provide shared understanding of common problems. So a farmer in the Lunar Border Hider probably has a very different way of understanding what problems are with managing that system than an ecologist. An ecologist different from someone who is in government or someone who's, uh, interested in the economy. And actually, you can start by building these causal loop models to see how all these things interact with each other.

Okay, um, a few examples. I'm not gonna go for this example in, in detail. Maybe if you come for some of my classes, uh, you'll play this one. So this is an example of a playable system dynamic model based on causal, uh, loop modeling. And I realized that, um, I've made this picture too big. So it says "Common Poo Resource Systems," and that should become "Poor Resource Systems." Um, in this, this is basically a simulation where you act as, um, fishing companies in the open ocean, um, system, and look at the tragedy of the commons. So you can, and this is, uh, kind of a quantified causal loop model in that you can kind of see how changes in things like net recruitment of fish affected by how many fish you catch in the ocean, that influences how much effort you put into catching, and you can see all these dynamics. And you can use this particular playable system dynamic model based on causal loops to see what happens when you're trying to manage open access resources, essentially. And you can run these models and see how they compare to the real world open ocean fisheries, for example. It's a very nice talk.

This is just a simple example of some of the system that system dynamics and causal loops used in this model. So here you have fish stock, which is the main thing you're interested in. And as the fish stock increases, so you can think of this as a slider here, then, um, there's more stock, more fish to make new baby fish. So you get a positive reinforcing blue kit loop here: more fish have more babies, have more fish, have more babies, reinforcing loop. Um, as the fish stock increases, the fish density increases, the amount of fish per area in the ocean increases. As the fish in the ocean increases, the fractional net recruitment, the amount of new fish, um, decreases because there's now, um, competition between those fish for the food that they need to eat. And therefore, um, as the net fractional, um, recruitment decreases, the total net recruitment decreases. So there's a positive relationship. So this is one positive, two positive, one negative, odd number of negative feedbacks, um, and therefore a balancing reinforcing loop here. So in other words, as you get more fish in the ocean, you get more fish until at some point, they're competing for resources, and then you have a balancing feedback loop. So you can't, then the amount of fish in the ocean doesn't expand forever.

Then you have the more, um, um, social side of this diagram. So as the fish density increases, the amount of fish you catch also increases because there's more fish per unit area. That encourages you to, um, cat, get more ships. As you catch more ships, so, um, that means that you're pushing the fish stock back because you're effectively, um, catching, uh, the fish which are in the ocean. So then the fish stock goes down, the fish density goes down, the number of fish caught goes down, the catch goes down, and again, it's a balancing feedback loop.

And then you can bring in all these other parts of the model. So think about revenue, what it means for fleet investment, technology, pressures to innovate. You can think about how this affects the broader market prices, how this affects, um, consumption and demand for fish, and so forth. So you can, you can keep expanding these models. Again, this comes to this notion of the system of interest. So you're only interested in, um, understanding and the relationship between a number of fleets you have and the number of fish stocks you have, but you might be interested in, uh, some of the more, uh, nuanced understanding of how technology affects these relationships, or how this affects the broader demand for fish in the marketplace, and how that, in the end, affects, uh, population growth, for example.

Another example looking at, essentially, the conflicts between nature conservation in New South Wales in Australia and cattle farming. And I've already spent a lot of time, so I'm actually gonna pretty much skip through this one, um, but you can see here, um, a model which is looking at the system dynamics for economic growth. So let's start somewhere. Um, you have an irrigation industry. So wetlands are being drained, and you're having irrigation in your farmland. As that increases, the amount of products from irrigated farming increases, profits from irrigation increase. This contributes to economic growth. This contribution to perception of prosperity in the community, that, um, contributes to a belief in the advantages of economic growth, uh, and those tend to lead to short-term economic returns being the focus, desire for increasing your agricultural inputs, making high profits, commitment to irrigation, therefore more irrigation. So there's a kind of reinforcing feedback loop here. And, um, but this exists within, uh, these other feedback loops which are kind of also, uh, reinforcing these things. So basically, in this model, what Uran was trying to do was, uh, understand why there's increasing pressure on natural resources in this system, because you have all these kind of reinforcing feedback loops which are pushing the system towards increasing, kind of, extraction of water from wetlands to use in farming systems.

And this, uh, next example is one which I was actually, um, involved with, and this is based in, um, Southern Transylvania in Romania. So this is an extremely important cultural landscape, consisting largely of small-scale farming and forest patches. Has some of the highest levels of biodiversity anywhere in Europe, and it has large numbers of bears, wolves, lynx, and very high diversity, plant diversity, butterfly diversity, and many other things, largely as a result of it being this mosaic of small-scale, low-impact, low-external-input farming system which has kind of evolved over the last five or six hundred years. But the system is threatened by land abandonment and land grabbing. So in other words, you have small-scale farming which isn't very profitable. So often what happens is people move to the cities and don't farm anymore. That land becomes abandoned, and this kind of complex mosaic of land uses disappears, which means you lose some of the ecological functions, and then you lose a lot of the biodiversity in that system. But you also lose the kind of, uh, culture of that system. Or the alternative is people abandon the land, um, rather than people abandoning the land, they sell it to large farmers, and it becomes more like the kind of agricultural landscapes that traditionally or more normally you see in other places in Europe. If you look at agriculture in Lower Saxony, so very large, uniform monoculture farming. And this is potentially, uh, problematic. So there's a threat to this valued system, culturally valued, socially valued, ecologically valued system, but also conflicting visions of the future.

So in this study area, we used causal loop modeling to try and understand what the dynamics playing out are in terms of things like land abandonment, land grabbing. And we did this through a participatory process, so workshops of 17 different organizations and key individuals. So again, trying to capture different perspectives on what's happening in this system from nature conservation, forestry, agriculture, tourism, uh, people of churches, uh, teachers. And we asked them, essentially, to talk about the main changes in the past and the present, and what they see as the future, and what's driving those changes. We use that to develop with these communities system diagrams, so causal loop diagrams about the region with a key focus on drivers of change. And you can kind of see this is an example of one of these causal loops. You can do these models pretty simply, just with post-it notes and pencils and a big bit of paper. And we brought, um, we had multiple iterations with mortal groups of people looking about these, and we brought them together and tried to develop a systems understanding of, uh, kind of this system as a whole.

So we can start here at one node, one element in the system, which is a probability of small-scale farming. So as the probability of small-scale, um, farming increases, the amount of land sold to foreign, uh, farmers decreases. In practice, what's happening in the system is the opposite. So, um, the probability of small-scale farming is decreasing, so the amount of land sold for farming is increasing. As the amount of land sold to foreign farmers increases, the amount of intensification of farming increases. Oops. As, as the intensity of farming increases, farmer and biodiversity decreases. But also, as the probability of small-scale farming decreases, the amount of abandoned farmland increases. The amount of abandoned farmland also decreases farmland biodiversity, profitability decreases, number of people leaving the village increases. So that's one set of, uh, feedbacks.

And then we can also think about how that affects the local economy. So that local economy weakens, this weakens infrastructure. This example I gave before, so there's a reinforcing feedback loop here. We can also think about how social capital, uh, so basically trust in the local economy, so social capital is decreasing, the level of education decreases, the amount of poverty increases, the amount of conflicts increases, the amount of social capital decreases. And so social capital decreases, number of people, uh, leaving the village increases, which increases the amount of abandonment and so forth. So you can start to see these, uh, type of things. A social capital decreases, there was in, uh, kind of increasing, uh, short-term profiteering. So people are much more interested in about just themselves in the short term, rather than the long-term functioning of this landscape. And that affects things like the amount of forest which is being grown, and that affects forest biodiversity. And they can also think about some kind of other impacts. So things like emigration of the Saxons. So this is an area which traditionally, traditionally or which had been settled by, um, German Saxon, um, immigrants five or six hundred years ago. So it had a mixture of, uh, kind of Romanians, Hungarian Saxons, and, uh, Roma. But after the fall of the Czech regime in Romania, many of those Saxons moved back to Germany, and that also had an effect on social capital. So you can start to see all of these different factors playing out in how this landscape changes.

And so we, um, use those causal loop models, and this is, there's actually more to this causal loop model than we've got here, but just to show you a bit of it, to develop scenarios about the future. So here, basically, based on two dimensions of potential external change outside of the system. So one in which on this axis, you have a kind of higher or lower social capital, and on this axis, governance, which is more focused on the economy or more focused on the environment. So you can see how these external factors play out in the system dynamics. And you can use that to think about what the outcomes in that system as a whole might be. So in other words, if you have a situation where you have increased social capital and pro-environmental, um, governance coming from the EU or the Romanian government, then you might end up with a kind of situation like this, where there's a maintenance of the current complex patches in the landscape, more tourism, uh, greater focus on maintaining cultural, um, kind of, uh, values in in this system. On the other hand, if you have low social capital and, uh, kind of governance based on the economy, you're likely to have more land grabbing and more simplification of this landscape. And you can see what the outcomes are in terms of social well-being and ecological, uh, conservation of biodiversity, for example.

So we then, in this example, we took these, uh, scenarios back to the local people to talk through again the system dynamics and what type of future they might want, and what might need a change in their system in order to, uh, lead to the type of system that they, that they would, um, ideally have. And you can also use those system dynamics to then to model out different things you might care about in the future. So here we were looking at these four potential scenarios and how it might mean on a spatial scale, so across all our study area. These are basically village catchment areas, each of those little pixels. Well, so what it might mean for intensification of farming, what it might mean for, um, land abandonment, what it meant for forest exploration, uh, tourism, local economy, social capital, immigration, impact of foreigners in this landscape. So you can see spatially across each of those different scenarios, these different things we care about, different nodes in this model, if you like, how they'll increase or decrease. So red means you've got an increase of intensification, um, blue means a decreased intensification, and so forth. And you can then play through these scenarios in different places, and they, the outcomes are different because different communities have, for example, um, different land use possibilities. So some areas are higher in the mountains and more slopey and less suitable for intensified farming. Some areas have different demographics, so more likely to be a land abandonment, and, and so forth. And so you can use this type of model not just to understand how the system is, but how it might change in the future, and how it might change spatially across, um, large areas. And you can take that back to the community who helped develop those maps and use them as, use these approaches as tools for thinking about what you might want to change in your system in order to lead to the desired outcome of that complex social-ecological system.

So, um, as imprecise and wishy-washy somehow as, uh, these models might be, I hope you can start to see how they have a particular use in, in kind of environmental science, sustainability science, which you don't necessarily get from doing this reductionist approach of just understanding what agricultural intensification might mean for, um, butterfly abundance in, in the field. You need that information as well, but this kind of allows you to feed in that detailed, kind of, jigsaw piece understanding into a broader system understanding of the dynamic interaction between all these different parts of these complex social-ecological systems.

So nearly done. Summing up. So causal loop models, causal methods, and potentially useful transdisciplinary components to more reductionist complements to more reductionist approaches to understanding and modeling complexity and kind of sustainability science. Um, I have a picture of, um, Donkey Shot by Pablo Picasso, partly because it's my favorite photo and favorite painting, but also because it's a nice example of, um, taking hugely complex reality, in this case, 1200 page long book or whatever it is, 600 page long page long book, and capturing the essence of of that book in a few hundred, uh, strokes of paint. And in the sense, what causal loop modeling is also trying to do is trying to find that essential, kind of, characteristics, node, system elements, interactions between systems, which determine the kind of behavior of that system as a whole, are the outcomes of those systems as a whole. So it's a, it's a way of understanding complexity in a relatively simple way, which you can explain to people from outside of, uh, kind of the, the kind of formal scientific modeling, uh, kind of scholarship. So you don't need to understand complex statistics in order to understand these causal loop models. You don't need to, kind of, have detailed understanding about any of the individual relationships. So you don't really need to understand exactly why increased nitrogen in the atmosphere leads to decreases in, um, butterflies in the landscape, but you can just see how more broadly that relationship relates to all the other things in the system which determine why you might have more nitrogen in the atmosphere, for example.

So I think one of the things I was asked to do, and this was actually last year, so I haven't checked if I was supposed to do this this year, but I've done it anyway. Homework. So if you want to try, um, to see how causal loop models work in practice, then here's an example you could think about from your own life. So try and create a causal loop model to explore the system dynamics related to whether or not you do the readings which are given at the end of the lecture. So I think at the end of each of these lectures in this course, Henry gives you, all the other lectures give you a reading, and you're supposed to read it. I know that not everyone reads it. So try and figure out what might be driving decisions to read or not read those readings. Then you might think about things like the amount of time available to students, the attitudes of other students, the attitude of the lecturer. So if you enjoyed the course, maybe there's a positive relationship between how good the lecture was and how likely you are to do the reading, or maybe there's a negative relationship between those two things. If the lecture was really good, you might feel you don't need to do the reading. You can think about how it affects your course grades, whether you like drinking beer, COVID-19, that should be, I think, COVID-18 wasn't so serious, etc. So just a simple way of you, um, kind of playing with this notion of, uh, calls or loops, really trying to think carefully about whether these relationships are positive or negative. You can see even early on, I struggled with one of those, but hopefully for the rest of the lecture, I was quite clear about which ones had a positive or negative relationships between nodes. And just to see like how big your model can get, how complex it, it gets, and where you draw your boundaries around your model of the world, because that also has a fundamental effect on, um, what's really driving these things. So you might go as far as societal pressures for you to get a degree, um, as one of the things which actually is driving whether you, uh, read a paper or not.

So what are the two papers, uh, um, I've got two here. I'm not sure I was supposed to provide two, uh, but I think there's, they're kind of slightly different in their approaches. Um, they're both introductory texts. So essentially, they will say many other things that I've, at least the first one, the Kirkwood one, will say much of what I've said, but in a slower and maybe more evenly paced way. Um, and it talks about some other things. It talks a bit more about the system dynamics related to causal loop diagrams. So that's a nice thing to read. And then there's a paper by Richardson, which is a bit more critical about some of the problems related to causal loop diagrams. So I've been talking quite positive about it, but of course, like any other method, it's not without its limitations or challenges.

And then there's just one online resource you can look at if you're interested from Systems Thinking and Causal Loop Construction, provides some step-by-step instructions for creating causal loop models. And if you're really interested in this type of stuff, um, here's a few, uh, online software which you can use to create these models. But of course, you can just do this with pen and paper, it's as much the same thing. Some of these causal loop modeling software allow you to actually quantify changes, so you can create genuine dynamics. You can, you can have, kind of, changes in the stocks and flows of the things you're interested in, but we don't need to go into that detail here.

So, uh, that's it from me. I hope you got something from this lecture. I'm sorry that I had to do it as a floating head, um, over some raindrops, but that's unfortunately the way it is right now. Okay, thank you, and goodbye.