 Okay, thank you for this invitation and for allowing me to speak to you. I will give in this presentation a reflection on how to include social science into computational models. And I do this from my experience in the last 25 years in working in different disciplines and I'm trained as a mathematician but worked mainly in the social sciences. So I may make some statements, especially about social science that might be a little bit exaggerated. So but I want to make some points that may help you put a little bit black and white sometimes. So the outline is that I give some background to this context. Brief history about integrated models of humans and the environment. Then a discussion of some challenges in the social, getting social science represented in integrated models. And I provide two examples in which I work with social scientists in Molling-Hunter-Garders and about what the message connects to the city and I conclude and I'm not familiar with your culture but I hope to have some discussion. So far there have been only folks so I hope to have some discussion. So why do I have the title to Molling-Cultures? It refers to a famous essay by C. P. Snow entitled Two Cultures. And that essay is from about six years ago. It's referred to these distinct cultures between the humanities and the sciences. And he argued that they were quite distinct communities where we're not really interacting with each other. So this essay got a lot of critique. But and I'm not arguing that the natural and the social science are completely separate. But I think it's useful to realize that if you want to interact with the social science they may have quite different practices and as least that's how I experienced it. And that might be good to know if you start or want to work across these different areas. For example, a lot of social sciences are not trained, are not familiar with quantitative methods or are opposed to use any quantitative or using mathematical tools to explore their scientific activities. And I will come to that and there are good reasons for that. So here I give in a way a perspective to Molling-Cultures as if they are in the social and natural sciences. So my background is in a way I rely on my background to make this thing because I experienced these cultures in the hard way. So I hope that you may learn something of that. So my training is an operational research. I worked when I was doing my PhD in a group on integrated assessment Molling. And there was not much social science that was more kind of the natural science group, the image group, the well-known integrated assessment model. Then I got more interested in the social science to start working more in economics, environmental economics, and then in political and behavioral economics. And then I got a tenure-track position in anthropology, the Department of Anthropology, while I had never took a course in anthropology myself. So you can become a professional anthropology without having any preparation for that. But the department was transforming to a school which was more interdisciplinary and my higher was part of that. So I got the culture shock there too. I got tenure there and came from out to the professor. But now I'm in the school sustainability where we have no idea. It doesn't matter what background you have. So by moving around these different disciplines, I have experienced that there are quite a lot of differences in how people approach the use of mathematics. And so I teach typically mathematics or computational methods to social scientists and to ecologists. And so that leads to an interesting discussion. So I give now a brief discussion of some integrated models. So I think some of you may know this, but it gives a kind of context of where I come from. So in the early 70s, it was a well-known study on limits to growth, the Valtry model, which is kind of one of the first integrated models of human and natural systems. Although you may argue there's not really much natural systems. Social scientists will argue there's not much social, but at least it was a first attempt to bridge the different components. This is a figure where they also included data. Let's see what I can point to this. But the data which you see here are the actual data. So this is not from the study of 72. These are some results. They reruns them in recent years. And it shows that the kind of predictions about the verse not aimed to be a predictive model, but the doom scenario, we are still on track. There were many different scenarios in that publication, including a stabilization scenario. That scenario is not possible anymore, given the current trajectory. So this model was not aimed for making predictions. They had a lot of different scenarios, but the doom scenario mainly stuck. This study got a lot of critique. And the main critique was that there were a lot of assumptions that were really kind of expressing the subjective subjectivity of the models themselves. One interesting paper by Bill Northhouse, which we will see later, was the measuring without data. Bill Northhouse is an economist and he argued that there was no economics in this model. So that was, of course, an important challenge. There is another interesting book where they show that a lot of the results were driven by the assumptions they made you. There are a lot of uncertainties and you can make choices in the assumptions. And a lot of the outcomes were driven by the assumptions they made. I think that Dennis Meadows and his colleagues, they knew a lot of these issues and they therefore explored a lot of different scenarios. But these kind of doom scenarios stick to the broader audience. I think this model was very useful as an eye opener. It was not aimed to be a kind of predictive, detailed, human-environmental systems model. There have been a number of follow-ups, which were more regionalized versions. But the next major step in integrated modeling was the development of integrated assessment models. There are two different types of brands, as I see that. So this is a picture of the image model, the group I was working in. So you see that on the left, the human component is basically input. The input is the economic development, the technology assumptions about technology and the demographics. Those are not influenced by the environmental system. They are more or less exogenous inputs to the system that generates emissions and are inputs to some integration of simplified versions of the earth system dynamics. And then the output is climate change, which has some impact on land use change. But the human component is very limited. So I got interested by working in this group to say, why are humans not represented? So that's why I got more interested in getting into the human component, because they were not in that area of research. There's also another type of integrated assessment modeling. Wetz is more coming from economics. He received Bill Northhouse again, who was complaining about World Tree Model. He had developed a very well-known integrated assessment model that is used a lot. It has a handful of equations to describe all the relevant components of the climate system and the economy. And in the typical models in economics, they calculate the optimal policy for the next 200 years, assuming that the representative agents have full understanding and full knowledge of the system and full control of the system. So there are very tight feedbacks between the actions and the impact. And of course, a lot of these underlying dynamics are highly simplified and highly aggregated. But the model has been very influential in policy. So these two types of integrated assessment models, which were kind of starting in the 1990s, they in a way, and I left this kind of field in the late 90s, but I think there's still the same issues at the moment. And when we had this meeting a year ago, it was kind of a deja vu from my dissertation work, because 20 years later, there's still the same debate between the kind of people working on human components and on the natural science components. So can we capture the complexity of both the social and environmental systems? I think that's one of our aims. And I think there are a lot of reasons why there's still this divide. And I think it also has a lot to do about the interaction between the disciplines. But I try to provide one, give one example of what I did in my dissertation, as I tried to use some at that time, when I did my dissertation, where new tools coming out of complex adaptive systems to represent agents who are able to learn and adapt. And so I combined elements of the image model and the dais model. And so I had the economics of the dais model and the climate system of the image model. And I had agents who are actually had something that might be controversial today, they learned from facts. So that is so that that was a kind of crude assumption. And sometimes it's really good to do money if you don't know much about a particular field. But I assume that that that agents will learn from facts. And and so I could explore different situations. What if the climate system was sensitive to the emissions? It was high climate sensitivity. Then the agents will experience a climate change more rapidly. And they may that may lead to a lower emission level. So a lower emission level go see with a higher temperature change because they cannot adapt fast enough. But if the climate sensitivity is very low, there's not much reason to adapt and you will lead to higher emissions. So this is an example of a kind of a more kind of integrated more. The agents are reacting to the environment system. But we made up a lot of what we were doing because we didn't really capture much. We didn't really understand much about social science. So but I got interested in these age-based models to explore that. What would that mean for these kind of climate integrated models? So I will say a little bit more about the human component. So some of you may think that, well, economics has been addressed. So we already addressed the human component. That might be true, but there are a lot of issues with economics. And that's also within economics. There has been a lot of discussion in economics about the kind of mulling they are doing, especially since the economic crisis 2008 has been a lot of debates about the way economies represent economic system. So this is by Nobel Laureate in economics, Paul Krugman, why economists get so wrong. Also, models, integrated assessment models by economists are criticized a lot. So this is an interesting abstract. What do the integrated models tell us? Very little. And if you go to more, they say, well, these models have crucial flaws that make them close to useless as tools of policy analysis. This is published in one of the leading journals in economics. So also within economics, the type of models they are using are criticized a lot. It's recognized that there are a lot of challenges. So if we want to include social science, what do we include? So there are a lot of other fields of social science. And last year, Jay mentioned that you had no idea that there were so many different disciplines in the social science. And there are a lot of different social science people ask when they people see that I work in social science. Oh, you're a sociologist. No, I'm not a sociologist. There are many different branches in social science. And they all do their own things. They have their own different theories, their own approaches. So they have different types of areas in economics and psychology, anthropology, sociology, all kind of other fields like history of philosophy. And so, but most of them are not as much involved in using quantitative models. So that's one of the challenges. The economics is very much more focusing and quantifying the work using mathematical models. So another issue is that if people read about social sciences, well, these are the results of the social science studies are quite obvious, which is true if you knew the answer. So, Duncan Watts wrote a very nice book about this issue. Duncan Watts is a physicist who became a sociologist. People may know him from the Small World Network research. So he did PhD in physics and became a sociology professor and now works for Microsoft. And typically, because of a lot of the uncertainties, a lot of the context related phenomena in what social scientists are studying, which I express typically in narratives in qualitative data, they have not very precise predictions. And so there might be alternative explanations, but they are kind of more more kind of general statements than very precise predictions. So, even if there are alternatives, if you are able to identify what is the most logical explanation that explanation sounds very obvious. So, one of the examples that Duncan Watts mentioned is why is the Mona Lisa painting the most famous painting in the world? Well, that may be because it's the quality of the painting or it is an accident. And I will leave it up for you to read the book. Of course, the answer is obvious, but I will leave it up to you to read what is the answer. So, what are the chances for the Mona? So, there are many different theories. In the United States, they like to disagree with each other, like to debate, and they are not have the culture to work together to develop this kind of more general theory of the studies they do. Ten years are often depending on writing a single authored book. So, it's less a culture of collaboration and contesting and building up this research and contesting each other's theories. A lot of the theories are qualitative, expressed in narratives, and so that makes it hard for Malas to use. Basically, we have to interpret, we have to translate the narratives into algorithmic expressions. So, that becomes a kind of subjective exercise. And a lot of the social science, they will argue that there's a lot of importance of the experience. You can only know something if you have experienced it and it's depending on the context, interpretation, what's the meaning, and that becomes very challenging in expressing it in algorithmic statements. And you may not have a model of human systems that you develop that could be applied in all parts of the world. So, what to do? So, some advice might be to take into account the diversity of the different theories that you acknowledge that there are a lot of, that you accept the subjective, that this is a subjective enterprise, and that you focus on some specific questions. So, I will now briefly show two examples to show the way we try to include these social dynamics. So, one is about hunting off by a chase. So, this is working with anthropologists. So, I worked with Kim Hill, who was an anthropologist, worked for 30 years in the Amazon, studying the hunter goddess, the hunt with them, et cetera. He had an enormous amount of data. And so, when you start working on a model, he had hours, many hours of anecdotes, how they chased the armadillos, and so this is all fine, but it doesn't help me to develop a model. And so, I secretly developed a very simple random walk model and then showed it to him and said, so what needs to be added to make it more in line with that we need to a better model. The model had, was in a very high resolution, kind of things that we are doing 60,000 cells. And we created a flow diagram of the decisions which fits very much in line with optimal fortune theory. So, it was kind of grounded in theory, but then contextualized to the case of the Ache and Paraguay. I will not go much into the details because of the time, but we had different levels of the model. So, we started with very simple random walk models. We included that they are in different camps, which moves every day and then they start moving through the forest more in a coordinated fashion. And even if they are, for some species, hunt together and we could increase the complexity, we could get better performance. So, but I had to force this process by providing a very simple version of the model so that my colleague could be more, yeah, express things that needs to be included makes more sense. We could then use this model to look at what is the optimal group size and actually the optimal group size was very much in line with what was observed. And if we could also, so what the Ache actually act in this landscape but would be an optimal case. So, on the reason that we do this kind of modeling of the undergarters that we are now developing a model in South Africa about 100,000 years where we don't have direct observations, of course, and that will be a more complex environment. But we can test different ways these ancestors have been moving around the landscape and interact with the landscape. And in the last few minutes, I would like to talk about urban water in Mexico City, which is a much more complicated situation on the goddess. Now we get into politics. That is very difficult to put into some equations. So a lot about flooding water scarcity. One reason why there is flooding is that Mexico City is built in a lake. So while the lake is not there anymore, but if it rains, it still end up in the center of Mexico City. And there is water scarcity because a lot of places don't have pipe water. And so there are major problems there. And so we have a large NSF project in which we try to integrate different components of the problem we look at. There are some hydrological models, some climate models, some growth models. And me and a poster focus on more the kind of policy environment. That's the age-based model where we focus on the interaction of the water authority with the neighborhoods. So the process here is that we have the water authority. They have the priorities. They have their mental models about what is important for where to invest. Do we invest to prevent flooding or to invest in to reduce water scarcity? And then these decision trees affect the landscape in the different neighborhoods, like 2,000 neighborhoods. And there are flooding events related to water scarcity that will have output impact of the neighborhoods. So neighborhoods may adapt, so neighborhoods may protest, and that will affect the actions of the water authority again. We exclude a lot of the elements that social scientists found important. We exclude all the informal economy. We don't include corruption. And so if I talk about this to social science, they're all upset because I didn't include all the important elements. So we focus really on some very specific elements about the interaction of the water authority and the neighborhoods about what they invest in the infrastructure. And the infrastructure then interact with the biophysical system. So we now have some initial version of the model that we are starting to discuss with stakeholders. Stakeholders are the water authority, but also neighborhoods. And we are actually next week we are doing some of the, we start doing some of these validation activities. We have some initial results about if we simulate this over the next 20 years, which areas do we expect to have more investments, which areas will have more water scarcity. And this will be done used in an interactive way with stakeholders. So to conclude, so a lot of social scientists qualitative and a lot of the theories are described as narratives. And so if you want to translate from these narratives to these algorithmic statements, you have to, this requires some subjective interpretation. So you make some points about predictive models. We don't really like to talk about predictive models in the social science that at least not for a lot of the things that we are doing for some issues you can do. But typically we want to explore a lot of different possibilities, explore a lot of the assumptions on the underlying different theories. And if you want to work with with the social scientists, I think it's important to really put in time to understand also what the social scientists want to. What is the importance of the social sciences to to represent in the malls. What is the interest for the social science to be involved in in the projects that you are working on, because they may have very different questions than we typically have. And that interaction may take years to really understand each other. At least that is my experience. And maybe I have one or two questions that I can take. Okay, thank you. That was great. Thank you. So, a while ago, after the Brexit, there was a lot of criticism of the Bank of England. Is there anybody from Britain here? Good. So if I could just go wrong, don't hesitate to correct me. Okay. This is kind of subtle, I think. So the Bank of England come out a lot of criticism for getting the model wrong. They sort of joined the cadre of these economists who are not able to model the future. And so there's an interview on the radio with the, you know, the guy in charge, Canadian, I can't remember his name. And he said, the model was not wrong. It's just that when the Brexit happened, when the vote was to leave Europe, and the UK economy was about to go down a tank. They had to invest a lot of money, a lot of new money basically into the economy, and a lot. I can't remember how much, but it's on the order of the billions and pounds. Now that wasn't in the model because they're not allowed to to think about what they might do if the outcome was of a particular direction. So that type of caveat, I guess you could call it. You put that into a model when when economists, particularly those who are highly regulated, like the chairman of Bank of England, they themselves can't put those into the model. How do you, how do you do that? So I was involved in some of the IPC scenario exercises and also there we got scenarios that we were told not to to to use in a way we got some results that were not acceptable, like that some parts of the world will really go down in some of the scenarios. So yes, there is a lot of politics involved in that way in some of these social science applications. So I don't know about this case in the UK, but I know a lot of these economic models are when they are used in practice, they are not literally you using the models, it's a lot of the qualitative understanding of the system to when they are giving advice. So that that has always been the case since they start using econometric models. So that's, so it's in the social science very common as well. Yes, you have a model but don't rely too much on it. There's a lot of qualitative understanding that we are not including. But then that is not always known beyond the way those those fields out is how these models are used. I'm Lawrence Buzha at the National Center for Atmospheric Research I run a group that interfaces between social and physical sciences. And this isn't a question is so much as a challenge for next year's program to have the same talk given by a social science. Because what I found I'm not a social scientist you spend a lot of time telling us that you weren't one either. What I found working with social scientists is that their theoretical basis just as deep as on the physics side and as an interface group I I'm always asking the social scientists, give me frameworks that let us work on this boundary that are so simple that even a physicist can understand them. I really love to hear this talk from a social sciences perspective next year. So whoever doing the program please think about that. Thank you. Hi Stephanie Kane I'm a cultural anthropologist. I'm wondering, it seems like complexity is the biggest obstacle in the modeling problem and I'm, I'm hoping to hear in the next few days if we can build flexibility into your model, rather than complexity, so that taking the example. Let's say the water management is the head guy is a corrupt guy. And he's saying I'm not giving water to these four neighborhoods or these 25 neighborhoods. That could be modeled in if you had access to that kind of information. Is it possible to build models such that once you get some information from on the ground and you know, what is the most relevant thing that you can factor rather than just sort of give it up and say, oh my God, this is social sciences to complex and then picking an arbitrary, which then gets called subjective parameter to include. Well, I don't think I disagree with that arbitrary is the same as subjective but so if we will know what the consequences if we had the idea of this is what the consequence will be of this corruption as opposed we can include it but we are trying to include the decision making of these water authorities for the next 2050 years so we cannot rely on on what we observe today by some individuals so that's why we have to model some processes and we purposely try to keep it as simple as possible to to make it to keep it transparent so it's not a predictive but we can use it with the different stakeholders so that they get an idea of what will be the consequence if you are combining these different simple rules of the neighborhoods of the water authorities and the biophysical system together. That is something that that people had not a good overview it if you combine them together and what will be the long term effect. But yes, we purposely keep it as simple as possible so that it is transparent and so that's maybe my mathematical background that we tried to keep it as simple, but it's not aimed as a predictive model. I think you that are for you you the modeling is only one of the different elements using this process. So that's that's I think also we should not over focus maybe focus too much on the modeling and some of our applications. This is about modeling so I was trying to jump into that but I think what you're doing is you're picking an arbitrary scale into the future and then and then making things simpler in the present. And that may or may not be relevant for the people that you're, you're trying to work with, but I disagree with the words arbitrage, they are not arbitrage. But in a complex world when you're choosing one or two or three things to put in the model, I would say is an ethnographer that I would want to talk to people who live there, or talk to someone who knows about the people who live there and then decide what my choice is going to be. And so then my model has to be flexible enough to be able to incorporate some of these things that I might not have sort of pulled in on before, but this is like three days.