Nina Fedoroff: Good morning and welcome to all of you. I am probably known to most of you. I am Nina Fedoroff. I am the Department of Science Adviser. Welcome to today's Jefferson Science Fellows Distinguished Lecture. Today's speaker is Professor Claudio Cioffi-Revilla. He holds an undergraduate degree in mathematics and science and doctorates in political science from the University of Florence and the State University of New York.
Claudio was the founding director of the Center for Social Complexity and the founding chairman of the Department of Computational Social Sciences at the Krasnow Institute for Advanced Study, both at George Mason University. Claudio came to the State Department as a Jefferson Fellow in 2006/2007, working in the Office of Geographer and in other parts of INR. Dr. Cioffi's research has been funded by Innocept DARPA and other agencies and has included collaborative research with the National Museum of Natural History, the Smithsonian Institution, as well as the State Department.
Professor Cioffi has published many papers, as well as five books, including one titled “Politics in Uncertainty” and another one titled “Power Laws in the Social Sciences.” Dr. Cioffi, as you can see, will talk about the relevance of computational social sciences for foreign-policy. Claudio.
Claudio Cioffi-Revilla: On a rainy day like this I think it takes special dedication and passion for science and policy to attend a gathering such as this. So, I thank you all.
Good morning to you all and I want to thank the department for this kind invitation [inaudible] especially Dr. Nina Fedoroff, the Science Adviser to the Secretary, as well as Ambassador Reno Harnish, my friends from INR, GGI, some of whom are here: Dr. Lee Schwartz and Sue Nelson and many other friends here at the department.
I also want to thank, before beginning, I'd like to just express some thanks to my own provost back at George Mason University because, without his consent, my year at the department would not have been possible, nor the continuation of the program through a number of episodes and exciting events. I also would like to thank the National Science Foundation and the Office of Naval Research for funding my research and that of my group and colleagues because it relates directly to what I will be discussing with you today.
At NSF, folks like Tom Baerwald, and David Lightfoot, and Rita Teutonico have been very supportive of the Human and Social Dynamics program that has made a huge difference for the field of computational social science, including highlighting opportunities for policy analysis and policy relevant scientific research; and that has been a rare occasion.
Well, in thinking what I would speak about today as part of this Jefferson Lecture, I thought that I couldn't do this absent the experience and what I learned in the year that I was in the building, so to speak, and I wanted to harness some of those ideas to put a different twist on this. So this talk is very different from what it would be if we were at a scientific conference or at the University. I have tried to keep in mind some of the most important things that I learned that year, in terms of the whole relationship between analysts and policymakers, or analysts and the upper floors of the building, because it was only after I was able to look in that direction that I was able to, I think, make some of my own scientific work that year a little bit more relevant to the work that takes place in the building. And, although I will not be able to speak about the work that I actually did at the bureau, you can see that I still [inaudible] I brought my INR hat and I'm very happy to display that. So…
Having had the opportunity to work on assessments and estimates was a really wonderful thing because it actually emphasized some of the very important principles of scientific research that somehow are actually highlighted when it comes time to apply scientific ideas in assessments and estimates for analysts and sometimes in the regular course of academic activity some of these things actually may not seem as obvious. For example, the necessity to work with multiple lines of evidence on data, the necessity to document and evaluate both validity and reliability of information, the necessity to take a very close look at the process through which information is fused into a conclusion, a judgment on whether something occurred or did not occur or how it occurred or where it occurred or when it occurred and all of these important features of analysis. I was struck by the similarity and the parallels that exist between the work of the foreign-policy analysts in the unit that I worked with and much of the academic work that we do at the University, except we produce for journals and analysts, of course, produce for the upper floors and decision-makers.
So I thought I would do two things today, given the limited time, because the first of these would be to explain what is this field of computational social science because I'm sure that, except that the few of you who know me and know something about this, the rest of you have really no reason to know what this field is about. It's a very new field and I'd like to spend a few minutes explaining some of its content and why I think it has relevance for the analysis of foreign policy.
However, I would also like to spend most of the time actually talking about some plausible applications in the area of foreign policy and I'm going to emphasize two kinds of concerns among many that could have been chosen, such as, for example, the issues of terrorism or nuclear proliferation; many other issues like these. I've chosen issues of governance and environmental change as those of primary interest to illustrate some of these ideas and then I'm going to give some closing remarks.
So, first of all, what is this field of computational social science that has such an impossibly long name and a difficult way to situate within the fields of science?
Computational social science is computational both in a theoretical sense and in an instrumental sense and I thought I would highlight this contrast by first bringing up the conventional, the traditional social sciences. When we talk about social sciences, we talk about a variety of fields: multiple fields, so it's really in the plural, right, and this traditional orientation is very disciplinary in orientation. There are academic disciplines that have parsed social science, although the initial founders of social science 200 years ago perhaps did not intend it that way. But today, it is what it is and we have very strictly parsed areas of social inquiry. These, mind you, these five classical disciplines of social science are the traditional ones that you see often invoked but there are other fields of social science that are not listed here but are as important and as attractive and active. The only difference is that they are somewhat less disciplinary in the sense of being somewhat less, somewhat more open to interdisciplinary collaboration.
The first one that comes to mind would be, for instance, geography, okay? The social science of geography is a very, very interdisciplinary, very porous social science; very open to innovation: methodological and theoretical innovations from any other fields from these five classical areas but from many other areas of mathematics and science as well.
Besides geography, another social science not listed here would be for instance management science and organization theory, it is very, very important; doesn't quite fit into any of these, perhaps, in part with sociology. There are also parts and subfields of these five traditional disciplines that have grown in their own identity to constitute quasi-autonomous disciplines. For example, criminology within sociology is a very robust, well defined with as many journals as the whole profession of sociology has.
But the main point about this is that these disciplinary social sciences have produced a myriad of theories and theoretical approaches, I would say, I would argue, in a fairly fragmented way. The classical methods of the traditional social sciences have been initially history but then also more specialized methods such as ethnographic methods and statistical and mathematical approaches. Among the statistical methods most famous of all are probably regressions, statistical regression models that some of you may be familiar with and among the mathematical methods unique to the social sciences, that is to say not shared with natural science, for example, would be game theory; game theoretic methods that were produced by social science need rather than imported from other fields.
Well then, these being the classical traditional fields of social science; what does computational social science look like? This is something by the way that I had -- I was not so cognizant of having to explain all this when I walked into the building three years ago but it was very important because people would ask me, "Well, what do computational scientists do? What do you do? I mean, I'm an economist or I'm a biologist. I know about computational biology but what is computational social science?" So computational social science has a different paradigm. First of all, it tends to be an integrative discipline and it tries to have itself spelled in the plural. Now, because of typos, the term computational social sciences has actually occurred even in some of the printed literature at the university, but the actual original name was actually in the singular because it tends to be an integrative discipline or field.
It has a theoretical paradigm, which is to say a point of view on social systems and processes that I would say, I would argue, is unique to this approach; not perhaps unknown to the traditional social sciences, but it is a defining point of view. And it is a defining point of view that we view social systems and processes as primarily information processing phenomena that -- where information plays a much more critical role in defining the organization of people and groups and entire organizations, more so than in the traditional social sciences. So, information processing is very, very important.
It's also very important to note that information processing is something that computational social science examines from the cognitive level of the human mind. That is to say, how we process beliefs and attitudes and other cognitive, mental components through human decision-making, the interactions of decision-making into groups, into organizations, entire societies and the global system. So, it's this multi-scale approach which necessitates integration in order to work in some scientifically coherent way.
The methodology of computational social science is also, I would argue, distinctive. Simulation models, especially of the recent generation of so-called agent-based models are already defining method in this field. We'll talk more about that in just a moment. And theory and methods from complexity science; these are ideas that have been recently rediscovered in the social science but were actually known for quite a long time. Again, I'll come back to that in a little while.
Methods from the analysis of networks are quite prominent in computational social science, although, here again, like everything in science, there are antecedents. The foundations of modern day network analysis and computational social science are traced back to graph theory in mathematics and the work of Euler and others who are very well known to mathematicians, applied mathematicians; somewhat known to sociologists and other traditional social science scientists but which, until recently, was not really very practical because it requires a computational detail that is very difficult and was very challenging until recently. And that is to say the manipulation and inversion of large matrices, which is computationally expensive and was rendered possible only through greater computer brawn in recent years. Work that we now do in matters of seconds or minutes used to require days to run on IBM 360s when I was in graduate school; and so there's a revival of network analysis in social science and computing because of this.
And finally, I would say that the fourth method that identifies and characterizes computational social science are these methodologies for automated information extraction. This used to be called content analysis in social science a long time ago except most content analysis began in a manual way and then slowly became automated. But today, powerful algorithms are used to extract information automatically from text but not only from text, increasingly so from imagery and other media.
So, computational social science is so in a theoretical way, due to the paradigm and the perspective on informational processing in human social systems and organizations. But it is also computational in the sense that it is a discipline enabled; enabled by modern computing. And I mean that in the same way that microbiology was enabled as a discipline thanks to the invention of the microscope, which opened for scientific investigation whole new areas of life that were plainly inaccessible before the invention of the microscope. The same is true of radio astronomy; another discipline that was to instrument enabled and driven. Nanoscience is the same as well. So, these fields computational social science shares with these other fields of science that have been first enabled and then driven by instrumental developments.
So, this means that the kinds of questions that computational social scientists investigate are distinctly different from the kinds of questions that our colleagues in the traditional disciplines usually investigate because we do so enabled by the power of computing. That means that we are able to integrate greater social complexity, examine it on many different scales, use media and data from domains that are normally not integrated or mixed from the traditional disciplines and so on. In other words, it is social science research where the instrumentation of computing adds value and renders investigations feasible in ways that would otherwise not be the case, okay?
So that's why we need to keep in mind the analogues of radio astronomy and microbiology and nanoscience because those two are areas that are simply not enabled, absent the instrument; the scientific instrument of investigation. So, the use of computers in computational social science is not just to speed up work and to do something fancy. It is because it actually allows us to see, to think about, to reason, to understand areas and parts of the social universe that would otherwise not be accessible. It's kind of exciting. Very exciting.
Here's a way in which you can think of these as two complementary perspectives, okay? Now we're back here to the idea of the paradigm and the point of view of computational social science that it's slightly different from that of the traditional disciplines. In the traditional social sciences if you think about election; political scientists, social scientists that mostly investigate and do research on elections and all kinds of theories and methods and so forth, but if they view elections mostly as a process of casting votes from the point of view of computational social science, an election is actually a large scale, multi-scale computation in a society of agents to try to compute, right; to figure out who is the leader, who is the leader with the strongest support? And everything else that we observe: the party system, the electoral process, the scheduling of primaries, and final elections, and so forth; all of this is part of the schedule leading up to that final computation. I would argue it's a somewhat different perspective, okay, not at odds with, not incompatible with, the casting of votes, okay? But it is a perspective that involves scales of human organization and time horizons and the flow of information processing way beyond what is necessary in traditional political science to talk about an election.
Another point of view would be, a contrasting point of view could be seen also if you think about public policy. Those of you that are familiar with the field of public policy and political science know that the allocation of resources is one of the major ingredients of this, albeit not the only one. But that's the main issue. How can scarce resources be allocated among competing priorities, which is a vexing problem of the policy process; no doubt about it. But the computational perspective is somewhat different because, as we'll see in a moment, policies are actually viewed as parts of the larger metabolism of a policy -- of a polity, excuse me -- and, therefore, a policy is actually an adaptive response by a system that is seeking to accommodate itself and regulate its internal state in an environment that poses challenges and opportunities. Those are very different perspectives, again, not at odds with each other but certainly they highlight different ideas.
What difference does all of this make? Well, the question about relevance of computational social science for foreign policy is very important and very apropos because, obviously, there would be no need for this discipline at all if it didn't have something important to say. To give an exhaustive account of this is impossible in the few minutes I have but I can tell you this: that work in computational social science that looks at human and social dynamics on many different scales is applicable and is about problems that occur at the local level, nationally or regionally, internationally or global. All of these scales of human and social dynamics are well integrated and are well accommodated within computational social science research and theory today. I'll give you a glimpse of some of these examples in a growing literature that can be better appreciated through the usual venues of professional meetings and journals and so forth.
I would raise two very important points. So, computational social science is not a field that is dedicated to the study of, say, bargaining and negotiation at the individual level, or has to do with long-range civilization dynamics at the global level. It really cuts across these different levels of human interaction.
There are two important points about this. First of all, computational social science is useful in highlighting the role of information processing; information processing across levels of complexity that have to do with human, with artificial and with natural organizations, natural systems.
The first of these human systems and human organizations have to do with people related by social bonds, whether informal or formal organizations. That's the universe of humans.
At the other end, we also have the world of nature with its dynamics between climate, biomass, morphology, hydrology, everything that constitutes the surface of the earth on which people live. Those natural dynamics have their own complexity and flows, not only information of course but also energy and other critical processes and systems.
In between these two, humans have invented artifacts; artificial systems that mediate the interface between the two as we gain in levels of quality of life that sustain higher and higher performance. And so today, for example, in a building like this; this is a huge artifact where we are able to be in a perfectly dry and comfortable situation. Look outside the street what's going on in nature, right? So we create these artificial natural environments and these artifacts are often human in their formal organizations. They are often physical in the form of engineering works, okay? It's important to distinguish that items of complexity exist for all three kinds: for social involving humans, for artifacts regarding the artificial constructs that we build, and then in the natural environment. This all sounds very obvious and [unintelligible]. The only problem is that traditional social science has never had – it's always had a very, very difficult time integrating these kinds of organizations in a cohesive way. I'll show you how that occurs in just a moment.
So the other point I would raise in terms of the relevance for foreign policy analysis is that this is a field that depends on the exploitation of the power of computing and that is just getting better and better and better every day. So, whatever we are able to do today, we will be able to do quantum better not too long from now. This is almost a sure bet. This is almost a sure bet because already what we are able to render today was impossible to even envision, even envision 10 or 20 years ago.
Okay, let me now turn to applications in foreign-policy. I've spent a little longer on this than I had anticipated but, nonetheless. Here what I would like to do is to focus on some important ideas. I won't say basic, but important ideas. Sometimes important ideas are very simple. And I'd like to start this with a standard model of a polity. Polities are obviously extremely important in foreign policy because we are often concerned with problems in friendly polities, in adversary polities, in polities that are neither friendly nor adversary but maybe we would like to make them more friendly than they are at the moment. So, understanding how a polity operates is a very basic and fundamental thing.
Usually political scientists in the traditional social science have sort of a corner on this issue because, after all, they are charged with studying government. The standard model of a polity you see here is quite simple in its bare, bare, fundamental essence. So, this is like looking at the fundamental structure of an atom or something like that, or a water molecule with basic bonds and -- we have a society consisting of people and organizations. There is a -- these people are affected by public issues that come up sporadically; some of them from inside, some from abroad. These issues affect society and government issues policies to deal with these public issues to mitigate the effects on society and lower the stress that they cause on humans and groups; and all of this is rendered possible through the contribution of resources in the form of taxes and other resource flows. And there are also ways in which the government, the system of government affects the dynamics of society. I won't go into that.
This, by the way, this is the way every imaginable polity in the real world is organized in a fundamental way, whether we’re talking about 5000 years ago or we’re talking about today. This is the basic structure of how a polity operates. What introduces a great deal of variation in practice is the relationship between society and government which can take on many, many, many, different forms. Many of these forms have been cataloged and rendered in elegant taxonomies by political scientists over time; and they range from democratic arrangements to very totalitarian ones and different kinds of regimes. That's where that linkage technically is called the constitutional, the constitutional form.
The same standard model with a little bit more detail and the detail is interesting because now in addition to the basic entities that we had before, we begin to add attributes that characterize each of these entities; and the interesting thing about these attributes is that they are like states of these entities. So, a society may be highly stressed or not so stressed, depending on public issues that come up and these issues may be of many different types. Some of them may be more salient than others. Some of them cost more than others in order to solve. Some of them are trivial. They might have a low priority; some of them higher priorities. So we can think of these issues as having a set of attributes. Government, of course, has lots of different attributes in terms of -- a very important one here is the capacity of government to actually be able to produce policies that solve these issues that are affecting society and so forth. So, what I intend to show in terms of this diagram is that, although the simple model of a polity is quite basic, the details are important and they are rendered feasible in a computational model. This is a point of view that has been, until the computational age, let's say, recently, nearly impossible to solve in classical social science, especially in political science.
The simple polity has a very strange feature which is that it works theoretically and we understand this as the basic mechanism of a polity and how it operates so issues come up, they affect society, the government issues policies and so forth. So, you have this basic metabolism that is well-documented and you can actually show many examples of this in practice. We had this big financial problem that came up. As it was rising already, government was beginning to think about doing some things and it took forever and then it finally started -- now are starting to maybe alleviate some stress. Maybe the worst is not over yet and so forth. Every time a big public problem like this occurs, obviously, the government has to mobilize with collective action to produce policies that solve these problems. So, it's like a problem-solving mechanism, if you will.
The important thing about this is that being so simple, it wouldn't work in practice. It wouldn't work in practice because it's very brittle and very fragile to flaws. If government is unable to obtain sufficient resources, then it would not be able to produce policies that would not be able to dampen the issues and so forth. There are all kinds of assumptions about the operation of the system that regard the availability of information, the reliability of implementation of policies and all kinds of things. So, in practice, this probably wouldn't work. The way it actually works in the real world is because some redundancy is built into these systems so that there are multiple ways in which governmental operations are supported. There are multiple agencies that track problems. There are multiple coalitions that can actually put pressure on a government in order to do something about issues and so forth. So, there is multiplicity in the real world.
The problem there, however, is that once we understand that it has been analytically impracticable for the traditional disciplines to deal with this problem and this is a deficiency for foreign policy analysis because we need to do a better job with that.
So in the real world, what happens is that polities are more complex. This is more or less the complexity that they have. In the real world you still have, of course, a society as we had before. You still have public issues that affect society, okay, so this part of the diagram, this part of the computational model is the same, okay? But what is added is two kinds of governance are now made explicit. Before we had a single system of government which was this one, of course. Now we differentiate into a polity that has an official system of government as we had before, okay, which still obtains resources from society and so forth and issues its own state policies and so on and so forth like before. So this part on the right-hand side here is a metabolism that still operates just fine.
But in the real world there are also any number of alternative polities operating because there are groups in society that can form into providers of public goods that can reach sectors of the population and, in some cases, substitute for the official government. But as far as the population is concerned, the society is concerned, okay, they receive resources and they have a set of their own policies, which we can call alternative policies. A good example of this is, for instance, in areas where the state has limited control of the public administration on the ground and the provision of public goods for policies and other groups, non-state groups, non-state actors substitute. Hamas and Hezbollah in the Middle East are typical examples of this.
Another recent example that I like is the – maybe you've seen this map that appeared in the New York Times just a few weeks ago. This is a map of Mexico with the areas of the various narco-cartels. These are areas where alternative polities operate; not the official Mexican State and in these areas the societies obtain the provision of public goods when they are able to obtain them through the action and operations: schools, hospitals, even transportation systems and so forth, provided for by the cartel, including security in some cases.
I also want to point out that no country in the world exists in the ether. Every country exists in some kind of a localized geography, in some kind of space. They may have deserts and mountains and rivers and so forth but it doesn't exist in the abstract. So, the simple polity that I showed you before, okay, is extremely simplistic and naïve in that sense. Every country, of course, exists in some part of the world. Part of the complexity of a complex polity has to do with the fact that it is a polity that is embedded in some kind of natural environment. The natural environment exists in the actual surface and location, as well as its climate and weather and so forth.
I'm beginning to suggest that what computational social scientists are up to is the integrative modeling of this kind of complexity that we are beginning to render in actual simulations, which is a scientific goal that goes way, way, way beyond what any of the traditional social scientists has been able to do in the last 200 years. That's a big claim. It is a big claim but it is true. It's a real claim. We are beginning to produce increasingly realistic models of countries and regions. This is already done, by the way, on the CD scale on metropolitan areas and so forth to simulate urban growth and transportation systems and interaction with microclimates and local, natural environments: estuaries. The Chesapeake Bay, for example, has an agent-based simulation model that's very interesting. We are beginning to do this on a country scale, on a regional scale and one day the global scale is not very far away and that is the integration of human and social dynamics with natural environments and not just the official, stereotypical form of governance but also these competing polities that arise in many, many interesting situations with foreign policy and policy around the world.
Beyond the Mexican example, which we could talk about this for a long time because one of the interesting things about this is that Mexico, as you know, also has many different biomes and ecological zones. These cartels; some of them operate on a fairly narrow range so their operations, governmental operations, as it were, are fairly limited. But some of them span various ecotopes and it's quite fascinating to see how they are actually able to do that while being an alternative, non-official type of system of government.
Here's another example: the current Swat Valley crisis in Pakistan. I was reading this the other day on Monday and maybe some of you read it as well I'm sure, and it's fascinating to see because here you have direct, current, important evidence of these two polities that operate within a complex polity. So, what is happening is that after the entry of the troops in the Swat Valley and the taking of the large urban centers, there are Islamicist organizations that are setting up their own provision of public goods in terms of shelter, food, water, medicines, medical care, and some of them, like this one was -- by this fellow here has been -- here you go -- supporting on the order of tens of thousands of people are being provided with basic needs that refugees have.
So here you have a concrete example right there in the Swat Valley; the formation of the official polity with the institutions of the State of Pakistan; a society with many displaced members: the issue being the current crisis; and the alternative polities of a variety, not a single one, but a variety of NGOs and other organizations, some of which are perfectly legitimate and have no further intents but some of them are not. There was a commentary that I thought was very striking at the end of this article and it was said by this fellow Mohammad Ahmad, who apparently is the executive director of something called the Initiative for Development and Empowerment Access. Now this initiative for, let's call it the IDEA. The IDEA; this is an unofficial purveyor of public goods, a non-state purveyor of public goods, okay, so it is a policy producer. He has it exactly right. We can win these people over if we give them the proper support. This is how alternative polities become situated in an area and then compete with the state.
So computational models to sum this up, especially the variety of models known as agent-based models are being developed with increasing realism. We have in my group, and we're not alone in this, developing multi-country models like this 10 country model here on the right where we are actually able to induce climate change that affects the economy of the society, that affects eventually what happens in the instability of the country. This increasing realism today is insufficient for purposes of prediction but very, very helpful in terms of providing insights and dynamics that we had not thought about before.
It's also a practical instance of couple, social, artificial and natural systems. They are all integrated here. Typically the scientific themes that make up these projects are composed of many different disciplines from the natural and the social sciences. And finally, they provide an experimental capability that has long been denied to social scientists, especially the capability in terms of being able to run experiments on computational simulation models to understand what if the drought next year is going to be 10 percent less than anybody imagines? What will this do to all other things? How does that play out in repercussions?
So in conclusion, I am going to leave you with three ideas here. First, is that there are some of these tools and ideas that are ready to add value. I had an opportunity to do some of this when I was in the building that year, in an assessment we are working on, but there will be more. There will be more but all of this must be held accountable scientifically, according to the normal canons of science, which have to do with verification, validation, peer review, publication and so forth. This is absolutely essential and no different than anywhere else in science and it should be that way.
Another point that I would like you to think about is that we do have a strong leadership in this field today in the states, but that is not guaranteed and there are very, very, very active players in Europe and Asia; very active. I believe that the Jefferson Science Program could be a very important help in this direction. It should continue its interest in this area and I hope that some of these opportunities expand in the future.
Nina Fedoroff: We have time for a few questions; would you go to the microphones and identify yourself if you have a question?
Male Speaker: That was very a interesting presentation and two questions: one is in the example, for example, of the climate change in the 10 countries. To what extent do you use iterative methods to refine these models? Do you sort of take a common sense look after you first run the model and go, "Wait a minute. That's wrong. We need to alter this element or that." And the second question is to what extent do you borrow theory from the existing social sciences because it seems that you are having some kind of theoretical construct when you say, “Let's feed climate change in as an independent variable and see what happens to the independent variables which are the people and all these things.”
Claudio Cioffi-Revilla: Great questions, thank you.
The first one; the process begins with best estimates based on the literature in terms of, for example, parameter values and relevant relations that have to be built into these models. And we try to -- Dan, my colleague from the Smithsonian here is smiling because this really rings a bell to one of the most important models that we've been running so far. Once we get the model in a configuration that it can actually do complete runs, we invariably need to make adjustments. Sometimes it's because it behaves in ways that it should not behave. Sometimes -- and the sooner we discover that, the better. Sometimes, unfortunately, those anomalies are not discovered until much later, but we do make every effort to incorporate real-world parameters into this. The model that Dan and I were developing at the Smithsonian. For example, we had an issue of climate change in inner Asia inducing change on the biomass on the ground and that turned out to be crucial because, as you know, nomadic populations follow herds and the herds follow the biomass, basically sort of. So, changes in the duration of seasons and the precipitation and so forth have an effect on this. We didn't have these parameters at our fingertips when we began the project so we did, yes, we went to experts and read the papers and got the estimates and double checked things and so forth. That's very, very important. It's important to do with increasing accuracy and increasing realism as the model tends to portray more and more fidelity of what's going on in the real world.
The other point is very good too because there are two ways of doing these models. In the first way, you don't use a whole lot of theory from the natural sciences or from the social sciences. But you can rely sometimes on what I call "artificial intelligence tricks," AI tricks. These are algorithms that sort of behave like the social dynamics behave or like climate behaves or so forth but they are not real physics and they are not real sociology. We do try to -- a distinctive feature of these kinds of models and we do try to incorporate known patterns of social behavior and known processes in physics to govern and regulate this. We think that's a very important thing to do. Thank you.
Dr. Deedar Tam: I am Dr. Deedar Tam, the Council of Science and Technology from the Embassy of India.
As I've understood from your talk, the computational social sciences; it's if you look into the traditional social sciences in the context of decision-making, it helps more on that qualitative basis. Is it the tool like just trying to convert it into their quantitative form? And if so, let the three systems which we have talked about, like specifically the social system, the natural systems, and the other, the artificial system. I think the natural system we have understood to a certain extreme because there are a lot of natural models which are available which simplifies. The most difficult because I have a little experience in this natural system modeling is the uncertainties because if we look into the uncertainties of the social system [unintelligible] what we understand. So how do you handle, how do you specifically align in the computational method? Thank you.
Claudio Cioffi-Revilla: That's a great question. Here's a slide that I did not use but I had prepared.
At the end of the day, it's really to what extent are these methods and these scientific contributions helpful for decision making? And here, based on social science theory, we can distinguish the kinds of situations, rights? Under certainty, this hardly ever happens in this building, right? Consequences are deterministic, are completely known. I mean, this is a different area. This is like manufacturing. You have decisive situations; not in foreign policy.
Under risk or under uncertainty; these are the two major classes of decision making that take place in public policy, doing foreign-policy and national security. In the first of these, it's under risk because we have some kind of a known probability distribution that we can use in order to manage risk and choose between competing risks. We can do that. But sometimes we are unable to know what the probabilities are, or what the consequences are for that matter, and in both of these areas, computational models are helpful because they can help you either generate some plausible distributions of probabilities in the absence of other information; that's helpful, especially if it's defensible, really defensible. And in other cases you might actually be able to tease out consequences that were unforeseen or no one had thought about because these processes are too complex.
It's a very brief answer to a very important question that would deserve a lengthier answer but…