Peter Turchin — Why Societies Fall Apart (And Why the US May Be Next) (#149)

Peter Turchin is a complexity scientist and one of the founders of cliodynamics — a new, cross-disciplinary field that applies mathematics and big data to test historical theories.

Transcript

JOSEPH WALKER: Peter Turchin, welcome to the podcast.

PETER TURCHIN: Glad to be here.

WALKER: Peter, I first encountered your work, I think, in 2019, and I thought it provided one of the most compelling explanations of what I was witnessing in the US (albeit from the shores of Australia). I've also had Jack Goldstone on the podcast, in 2021, and I've kept in touch with him since then; I told him that we were doing this today.

And then, of course, you're the father of cliodynamics.

TURCHIN: One of them.

WALKER: Yeah, one of them. And cliodynamics is the application of big data and mathematical models to find patterns in history. But it does this in a non-naive way that draws heavily on complex systems science. And you have a new book out called End Times.

Before we discuss that and a bunch of other topics, I wanted to start with this broad question of whether a science of history is possible. So, as you know, Karl Popper famously objected to the notion of a science of history.

And I think his objection is best expressed in the preface to his book The Poverty of Historicism, where he presents this neat little syllogism that essentially says: number one, the course of human history is strongly influenced by the growth of knowledge.

Number two, we cannot predict by rational or scientific methods the future growth of our scientific knowledge (because if, for example, someone in the Stone Age “predicted” the invention of the wheel, then ipso facto they would just have invented it).

And number three, we cannot therefore predict the future course of history.

So what do you make of Popper's objection?

TURCHIN: Well, first of all, the future is not predictable to an arbitrary degree of accuracy. We know that we cannot even predict the weather two weeks from now — and that's a purely physical system which is completely understood in terms of how it operates, because of what's known as mathematical chaos. Human societies are more complicated, more complex, and therefore accurate prediction of an arbitrary time in the future is impossible.

So, for example, the famous Foundation series by Isaac Asimov — at least the first volume before the Mule appears on the scene — is based on what we now understand to be impossible to do.

But on the other hand, some things are predictable.

First of all, we can do scientific prediction. We can extract predictions from different theories and then use what happens in the real world to test theories and reject some in favour of others.

Secondly, some aspects of societal dynamics are more predictable than others. So in my book I talk a lot about the structural trends, which are very important because these are the structural trends that undermine the resilience of societies to shocks. And therefore when we have low resilience, high fragility, that's when we expect the outbreaks of violence, including major ones like civil wars and revolutions, to happen.

But the actual timing of when such outbreaks happen depends very much on what we call triggers. Triggers could be ruler assassination or a symbolic act like self-immolation, or a bad harvest or bad climate event. Those triggers are very hard to predict — probably they are unpredictable —, especially when they depend on the free will of a human individual.

We have to keep in mind that some things we can predict because they develop slowly and more or less regularly, other things we can not — and the actual course of history is a combination of those two things. So that's the first thing.

But let's now get to the objection of Karl Popper.

It is curious that he chose the evolution of technology as his example of why history, the future of human civilisation (let's put it this way), is unpredictable, because that turns out to be one of the more predictable aspects of the future.

I'll give you two examples. First of all, Moore's Law. It's just amazing that it just keeps working many decades after it was proposed. There is work by former Santa Fe scientists such as my good friend Doyne Farmer who published together with his colleagues a number of articles on the development of green technologies, for example; those curves are quite predictable.

Also, if we look in the great aggregate, looking over thousands of years of human history of the evolution of technology — so at a very coarse level — we also see it’s quite predictable. We have collected data, for example, on how military technology developed over the past few thousand years, and it is amenable both to analysis and there are quite a lot of patterns.

Alright, now a few anecdotes. People tend to think that proposing new scientific explanations is somehow due to an individual genius. I don't deny that many of those famous scientists were geniuses. Let's say we take the discovery of the laws of motion or, even better, the invention of calculus. So calculus was in fact invented by two separate mathematicians, almost at the same time, and in fact there was a bitter feud between them as to who had the priority.

An even better example is the discovery of Mendelian genetics. Well, Gregor Mendel actually was the true discoverer, but he wrote his paper and he announced his discovery before the field was ready for it, and as a result he was completely forgotten. Nobody knew about him — until 40 years later when three separate scientists almost simultaneously discovered the gene. One of the three scientists, who was the “re-discoverer” of genes because he couldn't get priority, dug up Gregor Mendel's article; basically saying, “Alright, if I cannot be the one, then none of those guys will either.”

So first of all, in the aggregate, we see that human technology develops in a fairly predictable way. That's one thing. And secondly, at the micro level, many of the great scientific discoveries, such as calculus, genes, evolutionary theory by natural selection, happen at the same time. And that suggests that it is really not the individual genius, it's the collective action of many scientists and thinkers who prepare the world for the next discovery.

WALKER: I think I agree with that premise. Have you ever read the article Robert Merton wrote for the New Scientist magazine in 1961 on this?

TURCHIN: No, I don't know it.

WALKER: I'll send it to you. It's exactly on this question He goes through a bunch of anecdotes for why individual genius is overrated in scientific history, and if you actually drill down a lot of examples turn out to be pairs or groups of people.

TURCHIN: There are individuals and there is collective action, and so there is no tension really between them.

WALKER: It's a false dichotomy.

TURCHIN: Yes, it's a false dichotomy.

WALKER: But it's also true to say that networks are like the air we breathe.

So yeah, I agree with that premise, but what is the conclusion that you want me to infer from those micro examples?

TURCHIN: The conclusion that I'm working towards is that the evolution of technology is to a certain degree predictable — not perfectly, but we can predict it to a certain degree — and we know that technology drives a lot of other historical processes. So a year ago we published an article in Science Advances where we looked at why large-scale, complex societies evolved in human history, and it turns out that the primary engine is between-society competition, taking the form of warfare. I have a book called Ultrasociety which explains the idea behind it. But what drives the intensification of warfare? Military technology. And so that turns out to be the most important factor that drove the increase in the social scale at which people are organised in states and empires.

WALKER: Okay, got it. So putting the question of the predictability of military technology to one side because I don't know too much about that, I want to attack your argument on three levels.

TURCHIN: Please do.

WALKER: Your argument around those examples of more than one person coming up with a discovery. So firstly, knowing that a technology will emerge is obviously not the same as knowing when it will emerge. And the question of timing involves a lot of contingency. There are many examples, but one is Charles Babbage got pretty close to inventing universal computing in the 1830s but then was plagued by interpersonal conflicts and a lack of funding and so couldn't really bring it to fruition. Another example: Francis Crick's autobiography mentions the fact that if he and Jim hadn't discovered the double helix, it probably would have been another two or three years until one of the other competing research groups did.

Timing really matters, because technology interacts with contemporaneous other technologies, as well as social factors, in really complex ways. So even if you have a pretty good handle on what is going to emerge, you don't know when. And that's as much of a problem as not knowing what will emerge. So that's my first rebuttal.

TURCHIN: Let me address that. Well, first of all, when we look at the evolution of human societies over the timescales of hundreds and thousands of years, what's two or three years?

No prediction can be 100% accurate. So making a prediction which is 1% accurate — that's awfully good in my view. Or even less than 1%. That's one thing.

Second thing: the Babbage engine is a good example of Gregor Mendel. Essentially, he was ahead of his time. It wasn't just because of personal interactions and things like that. We now know that in principle his engine could have been built, but it was really beyond the bleeding edge. At that point there was no need for it, because what would it be used for? His engines were perfected, really, during World War II and after, because the computational capacities were needed to break codes, to do all kinds of things.

And at that point there was enough infrastructure for them. And also at that point the material science advanced to the point where we had vacuum tubes and things like that. So I think you're, as they say, pouring water on my mill. Your examples actually help my thesis, especially if you keep in mind that 100% accuracy is unattainable therefore we should not throw away the results which give you awfully good accuracy but not perfect accuracy.

WALKER: Okay, good rebuttals. I'll go to my second point. So to say the future growth of scientific knowledge is unknowable is an ontological claim. To say it's merely unknown is an epistemic claim. And so let's assume the epistemic claim, which is the claim you're making, is correct. It still seems to fail due to massive practical issues. Like, I just don't know how we would ever make use of it, because it would seem to imply you'd have to keep tabs on every proverbial entrepreneur working in their garage around the world.

TURCHIN: No, and that's the nice thing about cliodynamics: without denying the important role of individuals, we have first focused on their movements at the meso and macro level. So at the level of not individuals but cooperating groups or whole societies, actually states, polities. And at that level of aggregation, most of the time individuals don't make a difference.

I'll give you an example. Whichever way I vote in the next presidential elections, the state of Connecticut is going to go for the Democratic candidate, and therefore my action of free will be completely buffered out and it will not have any macro effect. On the other hand, if, let's say, Florida becomes the key state in which the final decision would be made on who wins, maybe even one individual vote may not be enough to sway it, but one individual going out and canvassing neighbourhoods, getting 10,000 people to swing their votes from one candidate to another, that could result in a macro level result.

So what we are talking about is known in dynamical systems as sensitive dependence on initial conditions. But even systems which are chaotic, that generate these erratic trajectories from purely internal causes, they are not a fluttering of the butterfly.

The proverbial fluttering of a butterfly's wings is only going to create a major hurricane if the butterfly is in the right place at the right time. So millions of butterflies would be fluttering, there would be no macro event; but one would have an effect. Again, it's the mixture of predictability and unpredictability. And individuals become influential because they happen to be in the right place at the right time.

So Mohamed Bouazizi was the fruit vendor who immolated himself in Tunisia. There are lots of other people who have immolated themselves. There were a bunch of American veterans who immolated themselves to prevent the Vietnam War in 1965. There was no macro level effect of that. And you don't know ahead of time who is going to have that macro level effect.

WALKER: Okay, so this actually leads nicely onto the question of contingency. I'm going to leave out my third rebuttal because it's kind of already addressed by your responses to the first two. It was sort of about computational irreducibility. But actually, before I move to contingency, out of curiosity, do you think the course of history is fundamentally deterministic even if it's not predictable.

TURCHIN: No.

WALKER: Okay.

TURCHIN: Because I believe in free will. This is a big question. We don't have complete empirical knowledge that free will is not just an illusion. So this is a religious question. And I choose to believe that we have free will because that makes my life more meaningful and comfortable as a result of that.

WALKER: Just as an aside, one of the interesting things I was talking about with Stephen Wolfram was that determinism and free will aren't actually incompatible in his paradigm because, basically, the rules by which a system, say technology, evolves, are deterministic. But you can't take a shortcut to the outcome using your brains or methods of analysis, because they're computationally as sophisticated as the systems that you're observing. So you don't know what's going to happen. But the rules by which the system is evolving are deterministic. And in that space, you can kind of call that free will.

TURCHIN: It becomes a philosophical question. I'm not a philosopher. I am a scientist. I want to see what are the practical consequences of our beliefs. Now, whether we do have or don't have free will, or whether history is perfectly deterministic or not, is immaterial. Because even if history was completely deterministic, we would not be able to use that to predict perfectly the state of humanity hundreds of years in the future, because we know that human society is a chaotic system. It suffers — or enjoys — the sensitive dependence on initial conditions. And since you cannot measure initial conditions so precisely… Just think about what kind of apparatus you would have to have in order to…

Let's give an example using the climate. The reason climate is unpredictable is because we cannot measure precisely the temperatures and pressures across the Earth precisely enough to predict it months ahead. The reason is because your measuring apparatus will be larger than the Earth, and you still won't be able to get perfectly the initial conditions. So you're much better off putting your efforts into climate control rather than trying to measure it.

Why measure it? If you don't want to have rain on this particular date, or you want to have rain, you know how to make rain. The Russians fly aeroplanes and during the May 9th demonstrations if they want to have clear skies, then they just create a big patch of blue skies. So it's the physics behind that.

My major point is that prediction is overrated. If you're sitting on a condemned row and you know that you're going to be shot to death at the crack of dawn, you have perfect predictability — and it's completely useless predictability. You might rather want to know how you can escape that.

We want to understand things so that we can actually nudge it or even engineer the outcomes that we want. And essentially, this is the long term goal of cliodynamics. We are not there by any stretch of imagination, but we want to get to the point where we can actually use it to engineer better social outcomes than unfavourable outcomes. Everybody agrees you don't want to have a civil war, right? So in the future we will be able to use something like cliodynamics to prevent such bad outcomes.

WALKER: I have a couple of questions just on that, but I'll save them for the end because there's some other context that I think we should bring out first.

So a couple of questions on contingency. As you've said, Peter, cliodynamics focuses on groups, not individuals. And that's not because you don't think contingency is important. It's just it's difficult to know how to actually model it. But do you think it's in principle possible that you could one day somehow include the effect of remarkable individuals in the theoretical framework of cliodynamics? Or is it just naive to think that a fine-grained theory would ever be possible?

TURCHIN: It's an empirical question. And in fact, one of the next steps that we are doing, I can tell you more about the historical databases that we are building. One of them is CrisisDB. It's a database of past societies sliding into crises and emerging from them. Now we are approaching 200 cases and eventually there will be 300 or more. And one of the driving forces behind collecting a large number of case studies is that now we see that the entry into crisis is fairly channelised. It's like a ball rolling down a narrow valley. There's only one place for the ball to go. But once you get to the cusp of the crisis, a whole bunch of different trajectories open up. So see, I'm thinking as a dynamical scientist.

That's where we see a huge variability of outcomes, and that's why we need a lot of examples. First, what we have done is characterise them statistically, to find out what is the frequency of really bad outcomes, good outcomes, and what's in between.

But secondly, the next step that I want to pursue (assuming that I can get funding, because this all takes quite a lot of work), is to build into our database the role and characteristics of different leaders. So it seems likely that leaders are important at these cusps. We were talking about this earlier. This is the trajectory divergence region. A small push may result in the trajectory going either to positive or to really catastrophic outcomes.

And so the characteristics of leaders: that's the next interesting question that you can ask. What are the characteristics of leaders whose decisions lead to good outcomes and what are the characteristics of leaders whose decisions lead to catastrophic outcomes? Now, I don't know if we will find any signal in that data, but that's an empirical question and we intend to find out.

WALKER: Interesting. How do you think about the interplay between contingency and broad impersonal forces? So take World War I, for example — and we can quibble over the details of this example — but many historians argue that World War I was a highly contingent event. And then that contingent eventually sets the stage for all of these structural forces that arguably lead to World War II. In a way, you could argue that it's contingency all the way down. So how do you deal with that? Is your answer again, well, cliodynamics just looks at larger timescales and contingency can't really shape structures over those larger timescales?

TURCHIN: It can, of course. And so again, we are back to the question of the limits to predictability. And in the dynamical systems approach, we can incorporate such contingencies in a reasonably straightforward way. So the contingency itself, or the event that has caused the trajectory to turn into a very different direction, is not predictable. But once that happens, the trajectory starts running now in a more understandable and predictable way. So this is what you mean by contingency: contingent on this event. So this event itself perhaps is not predictable, but we can investigate the trajectories contingent on such events that resulted in macro changes. So that's one thing.

But the second thing: here is another metaphor from dynamical systems science that's useful. If you think about systems in chaotic regimes, they are typically found on a strange attractor, which could be a fairly low dimensional attractor. So if you kick the system in one of the sensitive places, then the next peak might disappear or vice versa. Instead of not having a peak, you would have a peak. So there will be a macro level event.

But after the trajectory goes back, it will be still on the same strange attractor. So you may have delayed, let's say, a breakdown of a political system — or advanced it — but you haven't really changed much of anything.

So my interpretation of World War I is that if Serbian nationalists didn't shoot the Archduke, then something else would have happened, probably within a year, maybe two, because we know that Germans were really worried about Russia. Russia had a miracle decade from the end of the revolution of 1907. (Well, it was only seven years.) Their economy was growing at unprecedented rates, very rapidly, and the German Staff was very worried about Russia catching up, and therefore they were getting ready to have a preventive war. And so if Gavrilo Princip didn't assassinate the Archduke somebody, something else would have come along and triggered things.

WALKER: Not to belabour this point, but do you think there's any way in which we can use long-term and average tendencies to predict what will happen in a particular time and place.

TURCHIN: In a statistical sense, yes. I'll give you an example. People have been very impressed that we had the summer of 2020, the huge riots, lots of people actually getting killed, and then January 6 of 2021, and now things seem to be quieting down; the elections of 2022 went reasonably without major surprises and things like that. So does it mean that we are over?

Here is where we can use the knowledge of statistical patterns to suggest that it is unlikely to be so. Because typically these periods of political and social turbulence tend to last for many years. Sometimes systems collapse, of course, and you have 100 years of fragmentation. But typically the mode is between ten and 20 years or so in the data that we have examined and there are some good reasons why. But anyway, right now, just taking that as a statistical result, it means that it is unlikely that our society is so different from previous societies that all the turbulence will be over in just one year.

That means that we are likely to see more turbulence during the 2020s. I am particularly worried about 2024.

WALKER: Because it's an election year.

TURCHIN: Because it's an election year in America and we have two candidates who are both now under legal proceedings. Lawfare is going back and forth, and the rhetoric continues to escalate. And judging by previous crises of past societies sliding into crisis, it takes time. There is some inertia before people become ready to use violence, start killing other people. And the heating up of rhetoric is a very telltale sign that this is heating up.

Now, I hope that I'm wrong, because I live in this country and I don't want to have a civil war here. I'm too old for those types of things. But unfortunately, we're talking about statistical patterns. Chances are that we have a few more years of turbulence ahead of us.

WALKER: And would you ever attach a specific probability to that, or would you just say verbally that it's likely or unlikely?

TURCHIN: We can. Of course this would be contingent. So saying that assuming that our society is not terribly different, let's say, from the previous societies, then here is the probability. We just take the empirical distribution of times and that gives us an estimate of what is likely to happen to us.

WALKER: In terms of a specific probability.

TURCHIN: Yeah. “X many of these crises were done in seven years, X many in 8, 9, 10, up to 20, 25 and so on.” And so this gives us an empirical estimate of the probability of the length that our crisis will take to resolve.

WALKER: Some questions about data or big data, or what you call the cliodynamic macroscope, before we move to structural demographic theory and End Times. You mentioned CrisisDB, and that's obviously part of Seshat, this incredible database that you and colleagues and a team of research assistants have assembled. So I'd love to ask a couple of questions about that. Firstly, how big is Seshat?

TURCHIN: Well, it depends how you measure it. And let me just explain that we started building this database more than ten years ago, and the first set of questions was to test theories about how large-scale complex societies evolved. Why does nearly 100% of humanity now live in large-scale societies, which are typical only of the last 5% of our evolutionary history? CrisisDB is now the next step to test theories about why complex societies periodically break down.

Now, back to classic Seshat. We have about 450 societies, and Seshat is spanning the past 10,000 years from quite small-scale societies, such as Neolithic cultures, all the way to states and great empires, and up to 1800 or so — this database is for premodern societies. And by the way, the number keeps growing, so we have another 150. So we will have 600 societies very soon, because we are adding that to the database.

But anyway, to go back to the original set, for them we have coded hundreds of variables, of which 160 are well-coded. So multiply 160 by 500, roughly, and you get some idea about how many data records there are. So “record” is the value of this variable for this society. But each Seshat record is like a pyramid. It has not only values, it also has some other stuff associated with it. So, for example, what's the certainty or uncertainty? Whether there is agreement or disagreement? What are the references? So altogether, it bloats up to hundreds of megabytes of information.

WALKER: Wow. I imagine it can get quite difficult trying to convert historical evidence into digitised data that can then be fed into a cliodynamic model. Are there any stories you could share around that?

TURCHIN: Well, just to say that it was a process. It turns out that some variables are easier to code. For example, one variable is: does this society have swords and what metal are they made of? So that turns out to be reasonably easy to conceptually determine. You may lack the data because, let's say, there is no writing and very poor archaeology. But at least, for example, if you have enough burials, and there are no swords in burials but other weapons are present, then we can conclude with high degree of probability that they did not have swords.

But other variables are much harder. One of our research projects was on understanding the evolution of moralising supernatural punishment. So why did religions like Christianity teach that people get rewarded or punished in the afterlife? Or why did Buddhism teach how you escape the cycle of this terrible life and things like that?

There are a variety of theories, and that turned out to be quite contentious. And our first attempt did not work very well. So we had to essentially go back to the drawing board, redesign the approach, and now it finally all got published about a year ago. So that's an example of where things were quite involved.

Just think about it. How do you code whether a religion is moralising or not? It's a hard question. It requires a lot of thinking. It requires very close work with experts who really understand those societies. But experts are not enough, because each expert needs to be able to understand what we mean by this, and most of them would not bother reading the definitions. And so that's why a member of the project has to work very closely with an expert to elicit the correct information.

WALKER: It got me wondering, how much of a problem is it that labels and conceptual categories can vary across time and cultures?

TURCHIN: Exactly. From the very beginning, our definitions of variables were designed and then refined in several cycles in such a way that they could be applicable both to Aztecs in Mexico, Chinese during the Bronze Age, French in the Middle Ages. So those definitions often had to be rewritten as we encountered new, different societies. And then that's why it was so much work, because then we had to go back and recode the data that we had already coded using the not-so-good definition.

WALKER: It's an impressive project.

TURCHIN: Thanks.

WALKER: This is the last question I had on data. I thought it was really interesting how the creative kind of proxies that you use. Obviously, past societies didn't always have big government agencies or private pollsters churning out yearly statistics. And so you have to get creative about how you estimate the variables you're interested in, like violence, population growth and decline, et cetera. There are a couple of examples we might just touch on. Roman coin hoards: how do they illustrate population decline?

TURCHIN: Yeah, so essentially some numismatists — people who study old coins — noticed that there is a correlation between times of trouble and the number of hoards that you find. That makes a lot of sense because coin hoards are typically used as the store of wealth, and then at some point this wealth, to be useful, has to be dug up and used. So if a coin hoard was not dug up, that means that something terrible happened to the person who knew where it was buried. So one possibility is that they simply got killed. Another one is that maybe they were driven into exile or enslaved or something. So all of those are the result of violence. And so when we see, in one year, 50 hordes, whereas ten years before there were only two or three (because accidents happen all the time)... When we make a curve of the frequency of hoards, those curves trace quite closely the periods of internal violence or catastrophic invasions.

Typically external wars, if they happen around the periphery of a large state, don't generate a huge amount of hoards, because there are no armies marauding through. But civil wars are the primary producer of coin hoards.

If you think about it, how do you quantify how severe a civil war was? Well, perhaps by the number of people killed. That seems to be a good metric. I mean, it's a horrible metric, but it's good for science. So the number of people who are killed has some kind of a relationship to how many people who had buried hoards got killed. And so in the relative sense, if the number of hordes increases tenfold, that suggests that there was a roughly tenfold increase in the death rate. That provides us with a quantitative proxy for the severity of civil wars.

WALKER: Another one is: data on height is a key measure of popular immiseration, which is a concept we'll discuss more generally shortly. This is kind of obvious, but can you just explain why height is such an important indicator of biological wellbeing?

TURCHIN: Yeah, so human height typically gets set by the early twenties, and after that, by the way, sadly we start shrinking. So I'm a little shorter than what I was 40 years ago.

WALKER: Well, you must have been very tall in your early twenties.

TURCHIN: Well, at the time, yes. But now, of course, because of [inaudible] and everything, there are lots of much taller people. Anyway. So there are two growth spurts. The first one is the first five years or so, and the next one is the teenage years (different between males and females), but it turns out that both are important in determining your terminal height.

Now, the variation between individuals in height is mostly genetic. But if you are looking at a population of the same genetic composition over time, then shrinking heights indicate times of immiseration — which could happen for a variety of reasons, often several reasons together. One of them is that people, children and teenagers, don't get enough to eat. The second one is that they get sick all the time (because an organism needs energy to fight sickness). Or they're overworked. All those measures of immiseration result in declining population heights.

It is remarkable how sensitive this indicator is. It, of course, mostly gives us information about the general population, because heights for the nobility and elites typically don't shrink. But then you sometimes see a five, seven centimetre difference between nobility and peasants. This is a measure of inequality that you can get from skeletal material.

By the way, all you have to use is a femur. If you have a femur — that's the big bone in your leg, upper leg — that is closely correlated with overall height. And femurs have a pretty high probability of surviving. So you can estimate how population stature, height, average height, increased or decreased from skeletons.

WALKER: So you take the length of the femur, you use a table of correspondences and then you just average out the heights of each generation in a particular region.

TURCHIN: That's right.

WALKER: There's this really remarkable fact in your book End Times about how one of the reasons we know why American workers fared so poorly in the 19th century is because the average height of native born Americans declined by 5 centimetres.

TURCHIN: Yeah, two inches.

WALKER: Which is a lot.

So another way bones can be used is to measure violence. Why is a high frequency of breaks on the left ulna (also known as the forearm) in skeletons good evidence of violence?

TURCHIN: Yeah. Well, that's because… I wish your listeners could see us — I could demonstrate it on you. No, kidding. Well, most people are right handed. And so if somebody hits you with a club, you throw your arms up — and since they are right handed, they will hit your left forearm.

The other one is, of course, an arrowhead stuck in bones or just sitting inside your chest cavity.

WALKER: That's usually a good sign of violence.

TURCHIN: Exactly.

WALKER: Okay. So let's talk about structural demographic theory. Now, just to put this in context, there are many other empirical regularities in history that you've looked at across your body of work, and they're all fascinating. Sadly, we probably won't get time to talk about them all today. But, for example, besides secular cycles, there is the fact that huge empires tend to rise on step frontiers. That's really interesting. There's the autocatalytic models of religious conversion. That's, again, fascinating. But because we're speaking mainly about End Times I figured today we'll just focus on structural demographic theory.

TURCHIN: Those are the things that I talk about in my other books.

WALKER: Correct.

TURCHIN: Just to make sure that there's no false advertisement here.

WALKER: Yeah, good point. So could you give a summary of structural demographic theory in general, and how you've refined theory yourself?

TURCHIN: Sure. So the first thing is that large-scale complex societies, organised as states, have been around for about 5000 years. And we now have enough data to show that they can experience long periods of internal peace and order (notice at the same time they could be fighting quite fierce wars outside, but we're talking about absence of internal wars), often lasting for centuries, sometimes shorter, sometimes longer.

But inevitably such integrative periods, as we call them, end, and they get into end times or disintegrative periods.

Why? The most common feature of societies in the pre-crisis period is what we call elite overproduction. So let me unpack that. First of all, who are the elites? Simply put, a small proportion of the population.

WALKER: Like 1 to 2%.

TURCHIN: 1 to 2%, that concentrate social power in their hands. So think about the proverbial 1% here in the United States (although things are a bit complicated; we can get back to that). Or the Mandarin class in Imperial China. Or military nobility in mediaeval France, and so on and so forth.

This is a very important point: typically there is no sharp boundary between elites and non-elites. It sort of grades.

So in the United States, wealth is the best marker for elite status. So you can think about lower rank elites, like in the top 10% of the wealth distribution, then you have 1%, and then you have 1% of 1%. And so obviously, the more wealth you have, the more power you have. And the same thing in the parallel political pyramid. Obviously, as you work your way down from the president to a lowly bureaucrat, the amount of power decreases. So that's one important thing.

But the second important thing is in the dynamics. So how are elites reproduced and recruited? Typically, there are always more elite wannabes — in the jargon, elite aspirants — who are vying for a limited number of elite positions.

And some competition for such positions is good, because it feeds out better people. But it turns out that as competition becomes too intense, once you have two, three, four times as many aspirants as the positions for them, that is a bad sign. In my book, I use the game of musical chairs, modified musical chairs, to explain this. I don't know if your Australian listeners know.

WALKER: Oh we know.

TURCHIN: Alright, you start with ten musical chairs and eleven contenders, right? And one loses. But then instead of removing a chair, we add to the number of players. So you start with eleven, then it’s 15, 20, 30, 40. You can imagine, if you just try to think what would happen in this situation. I predict that within like 10, 15 or 20 minutes, there would be fist fights, right?

Because some people will want to break the rules. There's always some breaking, and that rule-breaking will spread. And soon enough you would have violence. Unless you're playing this game in Canada, because Canadians don't fight.

WALKER: The rule-breaking spreads because competition and cooperation are like an unstable…

TURCHIN: So some competition is good, but too much competition is bad because that's what corrodes the rules of the game. Humans are not agents, mathematical agents in game theoretic models who cannot break rules. Humans, when they see that they are not getting ahead, they will start breaking rules. Somebody will, and then it spreads.

We saw this in real life during the elections of 2016, when there were 17 Republican candidates during the primaries, and one individual was very good at breaking rules and getting ahead in the game. And everybody actually started — not everybody — most other candidates also started breaking rules (but they were not quite as successful in doing that). And since then, actually even before then, the rule-breaking started to happen because these conditions of elite overproduction started developing in the United States about 20 or more years ago.

WALKER: Okay, let's take a digression further down this path of elite overproduction, and then we can come back to structural demographic theory holistically. So Peter, elite overproduction is probably the one idea of yours I've referenced most over the years. And it's like one of those things that once you understand it, you start to see it everywhere (maybe more than you should see it, but you just can’t unsee it).

TURCHIN: I see it everywhere, especially because we have it quite strongly developed in the States right now.

WALKER: Yeah, that's for sure. So it'd be interesting to discuss some specific examples. How does elite overproduction predict cancel culture?

TURCHIN: Right. So in the United States, the ruling class is the coalition of wealth holders and credential holders.Unless you have wealth or become a self-made wealthy person, the route to political office is pretty clear: you want to get a law degree. But if you don't want to become president, but you just want to escape precarity, for example get into the top 10%, then you also want an advanced degree. It could be a PhD, medical doctor, or several others.

So as a result of elite overproduction, we have too many individuals who aspire to getting ahead. And so they are all trying to get credentials that would increase their chances. And as part of this, what we see is that some strategic individuals, but maybe not very nice ones, start thinking ahead, and therefore they want to clear the ranks of competitors a little bit.

And so how do you do that? In the old times — and we actually do see this, both in the 19th century and in the 17th century crises — the elite aspirants would have duels and kill each other using swords, pistols, or whatever.

Nowadays, we are more civilised than that. So it's character assassination that works well. It's an ugly side of things, but you want to attack both the established elites — so take professors, for example, because when a professor is fired, an extra place frees up — but also your competitors. So you want to clear the ranks and increase your own chances. And of course, it’s not necessarily that everybody who does this is consciously following this strategy. First of all, it could be more on a subconscious level. But secondly, once this game starts, once this elite overproduction game goes on, the norms of attacking competitors spread.

And so then many people actually might do it in self defence before they get accused. So this is the dynamic that results in the explosion of cancel culture.

WALKER: Right.

TURCHIN: So think about cancel culture as like duelling culture in previous, more brutal times.

WALKER: Yeah. So to put it back into the musical chairs metaphor, if I can get someone cancelled, that removes them from the game and makes it more likely that I'll get a seat.

TURCHIN: Exactly.

WALKER: Could we view the replication crisis in psychology as a consequence of intra-elite competition? [Note from Joe: By this I meant could we view the production of fraud, not the exposing of it, as a result of intra-elite competition.]

TURCHIN: But only partly. Partly this is the way that science advances — by critiquing previous approaches that did not work very well. And as a result of that, some of it is normal scientific process. In science, criticism is very important. To be effective at producing good science, you critique ideas, data, methods, but not people. Now, as a result of this cancelling culture, it spread to attacking people, ad hominem attacks. And that's the bad side of these critiques.

So the crisis of psychology, it probably would have happened anyway, and it had a positive effect on the quality of science, but as long as it’s kept from ad hominem attacks.

WALKER: Is intra-elite competition fractal? So if the proportion of elite aspirants to elites gets out of control, and there's a lot of competition between the 1%, is that also reflected in the 0.1%, 0.01%, and so on? Or is there like some threshold at which the competition ceases?

TURCHIN: No, and we just saw a great example of that, with Elon Musk and Zuckerberg. Now, I doubt it will ever come to pass, but they are making serious-sounding noises that they want to fight each other in the cage. So no, it's like with the turtles: all the way down, basically.

WALKER: If that fight does come to pass, who are you putting your money on?

TURCHIN: Cliodynamics does not have an insight on this. And it doesn't matter, because just the fact that they're fighting is a sign of competition heating up.

WALKER: How is the degree of polygamy among elites connected to elite overproduction?

TURCHIN: Yeah, that's one of the very interesting, very robust results from our analysis. I mentioned that complex societies go through these integrative phases, which are of variable length, and it turns out that in polygamous societies, the integrative phases are much shorter. Why?

This is actually a result which was noticed back in the 14th century by the great Arab historian and sociologist (I'm not afraid to name him that way) Ibn Khaldun. Ibn Khaldun noticed that dynasties in Maghreb, North Africa where he lived, tend to last for only three or four generations. So that would be 75 to 100 years, and then they would be replaced by another group coming typically from outside of this region along the Mediterranean border.

Why? The reason is that if you have polygamous elites, that means that they produce children at a much more rapid pace. So think about bin Laden, for example, who has like 100 brothers or siblings or whatever. So that's a very powerful engine for driving elite overproduction up very rapidly.

WALKER: Yeah, it's fascinating. It made me wonder, is it in some sense possible that the Western practice of rich people having less kids has been culturally selected for at the group level because it slows the rate of elite production and so makes those societies more stable?

TURCHIN: So, in fact, the number of children among the elites, even top elites, is quite variable. It changes with time. And so this is a separate topic. But certainly, I argue in my work, that monogamy spread as a result of cultural selection. Not only I. People like Joe Henrich, for example, also make this argument. And the reason is that polygamy generally is associated with negative effects at the society level. First of all, you run much shorter integrative phases, but also there are plenty of other things. In modern societies (we have good data), the crime rates, murder rates, for example, are much higher in polygamous societies. There are many negative effects.

This is what is sometimes known as selection by consequences. As far as we know, monogamy really was invented only once, in the Mediterranean, amongst the Romans and Greeks.

And it spread from there to the rest of the world. So most recently, Turkey, for example, about 100 years ago, switched from polygamy to monogamy, even though they stayed a Muslim country. Japan is the same way. China. So all those societies were formerly polygamous, but then they switched to monogamy. And the most likely reason is that people there, or elites there — thought leaders — realised that switching to monogamy makes the society more cohesive, more cooperative, and better to compete against other societies.

WALKER: Right. Why are lawyers so dangerous?

TURCHIN: Well, it turns out that lawyers are the most common profession amongst revolutionaries. Lenin was a lawyer. Castro. Robespierre.

WALKER: Lincoln.

TURCHIN: Lincoln. Gandhi. Gandhi was not a revolutionary, but he was certainly a very influential agent of chance.

In the United States specifically, as I mentioned earlier, if you want to get into political office, you want to get a law degree. And by the way, the best law degree apparently is from Yale Law School. In fact, Yale Law School produces both people who are very successful, but also counter-elites, those elite aspirants who are frustrated and then turn away.

WALKER: Like Stewart Rhodes.

TURCHIN: Stewart Rhodes. Exactly. The founder and the leader of the Oath Keepers. He got a Yale law degree. And several other populist politicians.

Now, again, taking the case of the United States, we have a horrible overproduction of lawyers. There are three times as many people getting law degrees as there are positions for them. And as a result of that we see a really bizarre distribution of salaries that newly minted lawyers receive.

I talk about this in my book. There is one quarter who get really huge salaries, close to $200,000. Then more than half get between 50 and 70,000, which is not enough to even pay off the debts that you have. And nobody in between.

WALKER: Bimodal.

TURCHIN: Yeah, it's bimodal. So this means that we know who gets those chairs and those who don't get the chairs. And so of those who don't get the chairs, they are ambitious, typically very smart, well organised, networked, energetic. And so the more of them are frustrated in their ambitions, the more turn to starting breaking rules and starting revolutionary movements.

Now things are getting even worse because it turns out that GPT-4 already can automate 45% of what lawyers do. So instead of three to one, we soon will have six to one.

WALKER: I hadn't updated on that, actually. That's a good point.

TURCHIN: Lawyers is the second profession, after secretary types, whose work is going to be automated massively. And as a result, it's going to be bad news unless you figure out where to put that energy into a productive manner.

WALKER: So many reasons to be pessimistic. As a side note, in terms of the significance of Yale Law School, I assume it's just a selection effect, where it's the most prestigious law school, so that's where the elite aspirants choose to go. But I was curious whether you had ever actually looked into the law school and its curriculum specifically, to see whether maybe there's something going on.

TURCHIN: I have not. I don't know why it's Yale, because it could be Princeton or Columbia or whatever, but it's an interesting question, but I don't know the answer to it.

WALKER: In 2020, I actually tried to find some data on lawyers in the Australian context, just in a very amateurish kind of way. I was just curious. I couldn't really find anything. But there was this study that Urbis did, where they looked at the number of solicitors practising in Australia nationally, and that number had increased by about a third between 2011 and 2018, whereas the general population had grown by about half that rate.

TURCHIN: That's right, yeah. So that would be interesting. Same thing we see in England in the run up to the Civil War of the 17th century. We see the great overproduction of Oxford, Cambridge graduates, and they had the third degree, which was basically a law degree to get to be a solicitor. I forget what it's called.

WALKER: And the reason you know that, to bring this back to data — and I think Jack Goldstone wrote about this in Revolution and Rebellion in the Early Modern World, which is sitting on your shelf behind me…

TURCHIN: He's the one who found this factoid.

WALKER: Yes, and the reason is you can look at the degrees, measure the credentials. And there was this explosion in enrollments at Cambridge and Oxford, which reached a peak in 1640, just on the eve of the Great Revolution, and then declined back to pre-1600 levels by the middle of the 18th century.

TURCHIN: That's right. So this is the race for credentials. And that's why it's a good proxy for elite overproduction.

WALKER: Yes. Just on Australia, quickly, have you ever looked into any Australian data generally?

TURCHIN: No, because keep in mind that getting all those studies, that's a lot of work, many months of work or sometimes even years.

I have just published a blog post where we invite other people to start collecting such data. We published a methodology article for them to use as a guideline for data collection.

WALKER: Okay. So last question on elite overproduction, then we can come back to structural demographic theory more broadly. I was wondering, to what extent can a solution be to just increase the elasticity of the supply of elite positions [i.e. the elasticity of the demand for elites]? So you could think about this at an institutional level — and this already exists to an extent, where the number of seats in Parliament or Congress just mechanically increases with population size. Maybe we want to change the ratio or something so that they're even more elastic. Or you could think about it at a technological level, so we could create outlets for elite frustration. So one example might be social media. Tyler Cowen argued in one of his earlier books What Price Fame? that fame remains positive-sum at its current margins. So you could let more and more elites just get famous and for quite some time, it won't become a zero-sum competition.

What do you think about that idea?

TURCHIN: Well, yeah. Keep in mind that people who are following the credential route, many of them don't necessarily want to become president or prime minister. They just want to get out of precarity. So one way to choke off that supply is to get rid of immiseration. This is something that you want to talk about next, I believe.

Because for the majority of the population in the United States their well being has been declining, that's one of the push factors that people want to get the credentials. And as a result of that, so many people now go into college that the college premium has been shrinking and now is essentially zero.

But there are many other things. So, for example, I'm very partial to historians. I want to have more historians around. They may not like cliodynamics, but that's fine because just by being historians they're churning out all the data that we want. So why don't we cut in half the horrendous budget for the military that we have in the United States, right, nearly a trillion dollars, and give some, even a one-tenth part of that savings, to just hire historians to give them stipends or something. So all you have to do is publish good work, hopefully more numbers for us, but whatever.

Or maybe Elon Musk is right that we need to go to the planets and so provide an outlet for some ambitious people to apply their energies elsewhere.

So in principle, once we start thinking about it, we don't want to increase the number of senators or something like that. That number should be really set by what is the optimum number for governing a country. But what we do want is to provide outlets for the energies of young people to have a meaningful life and to make meaningful change, positive change. And that is one of the reasons why we have so many difficulties: because our societies have failed to expand the opportunities for bright, energetic and ambitious people to apply themselves.

It doesn't have to be that they would get more power. Meaning in life could be achieved in other ways.

WALKER: Yeah, that's interesting. Have you heard of the effective altruism movement?

TURCHIN: Yeah.

WALKER: That seems to be an outlet for very talented people to seek meaning and status within a certain community that doesn't necessarily rely on wealth or income.

Actually, that just raises a more general question, which is what kind of cultural or social innovations could we create to provide that outlet?

So structural demographic theory strongly relies on the Iron Law of Oligarchy…

TURCHIN: Let's talk about that a little bit, because I want to tie it to the question of…

WALKER: Should we do popular immiseration?

TURCHIN: Yeah, but I can wrap it in one sort of package. So the question is that obviously elite reproduction is something that develops at some times, but not others, right? Because we have those integrative periods. So the question is, why does elite overproduction develop?

Well, the reason is that — let me just compress the long story into just a set of theses — if once a society has run for several generations enjoying internal peace and order, the elites, the ruling class, tend to assume that's an automatic thing, does not need to be nurtured. And they are tempted to reconfigure the economy in ways which would work not for everybody's benefit, but for their own benefit. And they can do it because they have power. So this is the Iron Law of Oligarchy.

And this has three bad consequences. So, first of all, this results in immiseration. Call it the wealth pump. It's the perverse wealth pump that takes from the poor and gives it to the rich. And there are many ways to do it, but for example, in the United States, by not increasing the minimum wage, by taking away the power of workers to organise and bargain with employers, and also by decreasing taxes on themselves, that's how the elites turned the wealth pump on in the States in the 1970s.

But this is a typical thing. This happens in mediaeval France, for example, or Rome, and so on and so forth, So then, first of all, this creates immiseration. The quality of life for the majority of the population declines and that drives their discontent and what we call mass mobilisation potential. So that's one force undermining stability.

Secondly, it results in overproduction of people with wealth, and many of those decide to go into the political arena. And so now you have the game of aspirant chairs, because in the United States, for example, the number of decamillionaires (people with $10 million or more of wealth) increased tenfold over the past 40 years. And so that created many more aspirants for positions in politics, such as Donald Trump, of course, but also Michael Bloomberg, or the failed ones, like Steve Forbes, for example, and many more. And some of them run themselves and others run candidates. So we have overcrowding. So that's the second problem.

The third problem is that by increasing immiseration, we now create another pump that essentially induces ambitious and energetic people from the immiserated class to try to get out of it, which drives the credential revolution, because that's how you get out. And so that creates overproduction of people seeking credentials.

WALKER: To escape.

TURCHIN: Right. And so as a result of that, we have overproduction of the wealthy people, overproduction of credentials, and they are the ones who eventually bring an end time to their societies.

So this is how the wealth pump, immiseration, and elite overproduction are connected at the dynamical level.

WALKER: Yes. So one of the key observations in that model is that crises aren't caused by the popular masses revolting. They're caused by the counter-elites who then mobilise and co-opt the masses. Is the reason that the masses don't initiate revolutions that they are, I guess, less talented than the elites? Or is it simply that because it's a larger group, it makes collective action difficult if not impossible?

TURCHIN: It's organisation. Why do we need elites at all, by the way? Because human society, in order to function properly, needs organisation. That's why we are organised as states, or in a business as firms.

Now, the commoners are not organised. Think about the Jacquerie, that very bloody peasant rebellion in France in the 14th century. They would happily have a revolution and overthrow nobles. In fact, they killed quite a lot of them. But then as soon as the first organised and well armed group of knights appeared on the horizon, they just rode them down and killed them off and dispersed the rest. The difference is in organisation, but also elites have better armour and weapons and things like that.

And it's the same thing nowadays. Social power means the capacity to organise. When elites are united and the state is strong, popular uprisings happen, but they are very ineffective and they result in a lot of bloodshed for the peasants themselves.

WALKER: Yeah, I'm curious about narratives that elites use to justify and defend popular immiseration to the workers themselves. So if the narrative used in the post 1970 period, in a word, was meritocracy — the idea that, hey, don't get envious or resentful, just work hard and you can be like us too — what was the narrative during the first Gilded Age or at other points during disintegrative phases? Are there any examples you can share?

TURCHIN: Sure. Details of the ideologies change, but the end result is the same. So in the Gilded Age, that was social Darwinism. Some people were essentially genetically worthy or equipped to be leaders and rich and so on and so forth.

Before that, during the 17th century, it was God, basically. In many Protestant versions of the religion, some people were preordained to be successful and others were not.

In the Middle Ages, the nobility said, “Because I had ten generations of ancestors, therefore I'm deserving to live better than you.”

But the end result is always the same. It's the elites justifying inequality, essentially.

WALKER: I see. Okay, so I think I've got about 25 questions left, but I'll pick the five best ones, and we'll try and get through them. So in 2010, in this now-famous Nature article, you predicted that the next decade is likely to be a period of growing instability in the United States and Western Europe. That prediction played out, obviously. We had the year from hell in 2020 in America. Question is, if the pandemic hadn't happened, do you think the prediction still would have played out?

TURCHIN: Possibly. It would still play out, but the timing could have been delayed. That's the most likely thing. So epidemic was one of those triggers, but on the other hand, remember that the distribution of those triggers tends to be much more frequent during the times of trouble. In fact, there is a very close correlation between end times and epidemics. They're much more likely during those disintegrative periods.

WALKER: Right. Makes sense, because people can't cooperate to prevent the spread of…

TURCHIN: Typically the wellbeing goes down, which makes people more susceptible to disease. Then you have globalizations, because wealthy people drive trade, and disease moves along the trade routes. So there is a variety of reasons why diseases tend to happen during those periods.

WALKER: Ah, I see. So before we move on to what we can do about all of this, is there anything else you'd say on either structural demographic theory generally or the trends that you've witnessed in the US specifically?

TURCHIN: No, I think because we have limited time, let's address those questions.

WALKER: Yeah. So at the beginning of the conversation, you mentioned that you've studied about 300 crises, and some have fairly benign endings, some have disastrous endings. Of the more optimistic cases you've looked at, do you know what caused those good endings, and to what extent are those near misses the result of individual agency or kind of heroic leaders versus maybe they could be the result of structural forces that you just haven't detected yet?

TURCHIN: Well, it's both. So let's take the chartist period in the 19th century British Empire. There were definitely some prosocial leaders, and there were some antisocial leaders. So like Duke Wellington, for example, the general who led British troops at Waterloo, he was a very conservative leader, and had a huge amount of power and status because of this. And so until he was out of the picture, really, no reforms could go forward. So here we have an individual who had a negative effect.

But also there were some structural things. The British Empire was huge. So, first of all, they shipped millions of people to places like Australia. There was also immigration to North America, and that relieved the labour oversupply and removed one of the engines driving the wealth pump. Secondly, they also shipped quite a bunch of surplus elites to positions in the empire. But those were all temporary mechanisms. This sort of worked to flatten the curve, if you know what I mean. It gave more time for the elites to get rid of non-cooperating individuals like Wellington, to address the deep causes of their problems.

WALKER: Of the crises that ended badly, how many of those do you think — and you can answer this very roughly — but how many of those could have been averted had the elites at those times understood your theory?

TURCHIN: Now, this is a difficult question, because most of these end times end up badly. Let's say you make me the Tsar in late 16th century in Russia, even though I understand now perfectly well what was going wrong, I would be unlikely to be able to do anything. Because persuading people that I know what the problem is, they would cut my head off before I would get very far. So my guess is that there were many examples of good prosocial leaders who in fact understood, at least intuitively, the problem, but they just couldn't get enough other people to cooperate with them. And so the whole thing collapsed.

WALKER: I guess there's also this problem of things like hyperbolic discounting where elites can prioritise their short term material interests over the long-term risk of going down with the ship.

TURCHIN: That's right.

WALKER: Have you thought about reflexivity? So you mentioned that the end goal of cliodynamics is to presumably transfer out of the ivory tower and persuade politicians and public policy to take its ideas seriously, to have a positive influence in the world. So if it is successful, if that does happen, I can see a couple of ways it could go. Maybe one is that people take the theory seriously; they see the leading indicators for a disintegrative phase and then they try to preempt it. Or alternatively, maybe they view the model as somehow inevitable and that reinforces it. So how do you think about how those dynamics feed back into the model? And how would you actually model those dynamics?

TURCHIN: Well, we already have a prototype — it's a published article — where I run forward the trajectories, and we can look at what sort of nudges and changes need to be done. This is a prototype, so should not be taken seriously, just to indicate where we need to put more research effort to develop it into a more full social engineering problem.

So I'm actually an optimist by nature. This time around, we missed the opportunity to head off the crisis. But next time the crisis comes around, by that point, I think that we will have much better theory and it will be possible to use it in a way to essentially get rid of those end times.

WALKER: Okay, three final questions. When did you read War and Peace? And was it compulsory reading in the Russian school system?

TURCHIN: Yeah, but I read it even before. I actually read it four times.

WALKER: Wow.

TURCHIN: Yeah. The first time before. Then for the class. Then I read it again for enjoyment. And the fourth time I read it more recently when I was writing my books, Ages of Discord and Ultrasociety.

WALKER: Okay, so to what extent has your view on history been influenced by Tolstoy.

TURCHIN: To some degree, yeah. I don't take everything that he said 100%. But several of his ideas have been quite influential for me. So mainly his idea, which is similar to Isaac Asimov's idea in the Foundation series, that you can make a lot of progress understanding the dynamics of societies by ignoring individuals and focusing on macro level and meso level dynamics.

WALKER: Right. And then I guess maybe his second idea to influence you is his version of asabiyyah.

TURCHIN: Yeah.

WALKER: Who has been your greatest intellectual influence?

TURCHIN: My father.

WALKER: Tell me about him.

TURCHIN: He was a physicist by education, but then he also, like myself (or rather me like him), switched more into cybernetics, fairly rarefied computer science, theoretical computer science. And he was a very remarkable individual, and he had a huge influence on my thinking.

WALKER: Final question. Complexity science applied to history reminds us that the veneer of civilization is thin and that seemingly stable civilizations, governments, can kind of collapse overnight. As we think about the American situation, what's your favourite example — or maybe you have a couple — from history, of revolutions or collapses that happened almost overnight.

TURCHIN: Well, they usually don't happen overnight.

WALKER: Not literally, yeah.

TURCHIN: There is some inertia. People need some time to psych themselves up for violence, at least normal people, not assassins or somebody like that. And so it typically takes days, weeks, sometimes months.

But yeah, since we are in the United States, we should think about 1860. In fact, the 1850s was a period when violence kept going up and up, and so we had Bloody Kansas and several other incidents, and no Americans really believed that what would happen in the next, you know, five years is that 600,000 people would get killed and a lot of real estate destroyed. And so when in South Carolina they attacked Fort Sumter, they clearly did not think that they would be completely devastated by this thing. This is the law of unintended consequences.

And so I agree with you that complex societies are very fragile. Most people who don't study history don't understand how fragile they are. Everybody thinks that this time is going to be different.

In fact, it's hard for me to imagine civil war in the United States. But it was hard for Americans in the 1850s to imagine civil war. Just because you can't imagine it doesn't mean it's not going to happen. And I'm not saying that it's 100% going to happen. But the probability is more than zero.

WALKER: Peter Turchin, I think I've got through about half of my questions, so we'll have to do this again sometime. But this has been such an interesting conversation. Thank you so much.

TURCHIN: My pleasure.