110 | Neil Johnson on Complexity, Conflict, and Infodemiology

Physicists have traditionally simplified systems as much as possible, in order to shed light on fundamental properties. But small, simple parts build up into large, complex wholes. Are there new rules and laws of nature that apply specifically to the realm of complexity? This has been a popular question for a few decades now, and we have some answers but not as many as we would like. Neil Johnson is an expert on complex systems generally, and information networks in particular. We discuss how self-organization can arise from individual units following their own agendas, and how we can mathematically characterize such behavior. Then we talk about information networks in the modern world, including how they have been used to spread disinformation and find recruits for radical fringe groups.

Support Mindscape on Patreon.

Neil Johnson received his Ph.D. in physics from Harvard University. He is currently professor of physics at George Washington University, where he heads an initiative in Complexity and Data Science. In 1999 he presented the annual Christmas Lectures at the Royal Institution in London. He was the recipient of the Burton Award from the American Physical Society in 2018. Among his books are the textbook Financial Market Complexity and the trade book Simply Complexity.

• Web page
• Google Scholar publications
• Wikipedia
• Amazon author page
• Lecture on Complexity in Human Activity

[accordion clicktoclose=”true”][accordion-item tag=”p” state=closed title=”Click to Show Episode Transcript”]Click above to close.

0:00:00 Sean Carroll: Hello, everyone. Welcome to the Mindscape Podcast. I’m your host, Sean Carroll. One of our long-standing interests here at Mindscape is in the idea of complexity, not just the simple systems that particle physicists or cosmologists like myself like to study, but the interesting phenomena that come into existence when things have many, many, many moving parts. And you can find emergent behaviour that might not have been immediately obvious just from thinking about the component parts. Now we’ve talked about complexity, we talked about it recently with Scott Aaronson, but that was in the sense of computational complexity, how difficult is it to solve a certain well-posed puzzle. Here we’re gonna be talking about complex systems, so either physical systems that you could build up, like a robot, or for that matter, a human being, but also social systems, information systems. The internet itself is a complex system. So Neil Johnson is one of the experts in this area, an incredibly prolific physicist who heads up a new initiative at George Washington University in complexity in data science.

0:01:00 SC: And our conversation has two different focuses. The first part of the talk will be about, what is a complex system? What makes something complex? How do you know? How do you characterise it? So various words that you might have heard before, like power law behaviour, will appear. And Neil explains very, very clearly how power laws are different than bell curves or other things that you might run across in the natural world. But then we turn our attention to this specific idea of information systems, and how people get information in the real world. A lot of Neil’s recent research interests are on how extremism or conflict or various forms of bad behaviour spread through complex networks. How do extremist groups recruit their volunteers? How does disinformation promulgate through social networks or through official news channels? So this is, as you might guess, scarily relevant stuff to the modern world. He’s a great explainer of things. This stuff is really important and fascinating at the fundamental level. I think we’re gonna learn a lot, so let’s go.

[music]

0:02:20 SC: Neil Johnson, welcome to Mindscape Podcast.

0:02:22 Neil Johnson: Great to be here with you, Sean.

0:02:24 SC: We’ve talked about complexity in some vague sense on the podcast before, most recently with Scott Aaronson, where we talked about computational complexity, P versus NP kind of things, but your kind of complexity, you’re interested in complex systems, which is a whole another kettle of fish. Why don’t you start us off by telling us what you think of are the defining features when we talk about a complex system?

0:02:47 NJ: Yeah, that’s a great question. And it’s very important to differentiate between something that’s complex and something that’s complicated. Our lives are now complicated because we have to use all these kind of online platforms. We don’t know if we’re gonna teach online or not. We don’t… There are complicated issues, but complexity is something more than that. And to give an idea, let’s just say, let’s imagine we’re doing something complicated tonight. We’re gonna make a pizza, but we’ve also gotta pick up one of the kids. And so we gotta schedule these things. That’s complicated, but imagine if each one of those pieces now depends in some really complicated way on what’s happened before, and there are other people doing things at the same time that impact us. So for example, if going out in the traffic, well it matters that other people, who may have decided to go out in the traffic with us, are now… And they’re now congesting the road, and therefore affecting my chain of events.

0:03:56 NJ: So that begins the move from complicated to complex, because there’s nothing that I could actually interconnect to the other meaning of complexity. There’s nothing easy for me to compute that would make the task easy… That would kind of predict what would happen other than me going out and actually doing it. So that’s how it connects to the idea of complexity in a computational sense, it’s sort of like the length of the code that you need, you can’t kind of simplify it. So that carries over to the complex systems view, in the sense that it’s almost like all the things that could happen in the system, feedback across scales, one local things affecting other thing, the driver next to me affecting me, but also the fact that it’s…

0:04:42 NJ: That all the drivers have come out at the same time for some unpredictable reason. All of those things interact and make the system complex. Now out of complex, out of that complexity, therefore, surprising things can happen. They can suddenly emerge. Traffic jams can emerge. Stock market crashes can emerge out of nothing. And that is really where the interesting complex systems comes to the fore.

0:05:09 SC: Yeah, so I’ve been doing these videos recently called The Biggest Ideas in the Universe on some ideas in physics, and I’ve been pushing the idea that physics makes progress via a spherical cow philosophy, where we ignore all the complications and simplify down to the bare bones, and so in some sense, what you’re saying is that complex systems are exactly where that doesn’t work, where all of the interactions between all the different pieces really do feed into the final result.

0:05:36 NJ: Correct. And some of the things that we, as physicists, love to do, like we assume that interactions don’t depend on time, positions can depend on time, but the form of the interaction doesn’t depend on time, and it doesn’t particularly depend on maybe whether that spherical cow has some kind of compass needle inside that makes it different on a Monday morning than on a Friday afternoon. So those simplifications that enable us to get so far in physics, they don’t work.

0:06:08 SC: Yeah, but nevertheless, like you just said, there are things we can say, so that’s where complex systems become interesting, where it’s not just, “Well, things are harder so we give up.” It’s, “There’s a new kind of thing we can start looking for.”

0:06:22 NJ: Correct, and the challenge is finding that thing, and I’ve been scratching around in my career, and a lot of people are doing a lot of important work in this area. We’re trying to get… You’d love to have something even approximate, but general in the sense that it’s reproducible in these systems. So that’s a real challenge, these systems that we’re talking about. The traffic, there’s no conserved energy. There’s no sense in which it’s the same at midnight as it is at 6:00 PM. So somehow, all these trends, all these symmetries that would have helped us in the problem are just not there.

0:07:08 SC: Is it safe to say that complex systems research is still in a pre-paradigmatic stage, where we haven’t decided what are the crucial questions or the basic systems we can keep returning to?

0:07:24 NJ: Yeah, I think that’s a great way of summing it up. Every semester, when I go in and I want to teach a class on complex systems, but what’s the homework for the complex… What’s the simple examples that they can work out that show complexity, and yet are simple enough in a homework that nobody’s gonna turn around and say, “Hey, that was… That homework couldn’t have even been done”? That’s where we are, and so it’s funny because it’s like a double-edged sword. There’s so much to explore, but it can look from the outside that there’s… Where’s your Schrödinger equation? Where’s your Hamiltonian? You have nothing. You have nothing to show. And so yeah, it’s a challenge. It’s certainly a challenge.

0:08:14 SC: But nevertheless, I’m sure you have some favourite examples that you do like to lay on people.

0:08:19 NJ: Yeah, I particularly got interested in this famous example of the El Farol problem, which I think we can…

0:08:31 SC: By the way, this is very famous within the complexity community. No one else has ever heard of it, so you should definitely share it with us.

0:08:37 NJ: Exactly, exactly. And I was in… So I’ll lead into the story, but the El Farol is a bar in Santa Fe, New Mexico. And I happened to be there recently before the lockdown, etcetera, and I was taking photos of myself against this El Farol Bar, and they came out and, “What are you doing? Who are you?” And of course, as soon as I speak with my accent, it’s like, “What are you doing? What exactly are you doing? Is this something of concern?” But of course, what I was really doing was taking a picture of a bar that has become famous in the complexity field, and in some sense, it was John Cassidy who said, “Solve the El Farol problem, and you solve complexity.” So let me maybe just explain what that El Farol problem is. It really embodies everything to do with complexity. Here’s the situation, and I’ll vary it a bit from Brian Arthur. Brian Arthur introduced it. Brian Arthur is at the Santa Fe Institute, and he introduced the following problem. I’ll vary some of the details, but it’s essentially the same.

0:09:39 NJ: So imagine there is some bar, let me call it, for example, El Farol, and you want to go there a particular night a week, say Tuesday nights. You just happen to want to go there, there’s a band playing that you like, you just wanna see them every Tuesday night. You’re not the only one. So there’s some other number of people who probably want to go there as well. Now here comes the problem, that the El Farol Bar, like many bars, only has a finite number of seats, and you want to sit, and so does everybody else.

0:10:14 NJ: So one can imagine the situation that every Tuesday rolls around, the problem will be, “Do I go or don’t I go?” Now if I go, I’ve gotta get in the car, I’ve gotta go there, I’ve gotta go through all that effort, and I might have to stand up. So I would lose in that situation. If I stay at home and I hear that, actually, there were enough seats. Maybe my friend tells me the next day, “Hey, you should have come to El Farol. There were plenty of seats,” then I’ve lost as well. And of course the flip side of that, there’s two ways to win. I can make all the effort and go, and I find a seat, I win. Or I stay at home, and then my friend says to me the next night, “Yeah, you were right not to come,” I win. And so there’s immediately… That sounds like a game, and so, “Aha! That’s not physics, that’s game theory.” No, because there is some indeterminate number of people, say 100, that I don’t know, I have no connection with, and they’re trying to do the same thing.

0:11:11 NJ: So imagine we all hit the same strategy, we all have the same… As we become identical objects, so we become rational in some way, so we all do the same thing. That means we either all go, or we don’t. It doesn’t actually matter what individual rule we follow. We’re either all going, or we’re not. And of course, that immediately becomes the wrong thing to do. And so Brian Arthur very cleverly, with this one simple example, set up the idea that, “I therefore have a system of objects that are heterogeneous, that don’t have full information about the system. Maybe they know the last few weeks, whether it would have been good to go or not, and they’ve gotta make some decision about what to do the next time. They’re all trying to do the same thing. If they all do use the same strategy, it becomes a losing strategy.”

0:12:09 NJ: So there’s this internal frustration. We begin to be… Anyone who knows… If you know… The people who know things about spin systems, immediately you think, “Okay, it sounds like there’s a whole bunch of metastable states, that this thing is gonna bounce around… ” And that’s the whole point. And…

0:12:26 SC: But by the way, it’s worth noticing that there is something subtle here because for the paradigmatic question about complexity, this is a really simple problem. The problem is not complex. You go to the bar or you don’t, that’s the question.

0:12:41 NJ: Correct, and that is the… You absolutely hit the nail on the head. It sounds so simple, and there are so many other examples you could think of in the world. Do I go to the supermarket? Particularly now, I don’t want it to be crowded in the sense that I may have to wait in line to get in. So do I make the effort, or don’t I? The traffic. I’ve got two routes to go home. Do I take this route or the other route? The market. Do I buy, or do I sell? It matters, not that there’s some intrinsic value in the… It matters whether everyone else is buying… Same with housing market. Do I buy when everyone else is buying? Bad idea. So you get the idea that there’s some very… And the dynamics that come out of this model, and people have studied this. The physics community have studied this a lot now since the ’90s.

0:13:37 NJ: The dynamics are so rich, and they go beyond any physical system that we can really think of because I, as one particle… That’s what I love about the problem, I could think of myself as a particle, what would I do? And of course, I take… I probably remember something about the last three weeks, and then I’ve got some kind of little rule in my head that says if that happened, then do this. And we’re all probably doing something like that. It’s a wonderful… So it’s non-Markovian, it depends on the past, depends on the history, and there’s definitely feedback. I will adapt. That’s the other thing. We’re particles that adapt. So we’re heterogeneous, we have memory of the past, and we adapt. So if it doesn’t work, I’m gonna change my system, and probably to a worse one, but I will change.

0:14:31 SC: And I guess… Probably to a lot of people who hear this problem, they’re gonna say, “Well, isn’t the answer just like, go 60% of the time randomly? Why isn’t the answer that simple?”

0:14:43 NJ: Yeah, so we could do it whereby I have some kind of loaded dice with a 10 on it, 10 faces, and I roll it. So when it’s the first six numbers, that’s my 60%. So I would go then, or I would stay behind. So that would be a probabilistic way of solving it. But of course, I’m not gonna sit around and wait for that. That’s not the way I roll, that’s not the way I work. I’m not a random object. I take… I have memory of the past, I will try and adapt. I’m trying to win. And so this is where it gets really interesting, and you could…

0:15:21 NJ: But think of it connecting to things in smart materials where you do embed kind of this agent behaviour, or even physics itself, that somehow the particle may be trying to optimise something, or certainly minimise its energy, or do something like this. And if it has some kind of classiness, some memory of the past, it’s not gonna just be the equivalent of flipping coins. And so this is where we’ve… The physics community actually… So this El Farol problem is what really got me interested in complexity. And I naively thought that the… For example, the economics and finance field would be absolutely… They would love this problem, but still nearly 20 years later, it really… They haven’t really embraced what Brian Arthur was really talking about in terms of this inductive behaviour, rather than absolutely being able to compute the ground state, and then sticking with it.

0:16:35 NJ: And I think it is because it doesn’t really fit in with any of their paradigms. There is no rational actor. You can’t even perturb the rational actor, there’s no equilibrium… There may be an equilibrium of sorts, but it’s really bouncing around between metastable states. So it’s much closer to what a physicist would like to hear about than what are the… Some of the other disciplines.

0:17:01 SC: So what are… I’m just trying to give the audience a little bit more of a feeling for what the solutions look like. Even if the problem is not completely solved, you mentioned that there are metastable states. Does that mean like, I go every week or I never go?

0:17:14 NJ: Yeah, so there’s one particularly interesting behaviour that emerges. So I’d mentioned in the past that complex systems can have emergent phenomena, that is something… So what is that? So there is one… A wonderful behaviour that happens, there’s a terrible… First of all, let’s give you the bad news, the terrible behaviour is, of course, when we all think we have the right rule, and then when we change our rule, we change it in similar ways, and so we’re always doing the wrong thing. That would be a huge crowd phenomenon, whereby a whole crowd go, or they don’t go. So the fluctuations in that system would be of the size of the number of particles, which is…

0:17:58 NJ: That’s a system instability, which is interesting in itself. But there’s a wonderful emergent state that arises that you can think of in the following way. Imagine you and I are following our rule for how we’re going to decide whether to go. And we begin to diverge in how we change our rule, and other people are following more closely you, just inadvertently, or me, in terms of how we adapt our strategy. And we can develop into effectively a crowd following one strategy. And a crowd, which I would call an anti-crowd, following the opposite strategy. Neither strategy is good nor bad because it just depends on what everyone else is doing. But imagine out of those 100 people, 50 are following a rule that says, “Go on the… “

0:18:52 NJ: It wouldn’t be as simple as that, but, “Go on the odd number weeks.” And then the other 50 are doing, “We’ll go on the even number of weeks.” So they don’t know each other, they haven’t coordinated in any sense, they just kind of ended up… But that crowd and anti-crowd…

0:19:07 SC: They don’t have a Facebook group. [chuckle]

0:19:08 NJ: That’s right, they had no Facebook group, which brings us onto… That’ll bring us onto other conversations.

0:19:12 SC: We’ll get there later, yeah.

0:19:12 NJ: Yeah, we’ll get there later. But without any coordination, without any local information, just global information fed back to them, like some external field sending them news, global news, they’ve managed to split themselves into a 50/50, which means that if you look at the… If you’re the bar manager and you’re looking at the system output, in other words, how many people are actually sitting in your bar, you’re hitting 50, 50, 50, 50, 50. You have no fluctuation. Perfect. And even if we all tossed coins that were… Actually not even based on 60/40, pretend we don’t even know what the seating is, we just toss a coin, decide where to go. That has fluctuations in it because 60 heads, 40 tails; 70 heads or 30 tails. But that end result, that system result, purely driven from competition of the 50 and 50 gives no fluctuation in the system.

0:20:14 NJ: So that begins to make you wonder, “Goodness, is it possible that I could even design then many complex systems? Whether they be, I don’t know, it could be drones, it could be things that don’t wanna get in each other’s way, it could be something in a biological system. Could I, if I… Could one way of controlling complex systems be to nudge them so that half of them are following one strategy, and half of them are following the opposite strategy? Or even if there were many, many strategies, make sure that they’re sort of paired up in opposites. That’s a wonderful outcome from that… Studying that very, very simple problem.

0:21:00 SC: Well, I think this is… I like that a possible solution has been given because that’s how we see that this very simple problem is a paradigm for complexity because of that emergent feature that you wouldn’t have guessed, right? Without central planning or even explicit cooperation, some kind of pretty good solution emerges out of people trying their best individually.

0:21:24 NJ: Correct, and actually, it’s probably even better than trying to do it top-down. Imagine you’re actually calling up the people and saying to them, “Well, you’re one of this group, or you’re one of the opposite group this week.” Would they listen to you? So the fluctuations that will come out of trying to force people to do something would be even worse… Would be worse than the ones that come out of competition.

0:21:49 SC: This is a slight deviation from the line here, but your discussion reminded me of a question that has been working at me for a while, like game theory, which is a theory of making decisions under situations of competition and perhaps incomplete information, etcetera, etcetera. Is game theory secretly physics? [chuckle] Can we reduce it all to statistical mechanics, optimising some utility function? Kinda sounds like minimising some Hamiltonian or some action or something like that.

0:22:20 NJ: Sure, and I would… I think you should go out and write that book, and I’ll be the first one to…

0:22:22 SC: I want to. [chuckle]

0:22:23 NJ: I will buy it, and I will… Because I think you’re right. In fact, there are lots of hints of that in the… Particularly in the type of probabilities that are used… The forms of the probability that are used by “agents”, look very much like two-level systems in a finite… At a particular temperature, the probabilities, the Boltzmann weightings. And so you’re absolutely right. I wish I’d have had that graduate course when I was doing my PhD, that would have laid the scene, instead I’ve had to learn it the hard way from standing up and having game theorists call out, “Oh, what you’re telling me is game theory,” and I feel like saying to them actually, “What you’re telling me is physics.”

0:23:08 SC: So this is something that you agree is, sounds kind of true-ish, but is not an accepted fact that it’s been well-known since von Neumann or something like that.

0:23:17 NJ: Yeah, although… There are even closer… It’s remarkably… There’s a area of game theory now, after all these years, after seeing A Beautiful Mind, and all these things with two-player games, there’s now an area of game theory called mean-field games.

0:23:38 SC: Oh, okay. There we go.

0:23:39 NJ: Mean-field game theory. In fact, I’m…

0:23:41 SC: The physicists have sneaked in.

0:23:42 NJ: Yeah. So this has emerged pretty recently, and it’s purely inspired by physics, the idea that mean-field approximation… It’s funny because in physics, if you go around saying, “Oh, I’m just gonna do mean-field theories,” people will very quickly tell you where mean-field theories break down. But the game theory community have… As they’ve tried to get towards more and more players, have realised that there’s no way they can write out exactly what each of these 10 to the 23 players will be doing. [chuckle] So they need some kind of re-normalisation of the system, and that comes through this mean-field game. So if there is an area which would really, really profit from people from the physics community joining with people from the game theory… Neither have got the right answer. But that is really a… Particularly for the dynamics of how things get towards equilibria or avoid equilibria, that’s the interesting bit.

0:24:53 NJ: As they said, “In the end, we’re all… In the long run, we’re all dead. Equilibrium is death.” So every system we’re interested in, all of us, whether… Even expansion of the universe, that’s a non-equilibrium system. So it’s the behaviour away from the dead state that we’re all interested in.

0:25:12 SC: But that is the hardest, and physicists themselves are not that great at that, as I know from experience, but…

0:25:18 NJ: Correct.

0:25:19 SC: But you also… You mentioned some phrases that I wanna repeat because it reminds us how this is not a simple physics problem, even if there is some version. You mentioned adaptive agents, and also rational agents. Part of the… What makes, in my mind, complex systems different than just physics systems is that we’re not dealing with particles that are dumb and memoryless and just bounce off each other. Part of what makes a complex system is that we assemble it out of individual pieces that themselves have some structures and maybe even goals and memories and things like that.

0:25:56 NJ: Absolutely correct. And so one of the challenges I’m particularly involved in in my own work is, how can I embody that in the simplest possible way? That kind of internal character. It always reminds me of Einstein’s phrase of the hidden variable. How can we put in some kind of characteristic that is simple enough, that captures what you just said, captures that adaptiveness, etcetera, but doesn’t go overboard and suddenly become, “I need a whole book just to explain one person”?

0:26:38 SC: Yeah, okay. But that’s why it’s interesting. So I think we have a good feeling now about what complex systems can be and how they arise, but despite the fact that it is still arguably pre-paradigmatic, some things appear over and over again, and one of the obvious things I’m thinking about are power laws. Maybe you can explain to the audience what a power law is, why we care. And are you someone who thinks that looking for power laws is a very important thing, or do you think it’s overrated?

0:27:08 NJ: That’s a fantastic question. So just to set the scene. We talked about coins, tossing coins. I think it’s… Imagine if we all toss coins, toss 100 coins. Pretty much 50, 55, 60, not so often 70, very infrequently 80 would come out as heads. So there’s a distribution of outcomes, and that distribution looks like a… Goes up, has a single peak, is a bell curve. It’s a bell curve. So as a society, we do a lot with bell curves. We assess risk, we think about medicine in terms of bell curves, we think about the average patient response, and we should do. That’s a very reasonable thing to do if, on average, people have some kind of average behaviour. Of course, in medicine, you’d think that the biology has an average, and maybe there’s fluctuations, but there’s some kind of average behaviour. But when you get in and look at social systems, and actually a lot of physical systems as well, these averages do not necessarily characterise the system very well.

0:28:28 NJ: Think about the… It’s a very topical thing now. Of course, wealth, the distribution of wealth in the country. A lot of people have little, very few people have a lot, but that distribution looks nothing like a bell curve. There is no kind of… Here’s the average, and then there’s a few outliers. One in a billion has more than twice, or more than three times the average wage. We all know that that is not true. In fact, that tail of wealth goes on a long, long way, and actually…

0:29:04 SC: So you’re in a situation where almost everyone is below average.

0:29:08 NJ: Right, and so if you’re doing… And of course the economists do not do this, building an economy around the average, and assuming that everyone has the average income, and then plus or minus $10 here and there, then all houses would have to be the same price, and there would be no high-end restaurants and low… It would just…

0:29:29 SC: A different world, yeah.

0:29:30 NJ: Yeah, it would… ‘Cause that would be, everything would be available to us. So we know that in other real-world settings, we have nothing like a bell curve. Now here comes the problem, bell curves are wonderful for calculating things like risk and the odds of something being an unlikely event. We can calculate lots of thing. We build buildings based on a bell curve. We assume that nobody comes in the room that is… Somebody can come in at 6 foot, but they’re not gonna come in at 6,000 feet of a height. And if we did… If that were the case, that every so often, every… Once in every nine months, you’d get someone of 6,000 feet, suddenly, buildings would be… We would have lost the purpose of buildings.

0:30:26 NJ: So it’s ingrained in the way we think, that some things are bell curves, and some things are other kind of distribution. The problem is we don’t really have a good mathematics of these other distributions that are so-called fat tails, where somebody can be earning 1,000 times what somebody else is earning. And so those fat-tail distributions are a problem. Now in physics, we are pretty used to those type of distributions. A particular form of them are called power laws, and the power law is characterised by the fact that what happens at one scale is exactly the same as what happens at another scale. It’s just twice as much, or half as much.

0:31:14 NJ: Now let me give an example. Scientists and physicists in particular are fairly convinced, seem fairly convinced that things like earthquakes follow a power law, or might follow an approximate power law, which means there’s a very deep consequence of that. It’s not just that there could be very big earthquakes. The consequence is that it’s scale-free, the behaviour. In other words, what’s happening? Earthquakes of a certain size is the same as what’s happening for earthquakes of twice the size, just at a different scale. So the consequence of that is if you’re building a theory of big earthquakes, dangerous big earthquakes, you really should just use the same theory that worked at lower scales. Now as the science carries over from scale to scale, it’s not like there’s some new phenomenon that suddenly appears as you get to earthquakes of a certain size. So that idea of scale-free is a very powerful thing in science, and particularly in physics, because that is the distribution that occurs when a system tends to be near some kind of phase transition, so called transition between system behaviours.

0:32:31 NJ: Just to give an example, forest fires are another example. We talked about earthquakes. Now let’s talk about forest fires. Forest fires at the transition between no fire across the system and fire just spreading everywhere, they tend to have a power law distribution in the clusters that are burning, for example. So this idea that power laws exist at critical junctures in complex systems is a very powerful one and a true one.

0:33:03 SC: Let me jump in just a second here because I think I remember back when I was a graduate student, and I was confused sometimes by the phrase “phase transition” being invoked in different circumstances ’cause I think of phase transitions of like melting ice or freezing water, and there are things that happen over time. But when you talk about the forest or the earthquakes having distributions of sizes, of events that are characteristic of phase transitions, it’s not because, I think, correct me if I’m wrong, the forest is literally undergoing a transition from one phase to another. It’s that there could be different phases, and the forest is perpetually stuck in between them. Is that basically right?

0:33:48 NJ: Correct. That’s a fantastic observation. You’re absolutely correct. So a shortcoming of the current understanding of power laws and phase transitions is that they happen exactly… You sit the system there, and it’s either there, or it’s in another situation. Getting it to… Describing how something would happen in time is the area of dynamical phase transitions, which is a really… That’s a much newer area of physics because it’s a system out of equilibrium. One can talk about systems… “In equilibrium, I’m gonna set the fire, the forest, just at its critical point, just where the fire would spread, or it wouldn’t spread.”

0:34:29 NJ: Yeah, okay, but that’s stationary in time. But if I’m thinking of something beginning to spread, and I’m fanning it, or I’m adding things in from the outside, more trees, it would be equivalent, or for a COVID outbreak, for example, more people, more susceptible people. That gets into the dynamics, the dynamical phase transition. That’s a much harder situation to describe, but the relevance of the power law is still there because you can imagine situations where I have a whole bunch of objects affected, which will be a huge event, that’s like having one person have a whole bunch of all the money in the world, or I can imagine many times, it would just be a few objects that are affected. So that’s the whole… The rest of us are not earning very much. So the power laws are pretty ubiquitous. Now there has been, of course, an interesting debate, and the literatures do, “Well, is it really a power law? Is it approximately a power law?” But putting… I think that’s a little bit… That’s getting a little bit too…

0:35:44 SC: Inside baseball. [chuckle]

0:35:45 NJ: Technical. Yeah, because in the end, we could do an experiment at home. Since we’re at home now, we can do an experiment on a phase transition, and we could try and map out the size of the clusters. Even though it’s a correct phase transition, because of the way we’re measuring it, we haven’t got all the information, we’re not recording it properly, we make mistakes. We’re never gonna get a perfect power law. So the power law is more indicative that the system is on some kind of tipping point.

0:36:13 SC: Right, I guess that’s what I wanted to get into because I think that sometimes from the outside, people see various kinds of disciplines. People show power laws, and they debate what the exponent of the power law is, or whether it is a power law or whatever, but the thing that gets missed is the reason why we care is because the existence of a power law might be pointing to some physical mechanism that creates this. It’s a different kind of physical mechanism than would create a bell curve.

0:36:45 NJ: Correct. And so the fascinating thing then becomes, “Hey, yeah, I may have two systems that have power laws.” If they’ve got the same slope, that’s a little bit like an exponent, that’s like saying two mountains that are both red slopes for skiing. If they’ve got that, there may be a similar mechanism behind them. And one of those mountains may be in Europe, the other may be in California, it doesn’t matter, there’s something about the way that mountain got created that gave it the same slope. And so it gives the hope that there may be some simple kind of mechanism that…

0:37:28 NJ: And this is where we get into… In my work in this area, we’ve worked on things to do with casualties in wars. You can get into a lot of hot water talking about this in this way, but it’s actually a fact that when you look at casualties from conflicts, for example, they could be very different conflicts from very different parts of the world, different motivations. If they have the… They have very similar power law slopes, exponents, which means that at the end of it, just somehow in the way that humans are doing things is the same in both places, the mechanisms. It’s how humans do conflict, like it’s how rocks get put together to make a red slope mountain.

0:38:21 SC: Well, this is a perfect segue, ’cause that’s exactly what I wanted to do, roll up our sleeves a little bit and get into the applications. And I know that you’ve been very heavily involved in conflict and radicalisation, and very… I don’t know why you’ve chosen a whole bunch of ways that people die as your conflict systems to study, but we won’t go into that. So what is the… So when we talk about these things, a power law of what? What is it that we’re actually studying?

0:38:49 NJ: Yeah, that’s a great question. Actually, I will do a little quick shout-out of why on earth I would end up in Steve Strogatz’ Cornell.

0:38:58 SC: A previous Mindscape guest.

0:39:00 NJ: Ah, fantastic. He once said that he studies the particular things, I think it was in reference to his Romeo and Juliet studies of non-linear dynamics, he finds it interesting to look at topics that are somehow taboo. And that stuck with me. That that’s an interesting area just because it’s unpleasant. The fact that there’s now data, a lot of data. So my interest in this grew through the Iraq war and the other wars that were going on. There was one… In Colombia, were going… That’s what… Suddenly, wait a minute, I can look… We can look at these two wars, lets forget all the thousands of books that have been written on their motivations and their consequence, just look at the numbers, the number of people that get fatalities per day, something like this, or per event. What would we see? So that’s how we got off on that.

0:40:04 NJ: And actually this goes back to… Lewis Fry Richardson was a… He’s a physicist, ended up looking at meteoric… What kind of weather patterns, et cetera, but he, in the First World War, got in his head that somehow to understand the human condition, human conflict, he would just track numbers of people in casualties… Sorry number of casualties in wars. So we did the same thing. He didn’t have the benefit of knowing in the individual events in a conflict how many people were killed, he just did… Had the totals for wars, which is more kind of messy… You get a messier result. But during the early 2000s when, unfortunately, of course, there was Iraq war going on, Afghanistan was going on, Colombia was going on, we started to look in exactly the same way, same lens at these three conflicts, and terrorism. There were terrorist events. We looked at all of these in the same way, and it was as follows. We looked at the number of casualties per event.

0:41:21 NJ: So for example, there may be five days when nothing happens, and then there’s… Something happens, and this number of people were killed. And we just literally plotted it out as a histogram like you would plot out how many people earn a certain amount of money. It’s now how many events had a certain amount of fatalities? And I think we did… I’m not sure what we expected to see. Isn’t that always the beauty? You’re not really sure when…

0:41:51 SC: [chuckle] Sure. Science.

0:41:52 NJ: People say, “What did you expect to see?” “I’m not sure.” But you might have expected to see something like a bell curve. You might have expected that, well, maybe in Iraq, where the particular capability among the insurgency, they would create a certain average number of fatalities every time they’ve decided to get involved in some event, and then there’d be fluctuations around that. So you might expect…

0:42:21 SC: Yeah, there’s some kind of best kind of attack that they just do over and over again.

0:42:24 NJ: Yes, yes. But that’s not what we found. So what we found was actually more like the mountain slope, more like the histogram of income. We found lots of events where very, very few were fatalities, and then there’s a whole spectrum up to a few events that were absolutely huge in fatalities. And when we… So we did the statistical tests, et cetera, et cetera, to check that it was a goodness of fit, as they say, for the power law. And when we looked at the slopes, what we found was… We were looking at Iraq, Colombia. By that time, word had got out, and we were being sent all these data sets from all these different wars that… Some of which, we’d never even heard of. But when we looked at them… Now there’s an important point, a lot of these were wars that were involving some kind of large state with some kind of organic, if we can call it that, opposition. When we looked…

0:43:32 SC: Not like the old World War II kind of wars.

0:43:34 NJ: Right, not like… And certainly not like the Hollywood movie, line them up at dawn, and two massive armies just… Well, which is an interesting point, interesting scientific point because that Hollywood picture of just line up at dawn on opposing sides, and just basically keep going till nobody’s left standing, that’s like… If you think about it in a kind of abstract way, that’s like a test tube of two reacting species that kind of just interact all day, and there’s nothing left at the end of it except the product. But these other… The modern wars, the insurgencies, and this is where, again, we got into pretty interesting territory with the social sciences of whether you call it an insurgency, or it’s a narco-guerrilla… If we… In the end…

0:44:21 SC: [chuckle] Better you than me.

0:44:22 NJ: These are the numbers. These are the numbers. We got power laws, and they all had the same slope, which is like saying, “I’ve gone around, my… I’m gonna go around the world, look at all the mountains, ‘Oh my goodness, they’ve all got the same slope.'” Why would that be?

0:44:38 SC: Which is not true in the real world for mountains, but…

0:44:40 NJ: Which is not true in the real world for mountains, but it was true, and still… And so we published this back in 2009, and we were expecting a deluge of people sending us results that were not power laws, therefore, [chuckle] and didn’t have… And even if they were power laws, didn’t have the same slope. But to this day, interestingly, now that we have much better data sets, and there are many, many people have now, into the counting the dead, if I can call it that, science, in the social sciences, and actually doing a remarkable job of counting fatalities per event. Still to this day, this idea that their power law holds up, and they all seem to have the slope around the red slope, which in this case is a 2.5 slope.

0:45:40 NJ: It’s remarkable, and it’s actually the same as the slope for terrorist events. So this is… Since then, we’ve come up with some physics models to try and explain that, physics-based models which have to do with the way that particles… Unlike the Hollywood situation of everybody lining up and mimicking what goes on at a test tube, it’s more the idea that objects, insurgents form into clusters. And it’s the… What you’re seeing is really the clustering… You see a cluster attack. And so it’s more to do with the way in which people form clusters. So human beings forming clusters between them.

0:46:18 SC: Maybe that’s worth going into more because when I think of these power laws, like you mentioned, classically, income, the explanation that comes to mind is the rich get richer. It’s not just a random distribution of money. Once you have money, you can get more, and so there’s a few billionaires and a lot of poor people. But for terrorist attacks, they seem more independent of each other, and so… But you’re hinting that the… That there’s some network phenomenon that explains some agglomeration of attackers that can help give us this power law.

0:46:53 NJ: Yes, correct. And it comes as a very, very simple story. If we, if you… And we could do this now, we sit here, and we could either write it out with pencil and paper, or do it in a simulation. If you take the simple rule that people, for example, and for this case, we talk about an insurgency, since they’re fighting some very strong state army or coalition army, they need to group together into something that is pretty substantial. So they’re trying to group together, they’re trying to coalesce into some kind of cluster. However, they are, every so often, being broken up because, just like you look at fish, fish under the… All these national, wonderful… Now that we’re all at home watching National Geographic programs, you see the fish gradually coalesce, they coalesce, they coalesce, and then they get… Sense that there’s a predator, and they scatter, they fragment. And that simple, simple, simple rule of coalesce, coalesce, and then occasionally fragment, remarkably gives exactly… Not only does it give a power law, it gives a power law with a 2.5 slope.

0:48:08 NJ: And so for all these years since 2009, we’ve been carrying this. This is our little story, waving our little banner saying that, “Well, I understand that, yes, they could be leftists, they could be rightists, they could be this, they could be that, they could be fighting in a jungle, they could be… They could have cell phones, they may not have cell… In the end, I’ll just give them two properties: They gradually aggregate… ” Well, not gradually, they just aggregate, and they fragment. Now I’ll make one slight addition to that: The distance between them is not as important as you might… In other words, it’s not. We’re not gonna… We’ll make it so that they can aggregate over arbitrary distances.

0:48:53 NJ: Now what do we mean by that? What we really mean is that they’re really just coordinating over arbitrary distances, and that’s actually known for most of those wars that we talked about in 2003 onwards, there were cell phones. There were people coordinating, so they would coordinate to do things in two places at once, for example. So the idea of coalescence without a distance and fragmentation gives you the 2.5. Now since then, lots of people have studied it. We’ve also done that around… Made our own efforts. You can put in distance, you could do all this. It doesn’t really get too far away from what I just said. And the beautiful thing is, and this is the graduate problem that I always set on my complex systems course, show with pencil and paper that that’s true. And it’s really… It’s two-and-a-half pages, and you remember, I remember doing a lot worse homework problems than that.

0:49:48 SC: Oh, yes. [chuckle] That’s very interesting. And actually, the way that you explained it there makes it seem very transparent to those of us who know a little bit about complexity, that it resembles the classic sand pile avalanche problem. You dribble sand onto the sand pile, and it will occasionally set off an avalanche. And so it grows slowly, but that it collapses all at once.

0:50:16 NJ: Correct. And so what I like about this is I always feel that every day I woke up in this complex systems field, I know less than I did the day before. But it’s… That model that I just… So it’s helping me understand. Now I understand what they meant by punctuated equilibrium. Now I understand what they meant by… So you can go back and look at these related ideas and say, “Aha! Now I see how to unify.” It’s not that we’ve come up with some fantastic mechanism and others didn’t have it. Others were probably on the point of having that as well. Maybe we’re getting towards a unified science.

0:50:57 SC: Yeah. So this is an overly ambitious question, so feel free to punt it, but one of our first guests on Mindscape was Geoffrey West, and he talked about his scaling laws that have these 1/4 powers, 1/4, 3/4s, etcetera.

0:51:10 NJ: Correct.

0:51:11 SC: And he claims to have a mechanistic explanation that relies on the fact that he’s talking about biological organisms or spatial systems that exist in three-dimensional space. And the four in his power laws is three plus one. If we were in different numbered dimensions of space, it would be a different fraction rather than four. Is there a similar feel good story you can tell about the 2.5 slope that you got?

0:51:37 NJ: I think he’s trumped me on that one, he got… He wins that one because that’s a fantastic way of seeing it. The 2.5, interestingly, is, it’s a little bit more involved, the explanation, but it… There are two levels of explanation. First is that power laws between two and three actually have… If you were to calculate their standard deviation, which is the size of the fluctuations, it actually diverges. And so there are systems that are, in some sense, infinitely adaptable. They can form small clusters, they can form large clusters. In a way, that’s like an evolutionary advantage.

0:52:21 NJ: So it being between two and three, and slap-bang in 2.5 sort of makes sense. Now that’s a less direct answer. But the more direct answer, the two and three is, interestingly, if you now add in the distance, and make those clusters form in a two-dimensional space, that exponent moves from 2.5 down to 2. And that makes more sense ’cause we’re now on a 2D space. 2.5 is really… It’s like clusters are forming. They’re not in 2D, they’re not in 3D, they’re sort of in-between. But I don’t like that particular explanation. What I do prefer is actually the one that 2.5 is a good thing for a system that is trying to compete and win.

0:53:10 SC: Well, I guess that was a set of questions I had as a physicist who’s just sort of trying to get into complexity myself, but who grew up doing nice, clean things like gravity and particle physics, etcetera. These are all open systems that you’re talking about. They’re dynamical, they exchange information and energy with their environment, they’re affected by the outside world. Is there some story to be told about entropy and free energy, and dissipation that helps us understand why these systems settle into this kind of power law, or is that something that would be different for every system, but we don’t even know yet?

0:53:53 NJ: I think that’s a fantastic question. I know that there are certainly work going along on that. There’s no definitive answer that I personally know of, but one would think that… And I think it’s such an important question that you just asked because in some sense, you answered that and you’ve answered this science of life because life itself is nothing more than us trying to keep ourselves away from the equilibrium. [chuckle] And so exactly what goes on in a non-equilibrium system, it’s heading towards the inevitable. However, on its way, it can do all sorts of wonderful things, and getting an idea of that beyond just the kind of idea that well over time, entropy increases, so we’re all gonna just… It’s the dynamics of that which is not worked out, so you’re absolutely right with that. I think it’s absolutely the pertinent question to ask. No clear answers yet.

0:54:57 SC: Yeah, it might be speculative, but can you imagine that there is some principle that we’ll someday hit on, that this power law scale-free distribution either maximises or minimises some entropy or entropy production rate? Is that the kind of thing we might be looking for, this complicated tree network system let’s us use up entropy as efficiently as possible?

0:55:25 NJ: Yes, I think that might be correct, and particularly if we take the vision of… Sorry, the interpretation of entropy as an information…

0:55:33 SC: Exactly, yeah.

0:55:34 NJ: Measure. Then in that case, I’m absolutely with you on that, that I do think it comes down to… And in that sense then, living systems are in some sense playing an information game with the rest of nature, and so they’re borrowing information for a while, they’re giving it back in a similar way that one might think of as an entropy of an entire system increasing, but the subsystem is locally, over time, giving and taking.

0:56:04 SC: Yeah, yeah, and it’s so exciting to talk about these things because it’s so hard in particle physics or cosmology to ask a question that on the one hand, we don’t know the answer to, but on the other hand, we can hope to find an answer. [chuckle] Whereas in complex systems, there’s all these questions we don’t know the answer to. They’re pretty easy to hit on.

0:56:22 NJ: Yeah. Sometimes, on a Monday morning, I’m really jealous of the others, saying, “What are we actually doing here?”

0:56:27 SC: The grass is always greener, I know, yeah.

0:56:30 NJ: Yes. [chuckle]

0:56:30 SC: But the story you told about the power law in these attacks sounds robust enough with the agglomeration and then dissipation, that it must be more widely applicable. What kinds of other systems should we be thinking about?

0:56:47 NJ: When we went and looked online at the growth in support for ISIS during 2014, 2015, there, the clusters are, they are communities online and we found a pretty perfect 2.5 power law. Because, again, people are trying to connect together into communities and occasionally, in that case, they were getting broken up by the social media company, by the moderator, because it violated the terms. They’re discussing violence and radicalism. It broke… So we’ve got, again, that idea of coalescence and fragmentation, and so we were able to show that you get exactly the same result of this 2.5 power law.

0:57:41 NJ: So in some sense, it helps think about and frame what people do when they’re trying to fight against the system, in some way, and so that gives you insight into probably the other… Could we even take something like some of these other far-right radical groups and do the same thing? Well the preliminary results suggest that we can and of course, that then completely changes the way that you might think about mitigating the build-up online. So most of the strategies, of mitigation schemes, that we see coming from the social sciences, they’re very much detailed, they’re very much focused on the bad actor, the bad apple. Which is a little bit like saying, “I’m gonna stop water boiling by taking out the bad molecule.”

0:58:41 SC: That’s a depressing but good analogy. Yes.

0:58:42 NJ: “Where’s the bad molecule?” “Okay, I’ll take the one with the fastest velocity going up perpendicular to the surface ’cause that’s the one that maybe most likely to… ” Get rid of that one, there’ll be another one, and then there’s another one. And so you really need… As physicists, we would never do it. Who’d imagine introductory physics where we’re gonna discuss every molecule in a glass of water before it boils? You’d never get anywhere and you’d be fired. So having a system view and saying, “Hey, as this extremist movement, whatever it is, ISIS beyond, the next version of it, which we hope there won’t be, but there will be extremes, the next one, it will… Whatever it is, it will be preceded by a build-up of clusters online that will form into a power law with a slope of 2.5. And the corollary of that is when you begin to see that happen, worry.” So it completely changes the focus of how you might go around trying to detect concerns, threats.

0:59:55 SC: So you can… Even though the slope might be universal, the hope is, nevertheless, there’s a normalisation? [chuckle]

1:00:02 NJ: Yes.

1:00:03 SC: There’s more little ones than big ones but still, the absolute number is something we can hope to increase or decrease.

1:00:08 NJ: Yes, and it then becomes down to maybe the world will then need to turn to the physicist to help with deciding policy.

1:00:15 SC: I’ve been telling them but they don’t listen to me. So I just wanna drive this home because the story you told can seem frustrating if you think that, well, as soon as these clusters grow big enough… We’re imagining these are clusters of things we don’t want like terrorist attacks, or extremist groups, or something. So they grow big enough and you can smash them and they disperse, but then they just re-agglomerate again. You’re not giving us just a sad story, you’re getting us a strategy as well?

1:00:49 NJ: Yeah, because… And you just said it, the big ones grow from the small ones. The big ones will probably be harder to tackle because they’re robust, they’ve got more eyes. We know they’re more aware of what’s around, but they come from small ones. So maybe there’s a strategy… That immediately suggest like the sand pile analogy. To prevent the big avalanches in the sand pile, just take care of the small ones so that they don’t build into big ones, ’cause the big ones need the small ones to build them. So, one could imagine an alternative strategy, and I’m not saying, it is not as simple as that, but there’s a systems engineering, complex systems engineering, whereby you’re not controlling…

1:01:42 NJ: Look at all the issue with the social media companies that they seem to go out and get the bad actors, bad influencers online. Well, that’s very hard to do. First of all, you’ve to go to find them, if they even exist. If we’re all on a spectrum of good to bad, which I… Maybe we could debate that, but maybe people are bad in certain moments, etcetera. First of all you’ve have to find them and then you’d have to stop being sued by them, if you try to remove them, they’re maybe they’re quite powerful people and so they’re gonna turn around… Or maybe they’ll… By you removing them, that will create even more support for them.

1:02:23 NJ: So you don’t… So, it gets off into those type issues. So is there a systems, a complex systems engineering? I don’t even know, there is no discipline of that yet, really.

1:02:35 SC: No.

1:02:35 NJ: But is there a way of engine… You’re a nudging the system in a way that then doesn’t impinge on… One of these objects can’t say, “Hey! You targeted me.” You’re more nudging the system. And that for me, is the light at the end of the tunnel of all of it. A lot of the systems that we worry about in society, it’s more… No, I think we’re thinking about it wrong, I think we should be thinking of almost nudging the… I don’t know how we could say it, nudging the chemistry in it.

1:03:14 SC: Changing the chemical potential, somehow, the sort of the background…

1:03:17 NJ: Something like that. Something like changing maybe if I could just stop some of those smaller clusters coalescing so fast, that would stop the flow through to the bigger ones. They would then be isolated in some sense. So yeah, I think there’s a lot of mileage there to be gained from the jump. And it’s gonna be an uncomfortable jump, but making that or building that bridge from what we do in our academic studies of complex systems to the policy field. And now I’m in DC. I actually have to say that I’ve had a remarkably good feedback and interaction with policy people here, I’m really pleasantly surprised.

1:04:05 SC: Well, we started by saying that one of the interesting things about these complex systems is you have these individual agents and they nevertheless seem to have emergent collective behaviour that was not given to them by top-down. And so in some sense, can we imagine the difficulties of these conflicts coming from the fact that the bad actors are acting like individual agents, complex systems coming together, whereas the white knights trying to combat them are more top-down? I remember reading something you did about how Facebook pages that were trying to give good advice to people to fight the pandemic, were all boring and they said the same thing over and over again, while the crazy ones said a million different things so they are more likely to hit someone in their sweet spot of vulnerability.

1:05:00 NJ: Yeah, that’s absolutely correct. And of course, this hits the big, big challenge for anyone who wants to work on complex systems, now is the time because this… The number one complex system in my head now is the online ecology involving trust, distrust of science, distrust of medical science, mixed in with the growing resistance to potential vaccines, misinformation, disinformation. That is the system to, in my view, be working on. And it’s precisely as you just said, that the… It’s almost like in… I don’t mean it in a political sense to call it an insurgency, but you’ve got some organic clustering going on, which in some sense is much stronger than the top-down go to this website to learn about vaccines. And because each one of these clusters in some sense to some degree autonomous for its kind, they’ve each develop their own flavours. And so giving them the, “Oh, he’s the best vanilla cream in the world.” I said, “No, I don’t want that. Actually, I like this other flavour. I’m not interested in that.” We see this actually in the online clusters, the communities and Facebook, we studied this a lot. We study across platforms, the clusters that form and they’re interconnections and how they evolve over time.

1:06:40 NJ: We see this that they’re talking about effectively strawberry ice cream, that they love it, although they’re talking really… To put it into concrete terms, they’re talking… Some of them are now… Some of them may be talking about, “Well, there’ll be no safety… There’s no five years of safety for the COVID vaccine.” And that’s accurate, they’re absolutely right. And then suddenly at the top appears where they had these lovely pictures of a baby’s hand, and they’re talking about these parents who love their kids. Of course, parents love their kids. So it’s a wonderful… They have a wonderful message. But then suddenly a banner appears at the top from Facebook, CDC saying, “Oh, we know, effectively, if you want to know about vaccines, go to this website.” It’s like parents coming into the room when you’re trying to develop some… It’s like, “Don’t do this, do that.” You’re not addressing the flavour of this cluster, so yeah, we’re studying that a lot now.

1:07:44 SC: Well, there is a famous story that I’ve heard attributed to John Wheeler, but it could be somebody else, or it could just be completely apocryphal, about physicists always get crackpot theories in the mail. And this physicist, who might have been Wheeler, had the strategy of putting the crackpots in touch with each other, so that they could talk to each other. But they always came back to him and said, “Why are you putting that person in touch with me? He’s a crackpot.” [chuckle] Just because you’re a crackpot doesn’t mean you’re sympathetic to other crackpots. So there’s many different ways to be wrong in this particular thing, but that actually helps the wrongness grow in some sense because there’s so many different buttons to push.

1:08:26 NJ: It does, and that’s what we saw with our… Why the support for ISIS grew so quickly. It wasn’t that there was some coherent message between them all, it’s just that what they were opposing was such an obvious… It was such an obvious target, and so you could oppose it in many ways. It’s like in a two-spin system, it’s not that you were the opposite spin, you could also be any… Any angle within the plane. So it didn’t really matter, you were opposing the main message, and the main message was bad. So you were allied by that. We see it with the far right, the hate, the online hate as well. They often disagree. Some of them will… These far-right groups, one of them will want a United Europe, the other wants to break it up and live in tribes in the caves; and yet that never gets played out because they’re both opposing the things that they rail against.

1:09:26 SC: I have to ask, certainly the way that you suggest thinking about the spread of the misinformation is remarkably analogous to the spread of the actual physical virus, [chuckle] as it gets from person to person. Is that…

1:09:40 NJ: Correct.

1:09:40 SC: How good is that analogy? Is it more or less the same thing, or are there secret differences that make all the difference?

1:09:45 NJ: That is a fantastic question. I just came off last week off the back of the World Health Organisation’s first conference. I was leading one of the science teams at the first conference on infodemiology. I’d never heard of that word. And they told me, “Well, you’ve never heard of it ’cause it didn’t exist.” They put together their earlier claim that the online misinformation was like an infodemic, which we… That was a few months ago, with the idea of, “Hey, maybe we should be therefore thinking of an infodemiology, to think about the online world.” But then there are a few important differences. Number one is that I can sneeze on one place, and somebody catches the cold or whatever on the other side of the planet instantly, ’cause it’s online. So the notion of distance. What does it mean to social distance online? That’s certainly not a strategy.

1:10:45 SC: The world is flat.

1:10:47 NJ: Yes. So certainly that’s not a strategy, but there are strategies. One thing, we’ve just finished a piece of work, we tried to calculate… A lot of governments managed to… Seemingly managed to handle the situation, talking about the R number, which is basically the number of people… If you’re infected, how many people are you gonna infect before you recover? ‘Cause if you infect another one, and then you recover, well you’re just propagating the same situation. If you infect more than one, the things are expanding. If you infect less than one, which probabilistically, you could, then it’s gonna die out in time. So this R number, whether it’s bigger than one or less than one, is an important number. But interestingly, it’s calculated for… In a very… In many ways, a ad hoc way in the epidemiology literature.

1:11:48 NJ: So the idea is… One of the reasons is that they just don’t know the network that people are in, and people are in networks all the time, and they may be in a social network, and yet never meet the person… Not meet them for five days or something. So in a way, they’re forced to think about… Epidemiology is forced to think about disease propagation again like a test tube. I stick in a test tube, susceptible people, and a couple of infecteds, they gradually all become infected, and then they recover. And so I’ve got that reaction problem. And so that’s how you tend to define the R number mathematically, and all these… The model and the science that we always hear about in government updates. But what does that become in the online world? Because in the online world, we have… Although you can instantly infect someone on the other side of the planet, there are different universes in the sense of, Facebook is its own universe, but it has no interaction commercially with something like Gab, which is another social media platform, or with 4chan, which is quite infamous.

1:13:08 NJ: It’s a relatively free and on unregulated platform. They’re separate universes and yet people tunnel between them. I literally think of this as universes with some kind of wormhole between them that people can go through across space and unfortunately, because they’re recorded… Because a lot of the postings, you can pull them up from so-called time machine back cache pages. It’s almost like across time as well, so postings come across time and space. So it’s almost like a worm hole, but you have to excuse me on this one, but it is like a wormhole and that these universes therefore are connected in some… It’s almost like some kind of multiverse. So in a way, it’s again ’cause I’m biased, but it’s a perfect place for physicists to jump in and think, “Well, I like to think about more multiverses. I like to think about universes and I like to think about couplings.” And these wormholes can come and go because they get shut down and sometimes they get opened up again, but they’re very real things. And so you’ve got a spreading phenomenon in that multiply connected space which is interesting, very interesting.

1:14:31 SC: Yeah, yeah. I don’t know if physicists would be good at this because we gab about wormholes, but usually we like talking about smooth space times, I think that’s our comfort zone. We should be better at talking about these sort of fractally connected places.

1:14:51 NJ: Yes, you’re absolutely right. So it does become a kind of spreading in some fractally connected space, exactly as you said. And what is being spread isn’t as clean as just a virus and no virus. It’s mutating, it’s there are multiple versions of it. And so I think it’s a fascinating challenge, this infodemiology, and it’s gonna become huge because as we get more… Some of these vaccines come on… Already, the rumors around some of these vaccines are just remarkable, and there are many of them. And who knows, it’s impossible that they all go well, some of them will go very well, and that’s just science. That’s the way it goes. But the opportunity for mis and disinformation is absolutely huge, so I think that’s… Just putting on the science hat for a moment, that’s a fascinating problem.

1:15:51 SC: Right, so the science hat as opposed to the moral or policy hat, which is… It can be frustrating, these problems.

1:15:58 NJ: Yes, which has me actually freaking out, yeah.

1:16:01 SC: Well, okay. So for one last little freak out, just to close things, I’ll get to one of my favourite issues here, which is, what are the implications of these ideas for the very functioning of democracy itself. There’s this idea that we need an informed populace, but the very notion of what it meant to be informed was different 200 years ago than it is now. I mean, are we gonna have to rethink what it means to be a citizen of a democratic society, or is it even a challenge to the idea if people can sort of fall into these local minima of crazy misinformation and that’s never gonna go away?

1:16:45 NJ: Yeah, that’s absolutely important, and a fantastic question. I do see this issue. I don’t think science has ever had a situation whereby it’s being done in real time, and the whole world is on top of every pre-print, and every misstep and discussing with… Imagine any of our own… Any of us in doing our own bit of science, if suddenly, every pre-print you wrote was hold over and discussed in Facebook, in online communities, and the limitation section was just expanded into some list of problems with science.

1:17:30 SC: To be fair, it is better than being ignored?

1:17:33 NJ: Well, that’s an interesting question to answer, I wonder sometimes. But the idea that the trust in science, in the science side. We all know this, if any science… If there were no unknowns in science, we wouldn’t have a job. So science is full of unknowns, but how to stop that being perceived as uncertainty in science is, I think a huge challenge. And so I actually think that, I’m biased again, but I actually think that is the core of what’s gonna go ahead, and I think democracy… I’m not a political expert anyway, but I do think that science is a key driver.

1:18:29 NJ: And I think trust in science is something that is absolutely central to this going forward, and people need to somehow, the public… We’re members of the public, when are we all gonna trust politicians? Well, maybe never. Maybe some of them one day, some of them, yeah. But science, science is sort of like, the breath of… Okay, we understand that they’re trying to get the answers. We understand that they may not always have the answers, we have… That level of trust, I think, is absolutely key going forward. And I think it underlies all these other issues ’cause as soon as you start losing trust in science, you start dis… Who can you trust? Who can you then trust? We will love members of our family. I have members of my family who have given me advice on COVID that I would… I’d probably be 6 feet under if I took it. I love them, however, I don’t trust them on the science and I need to trust.

1:19:32 SC: I think this is a genuinely subtle, sticky, difficult problem because obviously yes, I would love people to trust science more than they do. On the other hand, there’s two other hands. One is that science is the paradigmatic bottom-up thing, where there’s no pope of science who tells us what the truth is, there’s people who disagree with each other and fight, and they battle, and try to agree. So when you say trust science it’s not clear who you’re trusting. And the other is that even scientists are biased and wrong sometimes too, so the trust can’t be even absolute. So I’m torn about this. I want more trust but not absolute trust.

1:20:11 NJ: Correct. Yeah, no, you’re absolutely right. You know and I know that if we took members of… People saw what goes on in conferences, the arguing in the corridors and the backstabbing in terms of “your theory, my theory”. So when I mean trust in science, I do not mean trust in one particular result in science.

1:20:34 SC: Right, exactly.

1:20:35 NJ: I mean trust in the process.

1:20:37 SC: Yeah.

1:20:38 NJ: And so just as democracy is a process and we don’t have to necessarily trust one thing or another, I think trust in science it’s not the result of any particular paper, it’s the actual process. And so this idea that you could go to a conference and they’ll be people, we’ve seen it, we’ve all seen it, shouting in the halls at each other, and almost like separate groups in particle theory that won’t even talk to each other. We know these, we know these stories. But that’s…

1:21:11 SC: In the same department.

1:21:12 NJ: In the same department. But that’s probably necessary, that’s how humans do it. It’s the crowd and the anti-crowd. Together, they make the system.

1:21:23 SC: That’s right. The thing we should have trust in is that emergent process that… In the long term anyway, we’re allowed to critique it in the middle term. But in the long term, we think that we’re aiming in a good direction, scientifically speaking.

1:21:35 NJ: Correct, correct.

1:21:36 SC: Well, that gives us some hope. I always like to… I always tell my guests I like to end on an optimistic note, and that sounds like an optimistic note to me after you’ve given us some more sobering things to think about it. So Neil Johnson, thanks so much for being on The Mindscape podcast.

1:21:50 NJ: Thank you so much.

1:21:50 SC: This was great.

1:21:51 NJ: Thank you so much. Thank you. It’s been a pleasure, thank you.

[music][/accordion-item][/accordion]