COMPLEXITY: Physics of Life

Fractal Conflicts & Swing Voters with Eddie Lee

Episode Notes

Since the 1940s, scientists have puzzled over a curious finding: armed conflict data reveals that human battles obey a power-law distribution, like avalanches and epidemics.  Just like the fractal surfaces of mountains and cauliflowers, the shape of violence looks the same at any level of magnification. Beyond the particulars of why we fight, this pattern suggests a deep hidden order in the physical laws governing society.  And, digging into new analyses of data from both armed conflicts and voting patterns, complex systems researchers have started to identify the so-called “pivotal components” — the straw that breaks the camel’s back, the spark that sets a forest fire, the influential (but not always famous) figures that shape history.  Can science find a universal theory that predicts the size of conflicts from their initial conditions, or identifies key players whose “knobs” turn society in one direction or another?

Welcome to COMPLEXITY, the official podcast of the Santa Fe Institute. I’m your host, Michael Garfield, and each week we’ll bring you with us for far-ranging conversations with our worldwide network of rigorous researchers developing new frameworks to explain the deepest mysteries of the universe.

This week’s guest is SFI Program Postdoctoral Fellow Eddie Lee, whose work into “conflict avalanches” and swing voters gives a glimpse of the mysterious forces that determine why we fight — and how we may be able to prevent the next conflagration. In this episode, we talk about armed conflict as a fractal and a form of computation, swing voters in the justice system and influencers in pop culture, and what these studies have to say about the deep constraints that guide the currents of society.

Just a note that this will be our last episode before a short summer break, to give our scientists uninterrupted time to work on a torrent of new research. We have some exciting episodes scheduled for our return in mid-August…in the meantime, please be sure to subscribe to Complexity Podcast on your favorite podcast provider to make sure you stay in the know! And if you value our research and communication efforts, please consider making a donation at santafe.edu/podcastgive, or join our Applied Complexity Network at santafe.edu/action.

Lastly, we are excited to announce that submissions are open for this fall’s inaugural Complexity Interactive, a three-week online, project-based immersive course where you get a rare opportunity for mentorship by a large faculty of SFI professors — including Cris Moore, Melanie Mitchell, Simon DeDeo, Danielle Bassett, Luis Bettencourt, Melanie Moses, Ricard Solé, and many more. For more info and to apply, please visit https://santafe.edu/sfi-ci

Thank you for listening!

Eddie Lee’s SFI Webpage & Google Scholar Page

Papers we discuss in this episode:

A scaling theory of armed conflict avalanches

Sensitivity of collective outcomes identifies pivotal components

Emergent regularities and scaling in armed conflict data

Collective memory in primate conflict implied by temporal scaling collapse

Go further:

Time Scales & Tradeoffs, an SFI Flash Workshop [video]

Join our Facebook discussion group to meet like minds and talk about each episode.

Podcast Theme Music by Mitch Mignano.

Follow us on social media: TwitterYouTubeFacebookInstagramLinkedIn

Transcript coming soon!  Thanks for your patience...

Episode Transcription

Transcript produced by Podscribe and edited by Rayyan Zahid.

 

[Intro] 00:00 :

You know if you have a forest with a lot of dry wood that hasn’t been cleared away, by previous small conflagration you can have a major conflagration, right? I am not saying that, that is conflict. But you could imagine some kind of rule that a regulator would perform in society. I don’t know what that rule is, but definitely one of the ways it manifests is by preventing the largest outbreaks from happening. Or, an attempt to.

And, you know, one of the ideas behind some of the models that have been proposed for conflict is actually similar to this forest fire thing. Which is that, the conditions for conflict will simmer, and when it reaches that boiling, you can have one. And when you have one, it’s this big thing but it depletes the system in a way.

And so, you are always poised between never being able to have these big conflicts and having these sorts of having like these smaller conflicts. Would it be good to prevent all conflicts or would it be bad? Would that sort of lead to some sort of special situation where you suddenly had the nuclear Armageddon? I don’t know.

 

[Intro music] 01:00 :

 

Michael Garfield 01:24:

Since the 1940s scientists have puzzled over a curious finding. Armed conflict data reveals that human battles obey a power law distribution like avalanches and epidemics. Just like the fractal surfaces of mountains and cauliflowers, the shape of violence looks the same at any level of magnification.

Beyond the particulars of why we fight, this pattern suggests a deep hidden order in the physical laws governing society. And digging in the new analysis of data from both armed conflicts and voting patterns, complex systems researchers have started to identify the so-called pivotal components involved. The straw that breaks the camel’s back. The spark that sets the forest fire. The influential but not always famous figures that shape history.

Michael Garfield 02:11

Can science find a universal theory that predicts the size of conflict from their initial conditions? Or identity key players whose knob society in one direction or another?

 

Welcome to Complexity. The official podcast of the Santé Fe Institute.

I am your host Michael Garfield. And each week we’ll bring you with us on far ranging on rigorous researchers developing new frameworks explaining the deepest mysteries of the universe.

 

Michael Garfield 02:39:

This week’s guest is SFI program postdoctoral Eddie Lee. Whose work in conflict avalanches and swing voters gives a glimpse of the mysterious forces that determine why we fight. And how we will be able to prevent the next conflagration.

 

In this episode we talk about armed conflict as a fractal and a form of computation. Swing voters in the justice and influencers in pop cultures. And what these studies may have to say about the deep constraints that guide the currents of society.

 

Michael Garfield  03:11:

Just a note that this will be our last episode before short summer break to give our scientists uninterrupted time to work on a torrent of new research. We have some exciting episodes scheduled for our return in mid-august. In the meantime, please be sure to subscribe to complexity podcast on your favorite podcast provider to make sure that you stay in the know. And if you value our research and communications please consider a donation at SantaFe.Edu/podcastgive or join our Applied Complexity Network santafe.edu/action

 

Michael Garfield  03:43:

Lastly, we are excited to announce that submissions are open for this fall's inaugural complexity interactive. A three-week online project based immersive course where you get a rare opportunity for mentorship by a large faculty of SFI faculty.

For more info and to apply, please visit SantaFe.edu/sfi-ci

 

Michael Garfield 04:23:

All right. Well, shall we

 

Eddie Lee 04:25:

All right, let's do it.

 

Michael Garfield 04:26:

Yeah. Eddie Lee, it is a pleasure to have you here and our first ever in person, socially distanced podcast recording here across my patio. So welcome.

 

Eddie Lee 04:39 :

Thanks for having me. This is great.

 

Michael Garfield 04:42 :

I'd love to start as we typically do by just providing a little bit of personal background and having you talk a little bit about how you got into science and how you got into specifically an interest in the kind of research that you're doing at SFI and the stuff that we'll be talking about today.

 

Eddie Lee 05:00 :

Well, I think it was last week that actually the writer of the magic school bus series passed away. And, you know, I have to say that she really did leave a fantastic legacy and I'm part of that legacy. I grew up on, on the magic school bus and actually some after school science programs as well. So, I sort of had this nascent interest in science for a long time, something I enjoyed doing. I didn't really see myself becoming a scientist.

 

Eddie Lee 05:32:

And in fact, when I went to college, I actually tried to become an economics major surprisingly enough, but it turned out that I took this sort of fantastic course called integrated sciences. And it's sort of my preview in a way to complex sciences, because what it was a class telling us, showing us that the way of thinking in computer science, biology, chemistry, physics, we're all coming together in the context of some fascinating problems in biology.

 

Eddie Lee 06:03:

So, this idea that all these things were coming together was a really surprising thing for me. And I think what really sort of hit it home was when I read this book by Phillip ball called critical mass. And so, he brought up the aspect of society, also being part of this. And, you know, at some point in my past, I'd also read Isaac Asimov's foundation series, which is, which is the physics of social behavior and prediction. And so somehow all of these things were sort of swirling around in my head.

 

Eddie Lee 06:35:

I didn't know what to do about them. And I ended up imagining what I really need to do is go into sociology. So, I went to, I searched around, I'd try to talk with sociology professors. And I just talked with one, I think he was maybe at the head of the Taryn at the time. And he said, you know, I know this guy who might be able to direct you. And he couldn't remember the name, but he started searching online and he showed me the web page. And it was one of the professors who had taught that first class, that first course that I mentioned integrated sciences.

 

Eddie Lee 07:05:

So, he sorts of just sent me right back. Yeah. Boomerang style back into that professor's office. And he actually was familiar with SFI. So, he's the one who directed me to talk with Jessica and Dave Cracker, who now I've known for a long time. And that's sort of where all of this stuff started happening. And that's when I started really learning about what sort of science SFI works on and my sort of vague interest in it became, you know, crystallized.

 

Michael Garfield 07:37:

Awesome. Well, this is a great place actually, to just dig right in to the first set of papers I wanted to talk with you about today, which you co-authored with David and Jessica, as well as Brian Daniels and Chris Myers on scaling theory for armed conflict avalanches. So this idea that we can use physics insights to understand human social behavior from like an orbital perspective and like really much like in the foundation books, you know, that we can, we can come at the, the sort of vagaries of history from, you know, David would call a rigorous and principled quantitative approach.

 

Michael Garfield 08:23:

The history is not just the story of great men making decisions, but that those decisions occur within a landscape of physical constraints. And what are those constraints and how do they manifest themselves? And this is really interesting work and it's got a long history that you're building on. And so maybe the history of this strain of research is the right place to start. And then we can unpack it from there.

 

Eddie Lee 08:48:

Sure. Yeah. Like you're saying, it's interesting, but what we weren't focusing on was the particular stories of particular conflicts. And there's a lot to study about any particular conflict they're quite different, right? So, you might think of say the Libyan or Tunisian revolutions happening today, which are quite different from world war two, which were quite different from the British and French battling it out, say in 1812. So, these are all very distinct lines of history in and of their own.

 

Eddie Lee 09:20:

And so, what we were interested in was trying to understand whether or not there were universalities common features shared amongst these things that are clearly disparate. And as you were saying, this isn't a completely new idea. It was in 1948 that Louis Fry Richardson showed this amazing feature, which was that when you make a histogram of interstate Wars, so you count how many wars there are with F people that died, F fatalities. And you look at this histogram on a log plot, what you find is a straight line.

 

Eddie Lee 09:54:

And what's remarkable about a straight line being on a log plot is that it's, that's a power law. That means its scale free. In other words, sort of most, very naively speaking. If you were to look at a small conflict, it's sort of like a shrunk down version, statistically, speaking of a large conflict and so on and so forth all the way down. And so, this was very strange because interstate wars that you looked at were all quite different from one another. So why would you get this sort of regularity that emerges when you look across many different conflicts?

 

Eddie Lee 10:27:

And since then other people have noticed other interesting patterns in the timing of conflicts, as well, as well as terrorism and so on. And what we were interested in was looking for some of these patterns in a more recent dataset, which is called the armed conflict location event, data project, or ACLED for short. And what was remarkable about this data set was that they didn't take all these events, these small disparate individual conflict events that would occur in a war. They didn't group them together into wars.

 

Eddie Lee 11:00:

They just sort of disaggregated them into, into small localized defense. And so instead of having to take a quantity, that's already defined for you as a battle or a war or skirmish, or just a local riot, we got to connect these into ourselves, into these clusters of what we called conflict avalanches. So many ways similar to two battles or Wars, but without having assumed some of these connections from using sociopolitical factors. So just looking at the time and spatial patterns of these events, and what we found was that when you look at these conflict avalanches, various measures of their size, so not just how many people died, but how many of these reports did you obtain?

 

Eddie Lee 11:43:

How long did it last, how far did it spread? All of these different factors show these characteristic power law tales, like what Richardson saw and what we were able to do beyond that is to show that these pieces were all connected to one another in some mathematical way. And I think I had an analogy for this, which was that if you imagine you walk into a dark room and you're trying to figure out what is this object in front of you?

 

Eddie Lee 12:12:

And you find first, there's this sort of rigid pillar. Then you find this long sort of thing with a hairy end. You find this floppy disc, like thing, you start in your mind developing this picture of an elephant. And the way that that works is all these pieces are connected to one another, right? They're pieces of one whole. They're not independent, separate things. And so, you build a model in your head and you say, this must be connected. The legs must be connected to the body, must be connected to the head and so on.

 

Eddie Lee 12:43:

So, we did something like that with the mathematics of these various features. And we found that they were self-consistent in a way that seems to point to one sort of object driving the dynamics.

 

 

Eddie Lee 12:55:

And that's sort of really cool because it suggests that there is some sort of underlying thing that we're just getting hints of. We haven't found it yet. We haven't discovered exactly what it is. Obviously, it's something about war and conflict, but it's perhaps something bigger, right? That leads to universal patterns, across many different types of conflict.

 

Michael Garfield 13:17:

So, you draw some analogies in one of the preprints that you've done on this work, emergent regularities and scaling and armed conflict data. You draw analogies to forest fire models, which other researchers have tried to apply to this, to neural avalanches. And, you know, just from last week's episode with the SFI counter speech team and the network graphs that you see of the Twitter data that they've harvested, and the way that you see these conversations, these sort of battles between hate speech, organized, hate speech groups and organized counter speech groups on Twitter.

 

Michael Garfield 13:56:

And this look also like these, you know, you see these the sort of conflagration of point by point debate breaking out when there's a successful, an actual collision, rather than just the hate group eliciting engagement to its original post. But people are actually willing to go into the trenches with one another on social media. And so, I would love to hear you talk a little bit more specifically about this model as a, like a branching network diagram.

 

Michael Garfield 14:26:

Yeah. There's some really interesting figures, and we'll link to this stuff in the show notes so that people can, can actually get a look at what we're talking about here. Because it actually looks rather menacing.

 

Eddie Lee 14:39:

That's why we chose those colors.

 

Michael Garfield 14:41:

The map of a branching armed conflict here.

 

Eddie Lee 14:44:

Yeah. So, cascades are a real central conceptual foundation for understanding a lot of these different kinds of phenomena. And as you mentioned, if you look at neural activity in the cortex, what you find are, are these cascades of activity, and it's not just a bunch of neurons firing, you can actually trace it out to some sort of branching process. So, you imagine there's the center, the first neuron that activates other neurons, which activate other neurons and so on sort of in a conflagration type picture.

 

Eddie Lee 15:24:

And you can also think about various other phenomena with this sort of conceptual framework. So, forest fires are another way, right? So, you imagine you have a bunch of sorts of local clusters of maybe dried out patches of trees and you have a spark, maybe some lightning strikes, and then it sort of burns its way through all these connected clusters. And you can think about that analogy applying to, to social contagion as well. And so that's why we sort of started with this idea of thinking about conflict in the same way.

 

Eddie Lee 15:55:

And mathematically, it's a very generic way of thinking about it, right? All of that you need are different pieces that are connected to one another in this sort of branching way. And so, you can get this exponential increase. Actually, this is very relevant in the context of today with COVID right? So, when people talk about exponentially growing processes, this is exactly what they mean. Or you have someone who spreads it to some other people. And if each person on average spreads up to more than one person, then you have an explosion. So, we sort of were playing with this framework and thinking about conflict in this way as a social contagion.

 

Eddie Lee 16:29:

And we're not sure if that's exactly how it spreads, but we can at least say that the way that it spreads is consistent with this idea of the way that it spreads both across space, as well as across time and in size.

 

Michael Garfield 16:46:

So, this elicits for me, links back to this sort of giant network model and building of this, these conversations on the show with all of the other conversations that hopefully we'll publish in the SFI press one day and make a poster. This reminds me a lot of, of two things. One is the conversations I just had with Jeffrey West a few episodes ago, where he was talking about biophysical scaling laws as being extremely coarse and like not actually capturing all of the variation that we would see in like specific evolutionary instances. The Primo example, being that human beings, modern humans use something like 30 times more energetic resources than we would expect from mammals of this size.

 

Michael Garfield 17:33:

Because as we've talked about with David and some of the transmission episodes, the human being is basically just the visible corner of this hyper object now. And then each of us is actually like a cyborg spanning the globe with all of these electronic augmentations. Right? So, something like that is going on in this work with conflict. And when you actually peer into the data with a little bit more granularity, you see a lot of variation in this, not enough to challenge the thesis, but enough to suggest that there are some interesting details about the regional and temporal variations of these conflicts.

 

Michael Garfield 18:14:

And I would love to hear you talk about that. And then also just to like make this kind of unwieldy let's pin onto that since you brought up COVID, Chris Morris piece for the transmission series, where you talking about super spreader and that's and how the, you know, our zero as a measure of the transmissibility of a disease, doesn't actually tell you what's going on in that particular church or in that particular campus building or, you know, so there's. So, what are the features about the data set that you've noticed that seemed to be providing insights into why there's variations in the size of these conflicts around that mean.

 

Eddie Lee 18:56:

No, that's, that's great. I mean, I think it's really important to point out that a lot of this work 's goal isn't to explain all this variation, which, you know, eventually we want to do, but it's really to talk about the shared common features across all of these phenomena. And then of course it's important to mention that, that's the only thing it explains. And so, the features that I'm talking about in terms of scaling are the averages, but you can have scaling in the variation as well.

 

Eddie Lee 19:29:

And that's, that's really interesting. And as you mentioned, right, you know, Chris Moore wrote that very nice piece where it's really important to mention that are not, is an average and not necessarily going to tell you what's happening in a very particular location. And that's also exactly true for theories about metabolic scaling in force or conflict. And in fact, it turns out that that sort of variation manifests directly in our work looking at conflict.

 

Eddie Lee 20:02:

And one of the things that we find is that it's not only essential to account for the similarities between how conflict grows in different locations, but there's also variation in terms of this sort of coefficient in front of it. So, the way that it grows the shape of the curve may be similar, but the offset of that curve may be different. And that might depend on factors such as prosperity or governance. So, here's one example, right? So, you look at Eastern Somalia, weak governance, a lot of widespread poverty and a huge amount of conflict.

 

Eddie Lee 20:36:

One of the hot zones in this and the status set. But if you look at South Africa, it's quite prosperous, they have a pretty strong government and its relatively low levels of conflict. So even if it's the case, that conflict evolves in a similar way, you just don't expect to see as much conflict and South Africa, as you see in Eastern Somalia. And so, as you also said, these things will vary in time. So hopefully Eastern Somalia is not a hotspot forever, but maybe somewhere else will become one. So, there may be patterns also in the diversity of hot zones.

 

Eddie Lee 21:09:

And we see a hint of this. We can't say that will be the case forever because we only look at 20 years. But one of the sort of fascinating things is that Richardson, who I mentioned earlier, looked at wars between, I think, early 1800s and mid-1900s. And he found basically the same statistics. Aaron?? looked at the statistics all the way from Richardson's asset up to today and found the same statistics. So, there's some sort of weird preservation of conflict despite changing technology, different countries. And, and so on that either, I don't know, is, should you be amazed or should you be afraid?

 

 

Michael Garfield 21:47:

Well, I mean, again, two things come up here, right? One is Jen Dunn's work on food webs and how the structure of these food webs that their patterns of connectivity have been amazingly well-preserved conserved in there, in their form for like the last 500 million years as we've gone through all of these regimes in what is actually living in the oceans. So that's part of it, you know, and that's, that's sort of begs the question that I asked Brian Arthur, which is, do you think that we can use these models as a way of predicting like basins for future innovation, you know, and like predicting areas where we can expect potentially larger conflicts than we're seeing.

 

Michael Garfield 22:30:

And then the other piece of it is, bringing in the geopolitical component, Peter Turchin's work. When we're talking about a quantitative study of history and his writing on what he called the double helix of inequality and social instability, where he saw that they were very strongly negatively correlated over the last a hundred or 150 years. But, you know, you make a point in this paper that there are other geopolitical and geographic features that have to do with the coastlines and national borders and locations of these conflicts relative to large urban centers' populations.

 

Michael Garfield 23:11:

Yeah. So, I, again, I know that this is sort of beyond the scope of the paper, but I'm really curious, you know, what other, what other factors you think are playing into this and how that might be guiding your follow-up research in this area?

 

Eddie Lee 23:23:

I think what you can say is despite the fact that there are a lot of these universal patterns that we think we find is that there's a lot of work trying to understand the mechanisms behind why these patterns might appear. And I think there is yet a connection to be made there. And that would be sort of, I think the Holy grail for, for this sort of work is really understanding those connections and flushing them out.

 

Eddie Lee 23:53:

And somehow, they lie in the intersection of maybe social mobility, social prosperity, economic prosperity, technology, geography, right. You can't really fight someone if there's no one there,

 

Michael Garfield 24:08:

Tell that to our Facebook group.

 

Eddie Lee 24:13:

Yeah. I guess you could make up stories and so on, but there are a lot of different factors that somehow mashed together to generate these patterns. But as you were mentioning, Jen Dunn's work. That's not necessarily to say that you can't get universal patterns from that. So, you know, this is maybe too simple of an example, but one sort of that comes to mind right now, which is the central limit theorem, right? There are many reasons for why you get random likes statistics, right?

 

Eddie Lee 24:46:

Many different ways of generating them, but by virtue of their randomness, you end up getting some sort of regularity. So, we'll see. Maybe it's the fact that there are so many factors that do influence conflict that ended up getting simplification at higher scales, or maybe not, maybe, maybe it is contingent. And we, we just don't have enough granularity in our understanding to access that contingency

 

Michael Garfield 25:15:

Later on, in this paper, you talked about there being, you know, getting into this granularity, you say: "Unlike canonical cascade models, conflict also includes lattice style dynamics that evolve with geographic spread. The suppression of these dynamics away from the core could reflect social processes or geography that impact conflict evolution. Furthermore, our model suggests conflict is not only the result of local correlations and activity, but also regional and temporal disorder, perhaps reflecting memory of the severity of initiating events".

 

Michael Garfield 25:47:

So, this links directly to another paper that you wrote with Brian, David and Jessica on primate conflicts and temporal scaling collapse. And the idea, you know, Jessica's work on primate conflicts as a form of collective computation and the severity of those conflicts having to do with various factors that are performing computations at the level of like an entire primate troupe, right? So, this notion that the patterns that you're seeing in conflict data, that they hinge on there being certain evolutionary reasons for the distribution of conflict, duration, and severity is really interesting. I'm curious to hear you talk a little bit about this other paper and how it links in.

 

Eddie Lee 26:36:

Right. No, that's, that's interesting. Right? What are the functional properties of conflict, right? Is there a reason why conflict plays out the way that it does? It's hard to say for human conflicts. Obviously, there are reasons why people give for starting wars. But in monkey conflict, there's potentially one reason which is, has to do with a social hierarchy, right? So, you need to have a little bit of disorder and conflict in order to establish people's roles and where they sit.

 

Eddie Lee 27:07:

And the severity of the conflict at the same time cannot be too, too big, right. If it's too severe, it's actually detrimental to society. So, there's this idea that maybe what conflicts does is it serves an information gathering acquisition role, as long as it's not too bad,

 

Michael Garfield 27:26:

People have to survive, remember.

 

Eddie Lee 27:28:

Exactly. If you don't, if you know, if it's nuclear Armageddon again and that's it, we should remember that. But, I think one of the things that we were trying to study in that work was characterizing the features of monkey conflict and perhaps extracting from those features a potential functional role that, that these things, or these observations could, could conserve.

 

Eddie Lee 27:53:

And actually, what's really interesting about the monkey conflict is we studied a group, a society of ??. What we found was that the duration that conflicts lasted well, beta distribution, it was not a parallel distribution like human conflict, but how long they lasted that distribution of duration seemed to be a distribution that looked the same, whether you looked at small conflicts or large complex. And it's very strange because conflicts with 2 monkeys are not the same thing as conflicts with 10 monkeys.

 

Eddie Lee 28:24:

There are many more ways they can interact. There are many other things that could happen, go wrong. And so, this was very curious. And so what we found was that if you could think of these conflicts as being sort of the time for all the pairs to resolve their differences, and each of those pairs, or took some amount of time to resolve their differences, then you could get something similar to what we found in the data with the condition that the pairs that resolve their differences later remembered the sort of intensity with which the pairs resolved their differences earlier.

 

Eddie Lee 29:03:

So, in other words, if the fight started off with some intense pairwise interactions, like, you know, you bit me, how could you bite me? And that's, that's actually very apparently a very aggressive interaction that the other monkeys will not tolerate.

 

Michael Garfield 29:18:

I remember preschool. Yeah.

 

Eddie Lee 29:19:

And so, you end up having these long correlations from the beginning to the end, and if that's true, then you end up getting these, these sort of long tails of conflict duration. So, you have many conflicts that tend to last longer. And for us, when we, that was an indication of what we call collective memory, right. This idea that the entire duration of the conflict from beginning to end, some remember at the beginning. And that's what we mean by we say in that paper.

 

Michael Garfield 30:00:

Well, okay. So, you know, this, I have the luxury here of kind of like going out on the plank and speculating. Yeah. But this seems to be possibly why certain human conflicts are just insane, like the Hatfield’s and McCoy’s, you know, that the feud has a sort of durability built into it because of the way that the history of a family is prioritized. And like the, you know, the bigger the conflict gets, it would seem like certainly there are factors that are keeping resolution that are like forestalling resolution, but on the other hand, the more abstract the conflict becomes for people, the easier it is to get over.

 

Michael Garfield 30:45:

Right. So, you know, this, this is sort of like, you know, I wonder what this approach to understanding this might have to say to, or link with other work on the way that conflict between human beings has changed as war has become industrialized and has been more and more about foreign conflicts rather than defending your own, your Athens or whatever. From there are certain things that would seem to be kind of holding us in the United States in a position of kind of unending foreign Wars.

 

Eddie Lee 31:20:

Right. Well, I think it's a little bit unfair to say that these Wars are external. Right. For many people, these Wars are personal, right? Yeah. And that's their life, but sort of touching on this, on this question of, you know, how is it that conflict persists or why is it that conflict persists? I think one thing we suspect from looking at these patterns is that these memories, these core correlations in a conflict, are expressed not only in the history that people maintain about themselves, but also in the geography of how it spreads.

 

Eddie Lee31:57:

So, it's quite possible that what encodes the history of the conflict is not just what we think right, the stories that we tell ourselves, but potentially also factors of the environment. Structures that we build. And some of these structures are really obvious, right. I mean, we built the nuclear arsenal, right. It's going to be there. Right. This is really interesting to me, it sort of connects to some of the work that David, Jess and I have been doing recently with adaptation and outsourcing a memory into the environment.

 

Eddie Lee 32:30:

Right. Which is this idea that there are biological organisms that intentionally or use and manipulate the environment to couple their bad memory or the behavior with longer timescales that they need in order to better adapt. So, ants do this by building these trails. It's unlikely that individual ants can really remember what they're doing, but over the collective activity of many, they can establish these very long-lived persistent trails. And they can use this trail to harness resources.

 

Michael Garfield: like Google calendar.

 

Eddie Lee 33:01:

Exactly. No, I mean, let me do exactly the same thing, right. I mean, David has that example of the notepad, right? Multiply three times, six times, five times, 2,374. I can't do it in my head, but give me a piece of paper. I can just, basically you short term memory expressed on paper to do that calculation very efficiently. And in addition to opening up new ways of doing that calculation, that aren't accessible in my head to myself. So, there is this question, right? Sort of going back to the functional properties of armed conflict, you know, what purpose does armed conflict solve?

 

Eddie Lee 33:32:

Are there ways that we are sort of driving our conflict ourselves by embedding into our environment that we don't know of, that we do know of. And I don't think those questions are necessarily answered. At least I haven't seen those answers, but would be really interesting to think about, I mean, right, this is just one other phenomenon in nature, right? It's not just us that fight. We do fight with certain technologies and armed human conflicts or especially cause a lot of fatalities compared to other organisms. So, we're quite brutal, but.

 

Michael Garfield 34:02:

You know, to, to bring up a marvelous work of science fiction, that seems to have some direct bearing on the insights that you just said. I've been in a book club recently discussing the science fiction trilogy Lilith's Brood by Octavia Butler. Should I read this? It's an extraordinary hit. It's a very, very relevant and kind of evergreen piece of work by the first science fiction author to ever become a MacArthur fellow. It just highly awarded a black female science fiction author, you know, just an amazing, amazing mind.

 

Michael Garfield 34:36:

And I've been on the tip of trying to read more black Sci-fi right now to get a better understanding of the world spaces disclosed thereby. And in this book, this book is about humans interacting kind of non-consensually with a race of aliens that comes to earth and finds us in the aftermath of a nuclear apocalypse. And once they're masters of bioengineering and they want to reboot our planet and nurse humankind and the biosphere back from extinction.

 

Michael Garfield 35:09:

But in order to do that, they kind of have to change the rules for human beings. And one of the things that they do is they refuse to give the humans that are going to repopulate earth, any writing materials, like it's part of their thing, that they all have genetically engineered eidetic memory. And so, they would rather engineer us to have perfect recall, then allow us to record history on paper.

 

Eddie Lee 35:37:

It sounds like a curse. You know, you know, that, I guess that was the thing for a long period of human history, right? Oral tradition, Homer, obviously, until it was written down. Yeah. What is it that keeps us fighting? Is it, is it literature? And that would be even, there'd be very tragic and somewhat ironic actually, because we're supposed to learn from history.

 

Michael Garfield 36:05:

But you know, in this paper, I think it's, like you said, it's the end of the conflict remembers the beginning. Yeah. There's another piece here, which is about the beginning. And I think that this might give us a hinge to get into some discussion about your other work on swing voting, which is really, you know, also very timely and interesting. You and your co authors on this piece, talk about policers or the police macaques. And they're being like thresholds at which they will or will not intervene in order to break up a conflict before it becomes a conflagration.

 

Michael Garfield 36:44:

And I'm curious to hear you talk a little bit about how thinking about conflict as a collective computation illuminates, why we may make a choice to mediate a conflict or to de escalate it or to allow it to run its course. Right? Yeah. So, what are your thoughts on all of that?

 

Eddie Lee 37:04:

Yeah. Well, first the disclaimer is that I'm not an expert on mechanics society. So, so that, you know, when I talk with Jess later, I can say that I said this, but yeah. So, these police or macaque or high-power individuals that, I mean, you know, in short other macaque don't want to mess with them. So, if they get in a fight, you just don't hit the policer or per usual. So, they're serving some sort of regulatory mechanism and maybe in the sort of conceptual abstract sense, regulatory control is very important in terms of outbreaks and contagion, right?

 

Eddie Lee 37:40:

So, for example, for example, and I'm not saying that this is macaque conflict, but you know, if you have a fire or a forest with a lot of dry wood that hasn't been cleared away by previous small conflagrations, you can have a major conflagration, right? So, you have this idea that by preventing all, but the largest of conflagrations you've sort of saved the system, allowing it to renew itself. I'm not saying that that is a macaque conflict, but you could imagine some kind of role, right. That, that a regulator would perform in this sort of society.

 

Eddie Lee 38:13:

I don't know what that role is, but definitely one of the ways that it manifests is by preventing these, the largest, you know, outbreaks from happening or an attempt to. And, you know, that is sort of interesting. I don't know what the computational rule of conflict in human society is, but you know, one of the ideas behind some of the models that have been proposed for conflict is actually similar to this forest fire thing, which is that the conditions for conflict will simmer, right?

 

Eddie Lee 38:47:

And then when it reaches that boiling point, you can have one and when you have one, it's this big thing. But it depletes the system in a way. And so, you are always poised between never being able to have these big conflicts. I mean, truly massive ones, you know, think world war five and, and having these sort of like smaller conflicts, but I don't know, I don't know what would be, right. Would it be good to prevent all conflicts or would it be bad?

 

Eddie Lee 39:18:

Would that sort of lead to some sort of special situation where you suddenly had, you know, the nuclear Armageddon and I don't know, but it is, it is sort of an interesting question to think about,

 

Michael Garfield 39:29:

You know, pull in a totally non-scientific perspective on this from a rather sketchy character. I remember Osho Rajneesh, the disgraced guru that brought his cult into Oregon, had a line in one of his books. He said, you know, be aware of the pacifist because they're sitting on a volcano and. And you know, race had, ?? D'Souza just gave one of the flash workshop talks last week on timescales and tradeoffs, where she was talking about this in terms of exactly what you just said about, you know, if you have small forest fires every once in a while, it prevents this big one.

 

Michael Garfield 40:09:

So, you know, the question is in our efforts to expunge conflict from the planet. Are we really doing ourselves a long range disservice, are we making a tradeoff That's actually making the system dramatically more brittle than it otherwise would be? And then this starts getting into sort of more archeological or historical perspectives on the role of organized sports as a way of, of, you know, channeling our aggression into something that's less likely to result in. I mean, you might flip a cop car if you win the championship.

 

Eddie Lee 40:42:

Well, you know, I don't even know if it has anything to do with human nature, right. It might just be the way that modern society is structured. So, who knows? I mean, I think there's a real question here about what are the conditions required for armed conflict? Why does that happen? And, you know, could you get, could you develop a system where it just doesn't and totally avoid this whole issue? And I think it's actually quite complicated because conflict is associated with so many things. You know what I mean? It's not just like conflict by itself.

 

Eddie Lee 41:13:

Conflict comes with famine, comes with disease, comes with poverty, comes with deaths. I mean, you're, you're basically tearing a lot of things apart. So, it's just one side of the coin. And so maybe, you know, it's not pleasers you need, right. Maybe it's something else. Maybe you just need to change how people see one another.

 

Michael Garfield 41:34:

Yeah. At risk of being misunderstood as making a political statement here. I've been thinking a lot after talking with Jeff West about his work on, you know, the branching of the circulatory system as being a way of minimizing friction, minimizing turbulence in the bloodstream. And so, if we're going to apply that kind of biophysical scaling law insight to the way that we have established our systems of governance, then the question is, well, what does a truly fractal policing system look like?

 

Michael Garfield 42:06:

Because I used to work in festivals and, you know, there is this question of, if someone is trying to get into a fight at a festival, is it really appropriate to call the police? Can this be handled like burning man does with volunteer Rangers that are trained in de-escalation before it becomes a legal issue? And so, you know, this is, I think that your work is really inspiring for anyone who's interested in the conversation around police reform and for, you know, the kind of dynamic governance that I talked about with David in our transmission series about how many layers of modular decision making are necessary in order to get something that is both robust and isn't going to just implode periodically.

 

Eddie Lee 42:54:

Right, right. Yeah. I think, I think there were some big questions here, right. That goes far beyond the political issues at play today. Right. I mean, the questions that we're asking about complex systems are big questions about the system and might even question the sort of principles of the system. If you can get sort of a similar outcome with a completely or sorry, a different outcome with a completely different setup, that'd be really interesting. You know, I mean, especially in the context of armed conflict, because it really does seem for at least a couple of hundred years of innate features of the system.

 

Eddie Lee 43:31:

I mean what is it about the system that you can change? Can you change it? Is it human nature? Is it our construction and modern society? I think those are questions that haven't been answered and somehow, they're connected to, by the fact that conflict is associated with many of these other things, they're really connected with serious social problems that still have played to human society for a while. So, I think there is a chance here really by digging into one of these aspects, potentially being able to get an understanding of some of these broader social issues.

 

Michael Garfield 44:07:

So, this is the place for us to pivot, forgive the pun here into a discussion of the paper that you lead authored for Royal society interface on this is sensitivity of collective outcomes, identifies pivotal components. You know, we're coming up on the presidential election, the conversation around swing voters. And everybody wants to figure out who are the people to target that ended up sending the sand pile down one side of the Hill versus the other, right.

 

Michael Garfield 44:40:

So, could you unpack this particular paper and a little bit of the, you know, the sort of fundamental framing that you're doing here and, you know, get into the details on that place. Yeah.

 

Eddie Lee 44:52:

So, what we did here was we started with this idea of the median voter. So, the median voters, basically the person who sits in the middle, right, that's a swing better. So, imagine you had a bunch of people, let's imagine the simplest situation where you could just label them left and right. But then there's an odd number of people you decide by majority. And there's the person sending the middle, that's the median voter. Now that's a very simplified situation. And I think many situations are often simplified into this binary choice, which is not the case because there were so many things that actually come into play in a vote and across many votes, actually, right?

 

Eddie Lee 45:31:

Right? So, you never just have a median voter. The voter might be immediate on one issue, but not on any others and you might have variation, right? So, imagine that this bill is an obvious left, right? But you want it to pass, you know, what their median is going to do, then you change it. You make a more complex bill with many amendments and so on, right? So, you can play with this in many ways. And so, as a result, you're not really asking, or we weren't really asking what is the median voter, but once we account for all these sorts of potential complexities that manifest in the statistics, how could we identify that person who sort of plays that role more consistently than others?

 

Eddie Lee 46:11:

So, a little bit more of a subtle definition of median. And we call these pivotal voters who pivotal components. And this is a very general way of asking the problem, right? So, it's to say that if I were to push and tug on each of these people in different ways, how would the outcome change in the voting system? The outcome is the majority votes or different ways that people couldn't divide up into groups. But again, you can think about this and in different contexts, right? So, Twitter is an example. So, you imagine everyone sort of voting in a different way by choosing to use certain words.

 

Eddie Lee 46:43:

So, the vote is not just left or right, but it's whether I used it, I talked about this K-pop idol or this K-pop idol right. So, you make a choice about which community you start to belong to in terms of the words you use. And then we asked, like, if some people were to change their words, would that really substantially change the collective groupings of, of other people? And you can ask a similar question about stocks in that case. It's a little bit less about what you can change, but more of a sort of identification principle, right?

 

Eddie Lee 47:14:

Because it's very hard to change. For example, the, what was it, the SNP spider indices, you don't, you don't just go in and change the data. Those are an aggregation of many different things happening. So, it's more, just a sort of a way of understanding how they move with respect to one another. So, we took this sort of generalized idea of a median and try to understand, for example, and in congressional voting or Supreme court voting, could we identify these people that you would want to tug or push?

 

Eddie Lee 47:46:

And we actually asked quite a bit, sort of a general question. We didn't ask if there's one person, what if you could control anybody? Right. So, I could pay everybody a bribe, but to go in different directions. And the goal would be to change the outcome as much as possible. And what we found was that in some cases we found certain individuals that seem to matter a lot. And some cases we found that it actually wasn't one individual actually you'd have to sort of control the entire system to change things.

 

Eddie Lee 48:17:

And that's sort of interesting because you know, it sorts of hints at this idea of manipulation, right? So, a system where you have to push a lot of different things at once, you have to push all the buttons in different ways, all at the same time, that's hard, that's complex. Whereas if you just paid one guy and you get all the votes your way. Well, great. I know exactly what to do. So, we sort of look for that signal. And I, I guess maybe I'll point out one of the examples that we looked at, which was the Supreme court, and this is not the modern court, this is the Rehnquist, the second run core score, which was from 94 to 2005 when William Rehnquist was a chief justice.

 

Eddie Lee 48:55:

And what we found was that the primary signal was not about O'Connor and Kennedy, who are presumably the median voters that matter the most and that's, that's conventional wisdom. And that was really interesting. Right? So, suggest that there's something more complicated going on. I think given some of the work I've done, that's not so surprising, but as it turns out that it's a little bit more subtle, right? It's not that power is totally diffused throughout the court.

 

Eddie Lee 49:26:

If you ask whether courts, whether a vote was liberal conservative, and there's some ambiguity here in how you determine whether something is liberal conservative, but sort of pushing those aside for the moment. If you just ask whether the decision was conservative or liberal, and you asked who was the sort of most influential, according to this measure, then you do find O'Connor and Kennedy. So somehow the fact that you can identify them as being important voters hinges on your interpretation of their votes as partisan.

 

Eddie Lee 49:58:

If you don't, if you don't, if you forget about the partisan, you just say they voted in these ways, then they don't, they don't seem to be important. So somehow, right, the statistics are telling us something potentially very interesting, which is that influence is not only a measure of the changes that someone can impose, but your interpretation of those changes as well.

 

Michael Garfield 50:23:

Gosh, you know, I'm just thinking about this in light of research that was done right before the 2016 us presidential election on search result ranking. I forget who it was that published this, but they were talking about the difference in a news item, making it to the top search result versus the second search result and how they were able to swing by modulating that across the entire population of people actually issuing these searches.

 

Michael Garfield 50:53:

They were able to swing electoral results by up to 25% in one direction or the other. And they actually tested this on some mayoral elections in India, where they were able to demonstrate that they were, they were able to throw, and they didn't actually do it. Oh, I was okay with it, it was a retrodiction of results that had happened, like a matter of weeks before. Yeah. But they were, I mean, they basically said, look, you know, the search companies need to be aware that the structure of accountability within these organizations, when you have an opaque algorithm for search results and the company, they can throw somebody under the bus who was working under nondisclosure on the algorithm and say, Oh, this was a rogue agent.

 

Michael Garfield 51:40:

We lack effective safeguards in society to prevent this kind of thing from happening because the kind of manipulation at scale that you're talking about does seem possible.

 

Eddie Lee 51:51:

Right. So, what you're saying is that it's not just the facts that you give people, but the salience of those facts. Yeah. And that's, that's actually very subtle because, you know, I certainly actually, I, you know, it was just last, was it this past week I heard a talk about this, but I just can't for the gut of me remember, but something about salient being really important for determining how people interpret events around them.

 

Michael Garfield 52:18:

So, there's another dimension of this paper, in your discussion on this paper, you're talking about this, not only in terms of voting results, but in the susceptibility of population to disease or disinformation. And so, you know, David and Jeffrey and many other people at SFI have written about, and we, you know, we had Laura Hibbard, the friend and the scarp, you know, on the show also recently talking about cultural contagions. Yeah. And so, I'm trying to wrap my head around how this model could inform a strategy for fighting disinformation in this way. And like, what would that look like?

 

Eddie Lee 53:02:

Well, you know, it's, it's kind of hard to make that connection. I think formally because we're looking at the statistics. So, we look over many aggregate dynamics and so on. And I think probably for something like this information who really do want to understand the dynamics of how things are spreading. So, there is that sort of shortcoming, but you could still ask, I think in that context, you know, how changes in the system could result or what changes in the system could result in the largest changes in the outcomes.

 

Eddie Lee 53:35:

And I think sort of to do that, you sort of have to rely on the fact we're focusing on this mathematical technique for, for studying the response of a system to perturbation. So, we're asking, we're sort of thinking about that in a statistical sense. It doesn't have to be in a statistical sense, but actually it sort of connects to a really interesting set of ideas of sort of tangential to this. And what it should say is this, this field right, is relying on this idea of information geometry.

 

Eddie Lee 54:07:

And what information geometry is about is imagining that if you have some sort of mathematical description of a system, it has to be parameterized. You typically choose parameters in a way that makes sense relative to what you're studying, but you could imagine changing those parameters. And you could imagine an alternate universe where a different sort of dynamics range. And so, this idea is that these models are all connected to one another. If you change the parameters a little bit at a time, eventually you get to those crazy alternative universes, which is totally different.

 

Eddie Lee 54:42:

But the path that you take is described by information geometry. And what we say is the curvature of, of that geometry. So how quickly it changes tells us how sensitive the system is, right? So, if I change this parameter a tiny amount and all of a sudden, I am in this alternative universe, then it's highly sensitive. It's highly sensitive to perturbation, right? Whereas if it's totally insensitive, two changes, then this primary almost doesn't matter, right?

 

Eddie Lee 55:14:

You basically get the same universe. It's slightly different. Maybe the colors of the cars are not quite the same, but it's almost exactly the same and what it turned out to be the case in a lot of physical models of the world is that, and you have many parameters that actually don't matter. And you have a few parameters that really do. So, for example, in the biology of a cell, you have these sophisticated bottles with tons of parameters that you can never hope to measure.

 

Eddie Lee 55:44:

And in fact, you can't really measure them anyway. Okay. But if you plug those values into the mathematical model, you basically get a cell that works. So, this is incredible because you're sometimes off by a hundred percent, 200%, but it's the works. Why it's because the information geometry is actually multi-dimensional and in some directions of parameter space, it's totally flat. So, all of the universes look exactly the same along that dimension. And it's very, very sharp along some other dimensions. So, you have to get some things right. And those things matter a lot, but most things don't matter.

 

Eddie Lee 56:15:

And so, this idea that you have this hierarchy of sensitivity, right, is what allows us to understand that the world, because if everything mattered, everything's out the window, you can never do that. You can never understand nature, but the fact that there's a hierarchy means that I can first build a model that sort of right, right. Newton's model or Copernicus's model, it's sort of right. It's not exactly right, but it's good enough that people believe me. And then, you know, Newton comes by and then Einstein comes by. But all of those additional effects features are smaller and smaller.

 

Eddie Lee 56:45:

And so that's what I mean by hierarchy of parameters. Right? And so that's what allows us to, well, that's a claim, but you know, that's, that's what makes the universe easy to learn about. And hopefully that is true. Generally. It's not, it's not clear that's true for every complex system, right. Which is why the complex systems are hard, difficult, but what's sort of interesting in the context of pivotal components is if there is such a hierarchy, then it means that this system is easier to control, right? Because you don't need every degree of freedom.

 

Eddie Lee 57:16:

So perhaps in order to facilitate diffusion of power, you want to design systems that are completely impossible to control unless you have all the fingers on the right triggers at the right time. Whereas if you want it to be controllable, then you need to design it with this idea in mind. And you can imagine situations where you want A, and you don't want B or you want controllability and you don't want non-controllability. And that's sort of the connection. There's a long-winded way of getting to this connection with, you know, susceptibility in say, disinformation is one might want to ask, how are these systems designed and are they designed in a way that makes them easily manipulated or makes them hard to manipulate?

 

Eddie Lee 58:03:

And do you want one or the other? Because you know, it could be that disinformation is really hard to root out when things are very decentralized, right? So, it may actually be a clash of values, right? You may actually want one thing, but it just leads to the wrong thing.

 

Michael Garfield 58:19:

I'm reminded of a, was it around the 2004 presidential election? Someone had a video where they taught a chimpanzee to manipulate one of the Diebold voting machines. You know, you can like, you can teach a Chimp to hack this, which brings us back all the way back around to the primate stuff. Yeah. You know, as somebody who has a lifelong abiding interest in both the sort of philosophy around evolutionary theory, as well as time travel fiction, you know, these, what you just described are the two sort of worldviews that you see at war in the debate over evolutionary contingency versus inevitability, you know, it's, it's ultimately, it's a dispute between two different models.

 

Michael Garfield 58:56:

Like, do you have Ashton Kutcher's butterfly effect, time travel where every, you know, like you keep changing everything on accident or is it more like back to the future where it only matters who sleeps together? You know, that's, that's, you know, somehow that, you know, the timeline is like completely robust against these perturbations. And so, you know, that's, I guess until we have this figured out, my advice is to like, not get in the DeLorean. Right. I don't know. Well, dude, it's, it's been absolutely wonderful talking with you before we wrapped this.

 

Michael Garfield 59:31:

I just an opportunity to tell people a little bit more about what's on your plate right now as a researcher and what you're looking forward to, what are the burning questions for you right now? And, on the horizon, in the months of confinement to come here,

 

Eddie Lee 59:52:

Well, I'm, I'm writing the second Principia Mathematica, you know, because people keep saying, you know, this is the time when you really get some work done. And I'm sort of past that now. It's been so many months?

 

Michael Garfield:

And it's horrible.

 

Eddie Lee 1:00:12:

If you keep telling people you're working on this grand project, nothing comes out while you're still working on it. No, I have, I have several things on my plate right now that are sort of on and off. One thing is sort of extending this armed conflict stuff to start really digging into some of the, sort of maybe contingent factors, but trying to find patterns in the contingencies that unify them. We're also been looking at Jeffrey West Chris campus. And I have been looking at metabolic scaling theory or in the forest. So, trying to understand some of the dynamics that lead to scaling and forest populations and hopefully actually touch on armed conflict in some way.

 

Eddie Lee 1:00:52:

I've also been thinking about acquisition of information in firms. So, trying to see why is it that firm lifetimes are distributed in a very sort of regular way, surprisingly across sectors. And does it have to do with how they're learning, how they're acquiring information from around them. I'm finishing a project with Jess and David on adaptation learning and adaptation. And it actually, the idea is that we're thinking about different forms of memory embedded in either the organism in its behavior or in its environment, and trying to unify these various mechanisms or implementations of memory into a sort of optimal adaptation framework.

 

Eddie Lee 1:01:36:

I'm sure I'm forgetting something, but it's a lot and, um.

 

Michael Garfield 1:01:39:

You don't have to keep fighting it. Right. If you've forgotten it, if you've heard that ??.

 

Eddie Lee 1:01:47:

There's no glue, you know, at some point it will lead to some cascading failure.

 

Michael Garfield 1:01:52:

Fair enough. Well, yeah, thanks again for being on the show and folks check the show notes for a link to all of your research and the other stuff that we've discussed in here.

 

Eddie Lee 1:02:06:

Awesome. It was fun.

 

[Outro]

Michael Garfield 1:02:08:

Thank you for listening. Complexity is produced by the Santa Fe Institute. A non-profit hub of complex systems science located in the high desert of New Mexico. For more information including transcripts, research links and educational sources or to support our science and communication efforts, visit SantaFe.edu/podcast.Complexity Podcast

Fractal Conflicts & Swing Voters with Eddie Lee

 

[Intro] 00:00 :

You know if you have a forest with a lot of dry wood that hasn’t been cleared away, by previous small conflagration you can have a major conflagration, right? I am not saying that, that is conflict. But you could imagine some kind of rule that a regulator would perform in society. I don’t know what that rule is, but definitely one of the ways it manifests is by preventing the largest outbreaks from happening. Or, an attempt to.

And, you know, one of the ideas behind some of the models that have been proposed for conflict is actually similar to this forest fire thing. Which is that, the conditions for conflict will simmer, and when it reaches that boiling, you can have one. And when you have one, it’s this big thing but it depletes the system in a way.

And so, you are always poised between never being able to have these big conflicts and having these sorts of having like these smaller conflicts. Would it be good to prevent all conflicts or would it be bad? Would that sort of lead to some sort of special situation where you suddenly had the nuclear Armageddon? I don’t know.

 

[Intro music] 01:00 :

 

Michael Garfield 01:24:

Since the 1940s scientists have puzzled over a curious finding. Armed conflict data reveals that human battles obey a power law distribution like avalanches and epidemics. Just like the fractal surfaces of mountains and cauliflowers, the shape of violence looks the same at any level of magnification.

Beyond the particulars of why we fight, this pattern suggests a deep hidden order in the physical laws governing society. And digging in the new analysis of data from both armed conflicts and voting patterns, complex systems researchers have started to identify the so-called pivotal components involved. The straw that breaks the camel’s back. The spark that sets the forest fire. The influential but not always famous figures that shape history.

Michael Garfield 02:11

Can science find a universal theory that predicts the size of conflict from their initial conditions? Or identity key players whose knob society in one direction or another?

 

Welcome to Complexity. The official podcast of the Santé Fe Institute.

I am your host Michael Garfield. And each week we’ll bring you with us on far ranging on rigorous researchers developing new frameworks explaining the deepest mysteries of the universe.

 

Michael Garfield 02:39:

This week’s guest is SFI program postdoctoral Eddie Lee. Whose work in conflict avalanches and swing voters gives a glimpse of the mysterious forces that determine why we fight. And how we will be able to prevent the next conflagration.

 

In this episode we talk about armed conflict as a fractal and a form of computation. Swing voters in the justice and influencers in pop cultures. And what these studies may have to say about the deep constraints that guide the currents of society.

 

Michael Garfield  03:11:

Just a note that this will be our last episode before short summer break to give our scientists uninterrupted time to work on a torrent of new research. We have some exciting episodes scheduled for our return in mid-august. In the meantime, please be sure to subscribe to complexity podcast on your favorite podcast provider to make sure that you stay in the know. And if you value our research and communications please consider a donation at SantaFe.Edu/podcastgive or join our Applied Complexity Network santafe.edu/action

 

Michael Garfield  03:43:

Lastly, we are excited to announce that submissions are open for this fall's inaugural complexity interactive. A three-week online project based immersive course where you get a rare opportunity for mentorship by a large faculty of SFI faculty, including Simon DeDeo, Daniel Bassett, Melanie Moses, Ricard Solé, and many more.

For more info and to apply, please visit SantaFe.edu/sfi-ci

 

Michael Garfield 04:23:

All right. Well, shall we

 

Eddie Lee 04:25:

All right, let's do it.

 

Michael Garfield 04:26:

Yeah. Eddie Lee, it is a pleasure to have you here and our first ever in person, socially distanced podcast recording here across my patio. So welcome.

 

Eddie Lee 04:39 :

Thanks for having me. This is great.

 

Michael Garfield 04:42 :

I'd love to start as we typically do by just providing a little bit of personal background and having you talk a little bit about how you got into science and how you got into specifically an interest in the kind of research that you're doing at SFI and the stuff that we'll be talking about today.

 

Eddie Lee 05:00 :

Well, I think it was last week that actually the writer of the magic school bus series passed away. And, you know, I have to say that she really did leave a fantastic legacy and I'm part of that legacy. I grew up on, on the magic school bus and actually some after school science programs as well. So, I sort of had this nascent interest in science for a long time, something I enjoyed doing. I didn't really see myself becoming a scientist.

 

Eddie Lee 05:32:

And in fact, when I went to college, I actually tried to become an economics major surprisingly enough, but it turned out that I took this sort of fantastic course called integrated sciences. And it's sort of my preview in a way to complex sciences, because what it was a class telling us, showing us that the way of thinking in computer science, biology, chemistry, physics, we're all coming together in the context of some fascinating problems in biology.

 

Eddie Lee 06:03:

So, this idea that all these things were coming together was a really surprising thing for me. And I think what really sort of hit it home was when I read this book by Phillip ball called critical mass. And so, he brought up the aspect of society, also being part of this. And, you know, at some point in my past, I'd also read Isaac Asimov's foundation series, which is, which is the physics of social behavior and prediction. And so somehow all of these things were sort of swirling around in my head.

 

Eddie Lee 06:35:

I didn't know what to do about them. And I ended up imagining what I really need to do is go into sociology. So, I went to, I searched around, I'd try to talk with sociology professors. And I just talked with one, I think he was maybe at the head of the Taryn at the time. And he said, you know, I know this guy who might be able to direct you. And he couldn't remember the name, but he started searching online and he showed me the web page. And it was one of the professors who had taught that first class, that first course that I mentioned integrated sciences.

 

Eddie Lee 07:05:

So, he sorts of just sent me right back. Yeah. Boomerang style back into that professor's office. And he actually was familiar with SFI. So, he's the one who directed me to talk with Jessica and Dave Cracker, who now I've known for a long time. And that's sort of where all of this stuff started happening. And that's when I started really learning about what sort of science SFI works on and my sort of vague interest in it became, you know, crystallized.

 

Michael Garfield 07:37:

Awesome. Well, this is a great place actually, to just dig right in to the first set of papers I wanted to talk with you about today, which you co-authored with David and Jessica, as well as Brian Daniels and Chris Myers on scaling theory for armed conflict avalanches. So this idea that we can use physics insights to understand human social behavior from like an orbital perspective and like really much like in the foundation books, you know, that we can, we can come at the, the sort of vagaries of history from, you know, David would call a rigorous and principled quantitative approach.

 

Michael Garfield 08:23:

The history is not just the story of great men making decisions, but that those decisions occur within a landscape of physical constraints. And what are those constraints and how do they manifest themselves? And this is really interesting work and it's got a long history that you're building on. And so maybe the history of this strain of research is the right place to start. And then we can unpack it from there.

 

Eddie Lee 08:48:

Sure. Yeah. Like you're saying, it's interesting, but what we weren't focusing on was the particular stories of particular conflicts. And there's a lot to study about any particular conflict they're quite different, right? So, you might think of say the Libyan or Tunisian revolutions happening today, which are quite different from world war two, which were quite different from the British and French battling it out, say in 1812. So, these are all very distinct lines of history in and of their own.

 

Eddie Lee 09:20:

And so, what we were interested in was trying to understand whether or not there were universalities common features shared amongst these things that are clearly disparate. And as you were saying, this isn't a completely new idea. It was in 1948 that Louis Fry Richardson showed this amazing feature, which was that when you make a histogram of interstate Wars, so you count how many wars there are with F people that died, F fatalities. And you look at this histogram on a log plot, what you find is a straight line.

 

Eddie Lee 09:54:

And what's remarkable about a straight line being on a log plot is that it's, that's a power law. That means its scale free. In other words, sort of most, very naively speaking. If you were to look at a small conflict, it's sort of like a shrunk down version, statistically, speaking of a large conflict and so on and so forth all the way down. And so, this was very strange because interstate wars that you looked at were all quite different from one another. So why would you get this sort of regularity that emerges when you look across many different conflicts?

 

Eddie Lee 10:27:

And since then other people have noticed other interesting patterns in the timing of conflicts, as well, as well as terrorism and so on. And what we were interested in was looking for some of these patterns in a more recent dataset, which is called the armed conflict location event, data project, or ACLED for short. And what was remarkable about this data set was that they didn't take all these events, these small disparate individual conflict events that would occur in a war. They didn't group them together into wars.

 

Eddie Lee 11:00:

They just sort of disaggregated them into, into small localized defense. And so instead of having to take a quantity, that's already defined for you as a battle or a war or skirmish, or just a local riot, we got to connect these into ourselves, into these clusters of what we called conflict avalanches. So many ways similar to two battles or Wars, but without having assumed some of these connections from using sociopolitical factors. So just looking at the time and spatial patterns of these events, and what we found was that when you look at these conflict avalanches, various measures of their size, so not just how many people died, but how many of these reports did you obtain?

 

Eddie Lee 11:43:

How long did it last, how far did it spread? All of these different factors show these characteristic power law tales, like what Richardson saw and what we were able to do beyond that is to show that these pieces were all connected to one another in some mathematical way. And I think I had an analogy for this, which was that if you imagine you walk into a dark room and you're trying to figure out what is this object in front of you?

 

Eddie Lee 12:12:

And you find first, there's this sort of rigid pillar. Then you find this long sort of thing with a hairy end. You find this floppy disc, like thing, you start in your mind developing this picture of an elephant. And the way that that works is all these pieces are connected to one another, right? They're pieces of one whole. They're not independent, separate things. And so, you build a model in your head and you say, this must be connected. The legs must be connected to the body, must be connected to the head and so on.

 

Eddie Lee 12:43:

So, we did something like that with the mathematics of these various features. And we found that they were self-consistent in a way that seems to point to one sort of object driving the dynamics.

 

 

Eddie Lee 12:55:

And that's sort of really cool because it suggests that there is some sort of underlying thing that we're just getting hints of. We haven't found it yet. We haven't discovered exactly what it is. Obviously, it's something about war and conflict, but it's perhaps something bigger, right? That leads to universal patterns, across many different types of conflict.

 

Michael Garfield 13:17:

So, you draw some analogies in one of the preprints that you've done on this work, emergent regularities and scaling and armed conflict data. You draw analogies to forest fire models, which other researchers have tried to apply to this, to neural avalanches. And, you know, just from last week's episode with the SFI counter speech team and the network graphs that you see of the Twitter data that they've harvested, and the way that you see these conversations, these sort of battles between hate speech, organized, hate speech groups and organized counter speech groups on Twitter.

 

Michael Garfield 13:56:

And this look also like these, you know, you see these the sort of conflagration of point by point debate breaking out when there's a successful, an actual collision, rather than just the hate group eliciting engagement to its original post. But people are actually willing to go into the trenches with one another on social media. And so, I would love to hear you talk a little bit more specifically about this model as a, like a branching network diagram.

 

Michael Garfield 14:26:

Yeah. There's some really interesting figures, and we'll link to this stuff in the show notes so that people can, can actually get a look at what we're talking about here. Because it actually looks rather menacing.

 

Eddie Lee 14:39:

That's why we chose those colors.

 

Michael Garfield 14:41:

The map of a branching armed conflict here.

 

Eddie Lee 14:44:

Yeah. So, cascades are a real central conceptual foundation for understanding a lot of these different kinds of phenomena. And as you mentioned, if you look at neural activity in the cortex, what you find are, are these cascades of activity, and it's not just a bunch of neurons firing, you can actually trace it out to some sort of branching process. So, you imagine there's the center, the first neuron that activates other neurons, which activate other neurons and so on sort of in a conflagration type picture.

 

Eddie Lee 15:24:

And you can also think about various other phenomena with this sort of conceptual framework. So, forest fires are another way, right? So, you imagine you have a bunch of sorts of local clusters of maybe dried out patches of trees and you have a spark, maybe some lightning strikes, and then it sort of burns its way through all these connected clusters. And you can think about that analogy applying to, to social contagion as well. And so that's why we sort of started with this idea of thinking about conflict in the same way.

 

Eddie Lee 15:55:

And mathematically, it's a very generic way of thinking about it, right? All of that you need are different pieces that are connected to one another in this sort of branching way. And so, you can get this exponential increase. Actually, this is very relevant in the context of today with COVID right? So, when people talk about exponentially growing processes, this is exactly what they mean. Or you have someone who spreads it to some other people. And if each person on average spreads up to more than one person, then you have an explosion. So, we sort of were playing with this framework and thinking about conflict in this way as a social contagion.

 

Eddie Lee 16:29:

And we're not sure if that's exactly how it spreads, but we can at least say that the way that it spreads is consistent with this idea of the way that it spreads both across space, as well as across time and in size.

 

Michael Garfield 16:46:

So, this elicits for me, links back to this sort of giant network model and building of this, these conversations on the show with all of the other conversations that hopefully we'll publish in the SFI press one day and make a poster. This reminds me a lot of, of two things. One is the conversations I just had with Jeffrey West a few episodes ago, where he was talking about biophysical scaling laws as being extremely coarse and like not actually capturing all of the variation that we would see in like specific evolutionary instances. The Primo example, being that human beings, modern humans use something like 30 times more energetic resources than we would expect from mammals of this size.

 

Michael Garfield 17:33:

Because as we've talked about with David and some of the transmission episodes, the human being is basically just the visible corner of this hyper object now. And then each of us is actually like a cyborg spanning the globe with all of these electronic augmentations. Right? So, something like that is going on in this work with conflict. And when you actually peer into the data with a little bit more granularity, you see a lot of variation in this, not enough to challenge the thesis, but enough to suggest that there are some interesting details about the regional and temporal variations of these conflicts.

 

Michael Garfield 18:14:

And I would love to hear you talk about that. And then also just to like make this kind of unwieldy let's pin onto that since you brought up COVID, Chris Morris piece for the transmission series, where you talking about super spreader and that's and how the, you know, our zero as a measure of the transmissibility of a disease, doesn't actually tell you what's going on in that particular church or in that particular campus building or, you know, so there's. So, what are the features about the data set that you've noticed that seemed to be providing insights into why there's variations in the size of these conflicts around that mean.

 

Eddie Lee 18:56:

No, that's, that's great. I mean, I think it's really important to point out that a lot of this work 's goal isn't to explain all this variation, which, you know, eventually we want to do, but it's really to talk about the shared common features across all of these phenomena. And then of course it's important to mention that, that's the only thing it explains. And so, the features that I'm talking about in terms of scaling are the averages, but you can have scaling in the variation as well.

 

Eddie Lee 19:29:

And that's, that's really interesting. And as you mentioned, right, you know, Chris Moore wrote that very nice piece where it's really important to mention that are not, is an average and not necessarily going to tell you what's happening in a very particular location. And that's also exactly true for theories about metabolic scaling in force or conflict. And in fact, it turns out that that sort of variation manifests directly in our work looking at conflict.

 

Eddie Lee 20:02:

And one of the things that we find is that it's not only essential to account for the similarities between how conflict grows in different locations, but there's also variation in terms of this sort of coefficient in front of it. So, the way that it grows the shape of the curve may be similar, but the offset of that curve may be different. And that might depend on factors such as prosperity or governance. So, here's one example, right? So, you look at Eastern Somalia, weak governance, a lot of widespread poverty and a huge amount of conflict.

 

Eddie Lee 20:36:

One of the hot zones in this and the status set. But if you look at South Africa, it's quite prosperous, they have a pretty strong government and its relatively low levels of conflict. So even if it's the case, that conflict evolves in a similar way, you just don't expect to see as much conflict and South Africa, as you see in Eastern Somalia. And so, as you also said, these things will vary in time. So hopefully Eastern Somalia is not a hotspot forever, but maybe somewhere else will become one. So, there may be patterns also in the diversity of hot zones.

 

Eddie Lee 21:09:

And we see a hint of this. We can't say that will be the case forever because we only look at 20 years. But one of the sort of fascinating things is that Richardson, who I mentioned earlier, looked at wars between, I think, early 1800s and mid-1900s. And he found basically the same statistics. Aaron?? looked at the statistics all the way from Richardson's asset up to today and found the same statistics. So, there's some sort of weird preservation of conflict despite changing technology, different countries. And, and so on that either, I don't know, is, should you be amazed or should you be afraid?

 

 

Michael Garfield 21:47:

Well, I mean, again, two things come up here, right? One is Jen Dunn's work on food webs and how the structure of these food webs that their patterns of connectivity have been amazingly well-preserved conserved in there, in their form for like the last 500 million years as we've gone through all of these regimes in what is actually living in the oceans. So that's part of it, you know, and that's, that's sort of begs the question that I asked Brian Arthur, which is, do you think that we can use these models as a way of predicting like basins for future innovation, you know, and like predicting areas where we can expect potentially larger conflicts than we're seeing.

 

Michael Garfield 22:30:

And then the other piece of it is, bringing in the geopolitical component, Peter Turchin's work. When we're talking about a quantitative study of history and his writing on what he called the double helix of inequality and social instability, where he saw that they were very strongly negatively correlated over the last a hundred or 150 years. But, you know, you make a point in this paper that there are other geopolitical and geographic features that have to do with the coastlines and national borders and locations of these conflicts relative to large urban centers' populations.

 

Michael Garfield 23:11:

Yeah. So, I, again, I know that this is sort of beyond the scope of the paper, but I'm really curious, you know, what other, what other factors you think are playing into this and how that might be guiding your follow-up research in this area?

 

Eddie Lee 23:23:

I think what you can say is despite the fact that there are a lot of these universal patterns that we think we find is that there's a lot of work trying to understand the mechanisms behind why these patterns might appear. And I think there is yet a connection to be made there. And that would be sort of, I think the Holy grail for, for this sort of work is really understanding those connections and flushing them out.

 

Eddie Lee 23:53:

And somehow, they lie in the intersection of maybe social mobility, social prosperity, economic prosperity, technology, geography, right. You can't really fight someone if there's no one there,

 

Michael Garfield 24:08:

Tell that to our Facebook group.

 

Eddie Lee 24:13:

Yeah. I guess you could make up stories and so on, but there are a lot of different factors that somehow mashed together to generate these patterns. But as you were mentioning, Jen Dunn's work. That's not necessarily to say that you can't get universal patterns from that. So, you know, this is maybe too simple of an example, but one sort of that comes to mind right now, which is the central limit theorem, right? There are many reasons for why you get random likes statistics, right?

 

Eddie Lee 24:46:

Many different ways of generating them, but by virtue of their randomness, you end up getting some sort of regularity. So, we'll see. Maybe it's the fact that there are so many factors that do influence conflict that ended up getting simplification at higher scales, or maybe not, maybe, maybe it is contingent. And we, we just don't have enough granularity in our understanding to access that contingency

 

Michael Garfield 25:15:

Later on, in this paper, you talked about there being, you know, getting into this granularity, you say: "Unlike canonical cascade models, conflict also includes lattice style dynamics that evolve with geographic spread. The suppression of these dynamics away from the core could reflect social processes or geography that impact conflict evolution. Furthermore, our model suggests conflict is not only the result of local correlations and activity, but also regional and temporal disorder, perhaps reflecting memory of the severity of initiating events".

 

Michael Garfield 25:47:

So, this links directly to another paper that you wrote with Brian, David and Jessica on primate conflicts and temporal scaling collapse. And the idea, you know, Jessica's work on primate conflicts as a form of collective computation and the severity of those conflicts having to do with various factors that are performing computations at the level of like an entire primate troupe, right? So, this notion that the patterns that you're seeing in conflict data, that they hinge on there being certain evolutionary reasons for the distribution of conflict, duration, and severity is really interesting. I'm curious to hear you talk a little bit about this other paper and how it links in.

 

Eddie Lee 26:36:

Right. No, that's, that's interesting. Right? What are the functional properties of conflict, right? Is there a reason why conflict plays out the way that it does? It's hard to say for human conflicts. Obviously, there are reasons why people give for starting wars. But in monkey conflict, there's potentially one reason which is, has to do with a social hierarchy, right? So, you need to have a little bit of disorder and conflict in order to establish people's roles and where they sit.

 

Eddie Lee 27:07:

And the severity of the conflict at the same time cannot be too, too big, right. If it's too severe, it's actually detrimental to society. So, there's this idea that maybe what conflicts does is it serves an information gathering acquisition role, as long as it's not too bad,

 

Michael Garfield 27:26:

People have to survive, remember.

 

Eddie Lee 27:28:

Exactly. If you don't, if you know, if it's nuclear Armageddon again and that's it, we should remember that. But, I think one of the things that we were trying to study in that work was characterizing the features of monkey conflict and perhaps extracting from those features a potential functional role that, that these things, or these observations could, could conserve.

 

Eddie Lee 27:53:

And actually, what's really interesting about the monkey conflict is we studied a group, a society of ??. What we found was that the duration that conflicts lasted well, beta distribution, it was not a parallel distribution like human conflict, but how long they lasted that distribution of duration seemed to be a distribution that looked the same, whether you looked at small conflicts or large complex. And it's very strange because conflicts with 2 monkeys are not the same thing as conflicts with 10 monkeys.

 

Eddie Lee 28:24:

There are many more ways they can interact. There are many other things that could happen, go wrong. And so, this was very curious. And so what we found was that if you could think of these conflicts as being sort of the time for all the pairs to resolve their differences, and each of those pairs, or took some amount of time to resolve their differences, then you could get something similar to what we found in the data with the condition that the pairs that resolve their differences later remembered the sort of intensity with which the pairs resolved their differences earlier.

 

Eddie Lee 29:03:

So, in other words, if the fight started off with some intense pairwise interactions, like, you know, you bit me, how could you bite me? And that's, that's actually very apparently a very aggressive interaction that the other monkeys will not tolerate.

 

Michael Garfield 29:18:

I remember preschool. Yeah.

 

Eddie Lee 29:19:

And so, you end up having these long correlations from the beginning to the end, and if that's true, then you end up getting these, these sort of long tails of conflict duration. So, you have many conflicts that tend to last longer. And for us, when we, that was an indication of what we call collective memory, right. This idea that the entire duration of the conflict from beginning to end, some remember at the beginning. And that's what we mean by we say in that paper.

 

Michael Garfield 30:00:

Well, okay. So, you know, this, I have the luxury here of kind of like going out on the plank and speculating. Yeah. But this seems to be possibly why certain human conflicts are just insane, like the Hatfield’s and McCoy’s, you know, that the feud has a sort of durability built into it because of the way that the history of a family is prioritized. And like the, you know, the bigger the conflict gets, it would seem like certainly there are factors that are keeping resolution that are like forestalling resolution, but on the other hand, the more abstract the conflict becomes for people, the easier it is to get over.

 

Michael Garfield 30:45:

Right. So, you know, this, this is sort of like, you know, I wonder what this approach to understanding this might have to say to, or link with other work on the way that conflict between human beings has changed as war has become industrialized and has been more and more about foreign conflicts rather than defending your own, your Athens or whatever. From there are certain things that would seem to be kind of holding us in the United States in a position of kind of unending foreign Wars.

 

Eddie Lee 31:20:

Right. Well, I think it's a little bit unfair to say that these Wars are external. Right. For many people, these Wars are personal, right? Yeah. And that's their life, but sort of touching on this, on this question of, you know, how is it that conflict persists or why is it that conflict persists? I think one thing we suspect from looking at these patterns is that these memories, these core correlations in a conflict, are expressed not only in the history that people maintain about themselves, but also in the geography of how it spreads.

 

Eddie Lee31:57:

So, it's quite possible that what encodes the history of the conflict is not just what we think right, the stories that we tell ourselves, but potentially also factors of the environment. Structures that we build. And some of these structures are really obvious, right. I mean, we built the nuclear arsenal, right. It's going to be there. Right. This is really interesting to me, it sort of connects to some of the work that David, Jess and I have been doing recently with adaptation and outsourcing a memory into the environment.

 

Eddie Lee 32:30:

Right. Which is this idea that there are biological organisms that intentionally or use and manipulate the environment to couple their bad memory or the behavior with longer timescales that they need in order to better adapt. So, ants do this by building these trails. It's unlikely that individual ants can really remember what they're doing, but over the collective activity of many, they can establish these very long-lived persistent trails. And they can use this trail to harness resources.

 

Michael Garfield: like Google calendar.

 

Eddie Lee 33:01:

Exactly. No, I mean, let me do exactly the same thing, right. I mean, David has that example of the notepad, right? Multiply three times, six times, five times, 2,374. I can't do it in my head, but give me a piece of paper. I can just, basically you short term memory expressed on paper to do that calculation very efficiently. And in addition to opening up new ways of doing that calculation, that aren't accessible in my head to myself. So, there is this question, right? Sort of going back to the functional properties of armed conflict, you know, what purpose does armed conflict solve?

 

Eddie Lee 33:32:

Are there ways that we are sort of driving our conflict ourselves by embedding into our environment that we don't know of, that we do know of. And I don't think those questions are necessarily answered. At least I haven't seen those answers, but would be really interesting to think about, I mean, right, this is just one other phenomenon in nature, right? It's not just us that fight. We do fight with certain technologies and armed human conflicts or especially cause a lot of fatalities compared to other organisms. So, we're quite brutal, but.

 

Michael Garfield 34:02:

You know, to, to bring up a marvelous work of science fiction, that seems to have some direct bearing on the insights that you just said. I've been in a book club recently discussing the science fiction trilogy Lilith's Brood by Octavia Butler. Should I read this? It's an extraordinary hit. It's a very, very relevant and kind of evergreen piece of work by the first science fiction author to ever become a MacArthur fellow. It just highly awarded a black female science fiction author, you know, just an amazing, amazing mind.

 

Michael Garfield 34:36:

And I've been on the tip of trying to read more black Sci-fi right now to get a better understanding of the world spaces disclosed thereby. And in this book, this book is about humans interacting kind of non-consensually with a race of aliens that comes to earth and finds us in the aftermath of a nuclear apocalypse. And once they're masters of bioengineering and they want to reboot our planet and nurse humankind and the biosphere back from extinction.

 

Michael Garfield 35:09:

But in order to do that, they kind of have to change the rules for human beings. And one of the things that they do is they refuse to give the humans that are going to repopulate earth, any writing materials, like it's part of their thing, that they all have genetically engineered eidetic memory. And so, they would rather engineer us to have perfect recall, then allow us to record history on paper.

 

Eddie Lee 35:37:

It sounds like a curse. You know, you know, that, I guess that was the thing for a long period of human history, right? Oral tradition, Homer, obviously, until it was written down. Yeah. What is it that keeps us fighting? Is it, is it literature? And that would be even, there'd be very tragic and somewhat ironic actually, because we're supposed to learn from history.

 

Michael Garfield 36:05:

But you know, in this paper, I think it's, like you said, it's the end of the conflict remembers the beginning. Yeah. There's another piece here, which is about the beginning. And I think that this might give us a hinge to get into some discussion about your other work on swing voting, which is really, you know, also very timely and interesting. You and your co authors on this piece, talk about policers or the police macaques. And they're being like thresholds at which they will or will not intervene in order to break up a conflict before it becomes a conflagration.

 

Michael Garfield 36:44:

And I'm curious to hear you talk a little bit about how thinking about conflict as a collective computation illuminates, why we may make a choice to mediate a conflict or to de escalate it or to allow it to run its course. Right? Yeah. So, what are your thoughts on all of that?

 

Eddie Lee 37:04:

Yeah. Well, first the disclaimer is that I'm not an expert on mechanics society. So, so that, you know, when I talk with Jess later, I can say that I said this, but yeah. So, these police or macaque or high-power individuals that, I mean, you know, in short other macaque don't want to mess with them. So, if they get in a fight, you just don't hit the policer or per usual. So, they're serving some sort of regulatory mechanism and maybe in the sort of conceptual abstract sense, regulatory control is very important in terms of outbreaks and contagion, right?

 

Eddie Lee 37:40:

So, for example, for example, and I'm not saying that this is macaque conflict, but you know, if you have a fire or a forest with a lot of dry wood that hasn't been cleared away by previous small conflagrations, you can have a major conflagration, right? So, you have this idea that by preventing all, but the largest of conflagrations you've sort of saved the system, allowing it to renew itself. I'm not saying that that is a macaque conflict, but you could imagine some kind of role, right. That, that a regulator would perform in this sort of society.

 

Eddie Lee 38:13:

I don't know what that role is, but definitely one of the ways that it manifests is by preventing these, the largest, you know, outbreaks from happening or an attempt to. And, you know, that is sort of interesting. I don't know what the computational rule of conflict in human society is, but you know, one of the ideas behind some of the models that have been proposed for conflict is actually similar to this forest fire thing, which is that the conditions for conflict will simmer, right?

 

Eddie Lee 38:47:

And then when it reaches that boiling point, you can have one and when you have one, it's this big thing. But it depletes the system in a way. And so, you are always poised between never being able to have these big conflicts. I mean, truly massive ones, you know, think world war five and, and having these sort of like smaller conflicts, but I don't know, I don't know what would be, right. Would it be good to prevent all conflicts or would it be bad?

 

Eddie Lee 39:18:

Would that sort of lead to some sort of special situation where you suddenly had, you know, the nuclear Armageddon and I don't know, but it is, it is sort of an interesting question to think about,

 

Michael Garfield 39:29:

You know, pull in a totally non-scientific perspective on this from a rather sketchy character. I remember Osho Rajneesh, the disgraced guru that brought his cult into Oregon, had a line in one of his books. He said, you know, be aware of the pacifist because they're sitting on a volcano and. And you know, race had, ?? D'Souza just gave one of the flash workshop talks last week on timescales and tradeoffs, where she was talking about this in terms of exactly what you just said about, you know, if you have small forest fires every once in a while, it prevents this big one.

 

Michael Garfield 40:09:

So, you know, the question is in our efforts to expunge conflict from the planet. Are we really doing ourselves a long range disservice, are we making a tradeoff That's actually making the system dramatically more brittle than it otherwise would be? And then this starts getting into sort of more archeological or historical perspectives on the role of organized sports as a way of, of, you know, channeling our aggression into something that's less likely to result in. I mean, you might flip a cop car if you win the championship.

 

Eddie Lee 40:42:

Well, you know, I don't even know if it has anything to do with human nature, right. It might just be the way that modern society is structured. So, who knows? I mean, I think there's a real question here about what are the conditions required for armed conflict? Why does that happen? And, you know, could you get, could you develop a system where it just doesn't and totally avoid this whole issue? And I think it's actually quite complicated because conflict is associated with so many things. You know what I mean? It's not just like conflict by itself.

 

Eddie Lee 41:13:

Conflict comes with famine, comes with disease, comes with poverty, comes with deaths. I mean, you're, you're basically tearing a lot of things apart. So, it's just one side of the coin. And so maybe, you know, it's not pleasers you need, right. Maybe it's something else. Maybe you just need to change how people see one another.

 

Michael Garfield 41:34:

Yeah. At risk of being misunderstood as making a political statement here. I've been thinking a lot after talking with Jeff West about his work on, you know, the branching of the circulatory system as being a way of minimizing friction, minimizing turbulence in the bloodstream. And so, if we're going to apply that kind of biophysical scaling law insight to the way that we have established our systems of governance, then the question is, well, what does a truly fractal policing system look like?

 

Michael Garfield 42:06:

Because I used to work in festivals and, you know, there is this question of, if someone is trying to get into a fight at a festival, is it really appropriate to call the police? Can this be handled like burning man does with volunteer Rangers that are trained in de-escalation before it becomes a legal issue? And so, you know, this is, I think that your work is really inspiring for anyone who's interested in the conversation around police reform and for, you know, the kind of dynamic governance that I talked about with David in our transmission series about how many layers of modular decision making are necessary in order to get something that is both robust and isn't going to just implode periodically.

 

Eddie Lee 42:54:

Right, right. Yeah. I think, I think there were some big questions here, right. That goes far beyond the political issues at play today. Right. I mean, the questions that we're asking about complex systems are big questions about the system and might even question the sort of principles of the system. If you can get sort of a similar outcome with a completely or sorry, a different outcome with a completely different setup, that'd be really interesting. You know, I mean, especially in the context of armed conflict, because it really does seem for at least a couple of hundred years of innate features of the system.

 

Eddie Lee 43:31:

I mean what is it about the system that you can change? Can you change it? Is it human nature? Is it our construction and modern society? I think those are questions that haven't been answered and somehow, they're connected to, by the fact that conflict is associated with many of these other things, they're really connected with serious social problems that still have played to human society for a while. So, I think there is a chance here really by digging into one of these aspects, potentially being able to get an understanding of some of these broader social issues.

 

Michael Garfield 44:07:

So, this is the place for us to pivot, forgive the pun here into a discussion of the paper that you lead authored for Royal society interface on this is sensitivity of collective outcomes, identifies pivotal components. You know, we're coming up on the presidential election, the conversation around swing voters. And everybody wants to figure out who are the people to target that ended up sending the sand pile down one side of the Hill versus the other, right.

 

Michael Garfield 44:40:

So, could you unpack this particular paper and a little bit of the, you know, the sort of fundamental framing that you're doing here and, you know, get into the details on that place. Yeah.

 

Eddie Lee 44:52:

So, what we did here was we started with this idea of the median voter. So, the median voters, basically the person who sits in the middle, right, that's a swing better. So, imagine you had a bunch of people, let's imagine the simplest situation where you could just label them left and right. But then there's an odd number of people you decide by majority. And there's the person sending the middle, that's the median voter. Now that's a very simplified situation. And I think many situations are often simplified into this binary choice, which is not the case because there were so many things that actually come into play in a vote and across many votes, actually, right?

 

Eddie Lee 45:31:

Right? So, you never just have a median voter. The voter might be immediate on one issue, but not on any others and you might have variation, right? So, imagine that this bill is an obvious left, right? But you want it to pass, you know, what their median is going to do, then you change it. You make a more complex bill with many amendments and so on, right? So, you can play with this in many ways. And so, as a result, you're not really asking, or we weren't really asking what is the median voter, but once we account for all these sorts of potential complexities that manifest in the statistics, how could we identify that person who sort of plays that role more consistently than others?

 

Eddie Lee 46:11:

So, a little bit more of a subtle definition of median. And we call these pivotal voters who pivotal components. And this is a very general way of asking the problem, right? So, it's to say that if I were to push and tug on each of these people in different ways, how would the outcome change in the voting system? The outcome is the majority votes or different ways that people couldn't divide up into groups. But again, you can think about this and in different contexts, right? So, Twitter is an example. So, you imagine everyone sort of voting in a different way by choosing to use certain words.

 

Eddie Lee 46:43:

So, the vote is not just left or right, but it's whether I used it, I talked about this K-pop idol or this K-pop idol right. So, you make a choice about which community you start to belong to in terms of the words you use. And then we asked, like, if some people were to change their words, would that really substantially change the collective groupings of, of other people? And you can ask a similar question about stocks in that case. It's a little bit less about what you can change, but more of a sort of identification principle, right?

 

Eddie Lee 47:14:

Because it's very hard to change. For example, the, what was it, the SNP spider indices, you don't, you don't just go in and change the data. Those are an aggregation of many different things happening. So, it's more, just a sort of a way of understanding how they move with respect to one another. So, we took this sort of generalized idea of a median and try to understand, for example, and in congressional voting or Supreme court voting, could we identify these people that you would want to tug or push?

 

Eddie Lee 47:46:

And we actually asked quite a bit, sort of a general question. We didn't ask if there's one person, what if you could control anybody? Right. So, I could pay everybody a bribe, but to go in different directions. And the goal would be to change the outcome as much as possible. And what we found was that in some cases we found certain individuals that seem to matter a lot. And some cases we found that it actually wasn't one individual actually you'd have to sort of control the entire system to change things.

 

Eddie Lee 48:17:

And that's sort of interesting because you know, it sorts of hints at this idea of manipulation, right? So, a system where you have to push a lot of different things at once, you have to push all the buttons in different ways, all at the same time, that's hard, that's complex. Whereas if you just paid one guy and you get all the votes your way. Well, great. I know exactly what to do. So, we sort of look for that signal. And I, I guess maybe I'll point out one of the examples that we looked at, which was the Supreme court, and this is not the modern court, this is the Rehnquist, the second run core score, which was from 94 to 2005 when William Rehnquist was a chief justice.

 

Eddie Lee 48:55:

And what we found was that the primary signal was not about O'Connor and Kennedy, who are presumably the median voters that matter the most and that's, that's conventional wisdom. And that was really interesting. Right? So, suggest that there's something more complicated going on. I think given some of the work I've done, that's not so surprising, but as it turns out that it's a little bit more subtle, right? It's not that power is totally diffused throughout the court.

 

Eddie Lee 49:26:

If you ask whether courts, whether a vote was liberal conservative, and there's some ambiguity here in how you determine whether something is liberal conservative, but sort of pushing those aside for the moment. If you just ask whether the decision was conservative or liberal, and you asked who was the sort of most influential, according to this measure, then you do find O'Connor and Kennedy. So somehow the fact that you can identify them as being important voters hinges on your interpretation of their votes as partisan.

 

Eddie Lee 49:58:

If you don't, if you don't, if you forget about the partisan, you just say they voted in these ways, then they don't, they don't seem to be important. So somehow, right, the statistics are telling us something potentially very interesting, which is that influence is not only a measure of the changes that someone can impose, but your interpretation of those changes as well.

 

Michael Garfield 50:23:

Gosh, you know, I'm just thinking about this in light of research that was done right before the 2016 us presidential election on search result ranking. I forget who it was that published this, but they were talking about the difference in a news item, making it to the top search result versus the second search result and how they were able to swing by modulating that across the entire population of people actually issuing these searches.

 

Michael Garfield 50:53:

They were able to swing electoral results by up to 25% in one direction or the other. And they actually tested this on some mayoral elections in India, where they were able to demonstrate that they were, they were able to throw, and they didn't actually do it. Oh, I was okay with it, it was a retrodiction of results that had happened, like a matter of weeks before. Yeah. But they were, I mean, they basically said, look, you know, the search companies need to be aware that the structure of accountability within these organizations, when you have an opaque algorithm for search results and the company, they can throw somebody under the bus who was working under nondisclosure on the algorithm and say, Oh, this was a rogue agent.

 

Michael Garfield 51:40:

We lack effective safeguards in society to prevent this kind of thing from happening because the kind of manipulation at scale that you're talking about does seem possible.

 

Eddie Lee 51:51:

Right. So, what you're saying is that it's not just the facts that you give people, but the salience of those facts. Yeah. And that's, that's actually very subtle because, you know, I certainly actually, I, you know, it was just last, was it this past week I heard a talk about this, but I just can't for the gut of me remember, but something about salient being really important for determining how people interpret events around them.

 

Michael Garfield 52:18:

So, there's another dimension of this paper, in your discussion on this paper, you're talking about this, not only in terms of voting results, but in the susceptibility of population to disease or disinformation. And so, you know, David and Jeffrey and many other people at SFI have written about, and we, you know, we had Laura Hibbard, the friend and the scarp, you know, on the show also recently talking about cultural contagions. Yeah. And so, I'm trying to wrap my head around how this model could inform a strategy for fighting disinformation in this way. And like, what would that look like?

 

Eddie Lee 53:02:

Well, you know, it's, it's kind of hard to make that connection. I think formally because we're looking at the statistics. So, we look over many aggregate dynamics and so on. And I think probably for something like this information who really do want to understand the dynamics of how things are spreading. So, there is that sort of shortcoming, but you could still ask, I think in that context, you know, how changes in the system could result or what changes in the system could result in the largest changes in the outcomes.

 

Eddie Lee 53:35:

And I think sort of to do that, you sort of have to rely on the fact we're focusing on this mathematical technique for, for studying the response of a system to perturbation. So, we're asking, we're sort of thinking about that in a statistical sense. It doesn't have to be in a statistical sense, but actually it sort of connects to a really interesting set of ideas of sort of tangential to this. And what it should say is this, this field right, is relying on this idea of information geometry.

 

Eddie Lee 54:07:

And what information geometry is about is imagining that if you have some sort of mathematical description of a system, it has to be parameterized. You typically choose parameters in a way that makes sense relative to what you're studying, but you could imagine changing those parameters. And you could imagine an alternate universe where a different sort of dynamics range. And so, this idea is that these models are all connected to one another. If you change the parameters a little bit at a time, eventually you get to those crazy alternative universes, which is totally different.

 

Eddie Lee 54:42:

But the path that you take is described by information geometry. And what we say is the curvature of, of that geometry. So how quickly it changes tells us how sensitive the system is, right? So, if I change this parameter a tiny amount and all of a sudden, I am in this alternative universe, then it's highly sensitive. It's highly sensitive to perturbation, right? Whereas if it's totally insensitive, two changes, then this primary almost doesn't matter, right?

 

Eddie Lee 55:14:

You basically get the same universe. It's slightly different. Maybe the colors of the cars are not quite the same, but it's almost exactly the same and what it turned out to be the case in a lot of physical models of the world is that, and you have many parameters that actually don't matter. And you have a few parameters that really do. So, for example, in the biology of a cell, you have these sophisticated bottles with tons of parameters that you can never hope to measure.

 

Eddie Lee 55:44:

And in fact, you can't really measure them anyway. Okay. But if you plug those values into the mathematical model, you basically get a cell that works. So, this is incredible because you're sometimes off by a hundred percent, 200%, but it's the works. Why it's because the information geometry is actually multi-dimensional and in some directions of parameter space, it's totally flat. So, all of the universes look exactly the same along that dimension. And it's very, very sharp along some other dimensions. So, you have to get some things right. And those things matter a lot, but most things don't matter.

 

Eddie Lee 56:15:

And so, this idea that you have this hierarchy of sensitivity, right, is what allows us to understand that the world, because if everything mattered, everything's out the window, you can never do that. You can never understand nature, but the fact that there's a hierarchy means that I can first build a model that sort of right, right. Newton's model or Copernicus's model, it's sort of right. It's not exactly right, but it's good enough that people believe me. And then, you know, Newton comes by and then Einstein comes by. But all of those additional effects features are smaller and smaller.

 

Eddie Lee 56:45:

And so that's what I mean by hierarchy of parameters. Right? And so that's what allows us to, well, that's a claim, but you know, that's, that's what makes the universe easy to learn about. And hopefully that is true. Generally. It's not, it's not clear that's true for every complex system, right. Which is why the complex systems are hard, difficult, but what's sort of interesting in the context of pivotal components is if there is such a hierarchy, then it means that this system is easier to control, right? Because you don't need every degree of freedom.

 

Eddie Lee 57:16:

So perhaps in order to facilitate diffusion of power, you want to design systems that are completely impossible to control unless you have all the fingers on the right triggers at the right time. Whereas if you want it to be controllable, then you need to design it with this idea in mind. And you can imagine situations where you want A, and you don't want B or you want controllability and you don't want non-controllability. And that's sort of the connection. There's a long-winded way of getting to this connection with, you know, susceptibility in say, disinformation is one might want to ask, how are these systems designed and are they designed in a way that makes them easily manipulated or makes them hard to manipulate?

 

Eddie Lee 58:03:

And do you want one or the other? Because you know, it could be that disinformation is really hard to root out when things are very decentralized, right? So, it may actually be a clash of values, right? You may actually want one thing, but it just leads to the wrong thing.

 

Michael Garfield 58:19:

I'm reminded of a, was it around the 2004 presidential election? Someone had a video where they taught a chimpanzee to manipulate one of the Diebold voting machines. You know, you can like, you can teach a Chimp to hack this, which brings us back all the way back around to the primate stuff. Yeah. You know, as somebody who has a lifelong abiding interest in both the sort of philosophy around evolutionary theory, as well as time travel fiction, you know, these, what you just described are the two sort of worldviews that you see at war in the debate over evolutionary contingency versus inevitability, you know, it's, it's ultimately, it's a dispute between two different models.

 

Michael Garfield 58:56:

Like, do you have Ashton Kutcher's butterfly effect, time travel where every, you know, like you keep changing everything on accident or is it more like back to the future where it only matters who sleeps together? You know, that's, that's, you know, somehow that, you know, the timeline is like completely robust against these perturbations. And so, you know, that's, I guess until we have this figured out, my advice is to like, not get in the DeLorean. Right. I don't know. Well, dude, it's, it's been absolutely wonderful talking with you before we wrapped this.

 

Michael Garfield 59:31:

I just an opportunity to tell people a little bit more about what's on your plate right now as a researcher and what you're looking forward to, what are the burning questions for you right now? And, on the horizon, in the months of confinement to come here,

 

Eddie Lee 59:52:

Well, I'm, I'm writing the second Principia Mathematica, you know, because people keep saying, you know, this is the time when you really get some work done. And I'm sort of past that now. It's been so many months?

 

Michael Garfield:

And it's horrible.

 

Eddie Lee 1:00:12:

If you keep telling people you're working on this grand project, nothing comes out while you're still working on it. No, I have, I have several things on my plate right now that are sort of on and off. One thing is sort of extending this armed conflict stuff to start really digging into some of the, sort of maybe contingent factors, but trying to find patterns in the contingencies that unify them. We're also been looking at Jeffrey West Chris campus. And I have been looking at metabolic scaling theory or in the forest. So, trying to understand some of the dynamics that lead to scaling and forest populations and hopefully actually touch on armed conflict in some way.

 

Eddie Lee 1:00:52:

I've also been thinking about acquisition of information in firms. So, trying to see why is it that firm lifetimes are distributed in a very sort of regular way, surprisingly across sectors. And does it have to do with how they're learning, how they're acquiring information from around them. I'm finishing a project with Jess and David on adaptation learning and adaptation. And it actually, the idea is that we're thinking about different forms of memory embedded in either the organism in its behavior or in its environment, and trying to unify these various mechanisms or implementations of memory into a sort of optimal adaptation framework.

 

Eddie Lee 1:01:36:

I'm sure I'm forgetting something, but it's a lot and, um.

 

Michael Garfield 1:01:39:

You don't have to keep fighting it. Right. If you've forgotten it, if you've heard that ??.

 

Eddie Lee 1:01:47:

There's no glue, you know, at some point it will lead to some cascading failure.

 

Michael Garfield 1:01:52:

Fair enough. Well, yeah, thanks again for being on the show and folks check the show notes for a link to all of your research and the other stuff that we've discussed in here.

 

Eddie Lee 1:02:06:

Awesome. It was fun.

 

[Outro]

Michael Garfield 1:02:08:

Thank you for listening. Complexity is produced by the Santa Fe Institute. A non-profit hub of complex systems science located in the high desert of New Mexico. For more information including transcripts, research links and educational sources or to support our science and communication efforts, visit SantaFe.edu/podcaComplexity Podcast

Fractal Conflicts & Swing Voters with Eddie Lee

 

[Intro] 00:00 :

You know if you have a forest with a lot of dry wood that hasn’t been cleared away, by previous small conflagration you can have a major conflagration, right? I am not saying that, that is conflict. But you could imagine some kind of rule that a regulator would perform in society. I don’t know what that rule is, but definitely one of the ways it manifests is by preventing the largest outbreaks from happening. Or, an attempt to.

And, you know, one of the ideas behind some of the models that have been proposed for conflict is actually similar to this forest fire thing. Which is that, the conditions for conflict will simmer, and when it reaches that boiling, you can have one. And when you have one, it’s this big thing but it depletes the system in a way.

And so, you are always poised between never being able to have these big conflicts and having these sorts of having like these smaller conflicts. Would it be good to prevent all conflicts or would it be bad? Would that sort of lead to some sort of special situation where you suddenly had the nuclear Armageddon? I don’t know.

 

[Intro music] 01:00 :

 

Michael Garfield 01:24:

Since the 1940s scientists have puzzled over a curious finding. Armed conflict data reveals that human battles obey a power law distribution like avalanches and epidemics. Just like the fractal surfaces of mountains and cauliflowers, the shape of violence looks the same at any level of magnification.

Beyond the particulars of why we fight, this pattern suggests a deep hidden order in the physical laws governing society. And digging in the new analysis of data from both armed conflicts and voting patterns, complex systems researchers have started to identify the so-called pivotal components involved. The straw that breaks the camel’s back. The spark that sets the forest fire. The influential but not always famous figures that shape history.

Michael Garfield 02:11

Can science find a universal theory that predicts the size of conflict from their initial conditions? Or identity key players whose knob society in one direction or another?

 

Welcome to Complexity. The official podcast of the Santé Fe Institute.

I am your host Michael Garfield. And each week we’ll bring you with us on far ranging on rigorous researchers developing new frameworks explaining the deepest mysteries of the universe.

 

Michael Garfield 02:39:

This week’s guest is SFI program postdoctoral Eddie Lee. Whose work in conflict avalanches and swing voters gives a glimpse of the mysterious forces that determine why we fight. And how we will be able to prevent the next conflagration.

 

In this episode we talk about armed conflict as a fractal and a form of computation. Swing voters in the justice and influencers in pop cultures. And what these studies may have to say about the deep constraints that guide the currents of society.

 

Michael Garfield  03:11:

Just a note that this will be our last episode before short summer break to give our scientists uninterrupted time to work on a torrent of new research. We have some exciting episodes scheduled for our return in mid-august. In the meantime, please be sure to subscribe to complexity podcast on your favorite podcast provider to make sure that you stay in the know. And if you value our research and communications please consider a donation at SantaFe.Edu/podcastgive or join our Applied Complexity Network santafe.edu/action

 

Michael Garfield  03:43:

Lastly, we are excited to announce that submissions are open for this fall's inaugural complexity interactive. A three-week online project based immersive course where you get a rare opportunity for mentorship by a large faculty of SFI faculty, including Simon DeDeo, Daniel Bassett, Melanie Moses, Ricard Solé, and many more.

For more info and to apply, please visit SantaFe.edu/sfi-ci

 

Michael Garfield 04:23:

All right. Well, shall we

 

Eddie Lee 04:25:

All right, let's do it.

 

Michael Garfield 04:26:

Yeah. Eddie Lee, it is a pleasure to have you here and our first ever in person, socially distanced podcast recording here across my patio. So welcome.

 

Eddie Lee 04:39 :

Thanks for having me. This is great.

 

Michael Garfield 04:42 :

I'd love to start as we typically do by just providing a little bit of personal background and having you talk a little bit about how you got into science and how you got into specifically an interest in the kind of research that you're doing at SFI and the stuff that we'll be talking about today.

 

Eddie Lee 05:00 :

Well, I think it was last week that actually the writer of the magic school bus series passed away. And, you know, I have to say that she really did leave a fantastic legacy and I'm part of that legacy. I grew up on, on the magic school bus and actually some after school science programs as well. So, I sort of had this nascent interest in science for a long time, something I enjoyed doing. I didn't really see myself becoming a scientist.

 

Eddie Lee 05:32:

And in fact, when I went to college, I actually tried to become an economics major surprisingly enough, but it turned out that I took this sort of fantastic course called integrated sciences. And it's sort of my preview in a way to complex sciences, because what it was a class telling us, showing us that the way of thinking in computer science, biology, chemistry, physics, we're all coming together in the context of some fascinating problems in biology.

 

Eddie Lee 06:03:

So, this idea that all these things were coming together was a really surprising thing for me. And I think what really sort of hit it home was when I read this book by Phillip ball called critical mass. And so, he brought up the aspect of society, also being part of this. And, you know, at some point in my past, I'd also read Isaac Asimov's foundation series, which is, which is the physics of social behavior and prediction. And so somehow all of these things were sort of swirling around in my head.

 

Eddie Lee 06:35:

I didn't know what to do about them. And I ended up imagining what I really need to do is go into sociology. So, I went to, I searched around, I'd try to talk with sociology professors. And I just talked with one, I think he was maybe at the head of the Taryn at the time. And he said, you know, I know this guy who might be able to direct you. And he couldn't remember the name, but he started searching online and he showed me the web page. And it was one of the professors who had taught that first class, that first course that I mentioned integrated sciences.

 

Eddie Lee 07:05:

So, he sorts of just sent me right back. Yeah. Boomerang style back into that professor's office. And he actually was familiar with SFI. So, he's the one who directed me to talk with Jessica and Dave Cracker, who now I've known for a long time. And that's sort of where all of this stuff started happening. And that's when I started really learning about what sort of science SFI works on and my sort of vague interest in it became, you know, crystallized.

 

Michael Garfield 07:37:

Awesome. Well, this is a great place actually, to just dig right in to the first set of papers I wanted to talk with you about today, which you co-authored with David and Jessica, as well as Brian Daniels and Chris Myers on scaling theory for armed conflict avalanches. So this idea that we can use physics insights to understand human social behavior from like an orbital perspective and like really much like in the foundation books, you know, that we can, we can come at the, the sort of vagaries of history from, you know, David would call a rigorous and principled quantitative approach.

 

Michael Garfield 08:23:

The history is not just the story of great men making decisions, but that those decisions occur within a landscape of physical constraints. And what are those constraints and how do they manifest themselves? And this is really interesting work and it's got a long history that you're building on. And so maybe the history of this strain of research is the right place to start. And then we can unpack it from there.

 

Eddie Lee 08:48:

Sure. Yeah. Like you're saying, it's interesting, but what we weren't focusing on was the particular stories of particular conflicts. And there's a lot to study about any particular conflict they're quite different, right? So, you might think of say the Libyan or Tunisian revolutions happening today, which are quite different from world war two, which were quite different from the British and French battling it out, say in 1812. So, these are all very distinct lines of history in and of their own.

 

Eddie Lee 09:20:

And so, what we were interested in was trying to understand whether or not there were universalities common features shared amongst these things that are clearly disparate. And as you were saying, this isn't a completely new idea. It was in 1948 that Louis Fry Richardson showed this amazing feature, which was that when you make a histogram of interstate Wars, so you count how many wars there are with F people that died, F fatalities. And you look at this histogram on a log plot, what you find is a straight line.

 

Eddie Lee 09:54:

And what's remarkable about a straight line being on a log plot is that it's, that's a power law. That means its scale free. In other words, sort of most, very naively speaking. If you were to look at a small conflict, it's sort of like a shrunk down version, statistically, speaking of a large conflict and so on and so forth all the way down. And so, this was very strange because interstate wars that you looked at were all quite different from one another. So why would you get this sort of regularity that emerges when you look across many different conflicts?

 

Eddie Lee 10:27:

And since then other people have noticed other interesting patterns in the timing of conflicts, as well, as well as terrorism and so on. And what we were interested in was looking for some of these patterns in a more recent dataset, which is called the armed conflict location event, data project, or ACLED for short. And what was remarkable about this data set was that they didn't take all these events, these small disparate individual conflict events that would occur in a war. They didn't group them together into wars.

 

Eddie Lee 11:00:

They just sort of disaggregated them into, into small localized defense. And so instead of having to take a quantity, that's already defined for you as a battle or a war or skirmish, or just a local riot, we got to connect these into ourselves, into these clusters of what we called conflict avalanches. So many ways similar to two battles or Wars, but without having assumed some of these connections from using sociopolitical factors. So just looking at the time and spatial patterns of these events, and what we found was that when you look at these conflict avalanches, various measures of their size, so not just how many people died, but how many of these reports did you obtain?

 

Eddie Lee 11:43:

How long did it last, how far did it spread? All of these different factors show these characteristic power law tales, like what Richardson saw and what we were able to do beyond that is to show that these pieces were all connected to one another in some mathematical way. And I think I had an analogy for this, which was that if you imagine you walk into a dark room and you're trying to figure out what is this object in front of you?

 

Eddie Lee 12:12:

And you find first, there's this sort of rigid pillar. Then you find this long sort of thing with a hairy end. You find this floppy disc, like thing, you start in your mind developing this picture of an elephant. And the way that that works is all these pieces are connected to one another, right? They're pieces of one whole. They're not independent, separate things. And so, you build a model in your head and you say, this must be connected. The legs must be connected to the body, must be connected to the head and so on.

 

Eddie Lee 12:43:

So, we did something like that with the mathematics of these various features. And we found that they were self-consistent in a way that seems to point to one sort of object driving the dynamics.

 

 

Eddie Lee 12:55:

And that's sort of really cool because it suggests that there is some sort of underlying thing that we're just getting hints of. We haven't found it yet. We haven't discovered exactly what it is. Obviously, it's something about war and conflict, but it's perhaps something bigger, right? That leads to universal patterns, across many different types of conflict.

 

Michael Garfield 13:17:

So, you draw some analogies in one of the preprints that you've done on this work, emergent regularities and scaling and armed conflict data. You draw analogies to forest fire models, which other researchers have tried to apply to this, to neural avalanches. And, you know, just from last week's episode with the SFI counter speech team and the network graphs that you see of the Twitter data that they've harvested, and the way that you see these conversations, these sort of battles between hate speech, organized, hate speech groups and organized counter speech groups on Twitter.

 

Michael Garfield 13:56:

And this look also like these, you know, you see these the sort of conflagration of point by point debate breaking out when there's a successful, an actual collision, rather than just the hate group eliciting engagement to its original post. But people are actually willing to go into the trenches with one another on social media. And so, I would love to hear you talk a little bit more specifically about this model as a, like a branching network diagram.

 

Michael Garfield 14:26:

Yeah. There's some really interesting figures, and we'll link to this stuff in the show notes so that people can, can actually get a look at what we're talking about here. Because it actually looks rather menacing.

 

Eddie Lee 14:39:

That's why we chose those colors.

 

Michael Garfield 14:41:

The map of a branching armed conflict here.

 

Eddie Lee 14:44:

Yeah. So, cascades are a real central conceptual foundation for understanding a lot of these different kinds of phenomena. And as you mentioned, if you look at neural activity in the cortex, what you find are, are these cascades of activity, and it's not just a bunch of neurons firing, you can actually trace it out to some sort of branching process. So, you imagine there's the center, the first neuron that activates other neurons, which activate other neurons and so on sort of in a conflagration type picture.

 

Eddie Lee 15:24:

And you can also think about various other phenomena with this sort of conceptual framework. So, forest fires are another way, right? So, you imagine you have a bunch of sorts of local clusters of maybe dried out patches of trees and you have a spark, maybe some lightning strikes, and then it sort of burns its way through all these connected clusters. And you can think about that analogy applying to, to social contagion as well. And so that's why we sort of started with this idea of thinking about conflict in the same way.

 

Eddie Lee 15:55:

And mathematically, it's a very generic way of thinking about it, right? All of that you need are different pieces that are connected to one another in this sort of branching way. And so, you can get this exponential increase. Actually, this is very relevant in the context of today with COVID right? So, when people talk about exponentially growing processes, this is exactly what they mean. Or you have someone who spreads it to some other people. And if each person on average spreads up to more than one person, then you have an explosion. So, we sort of were playing with this framework and thinking about conflict in this way as a social contagion.

 

Eddie Lee 16:29:

And we're not sure if that's exactly how it spreads, but we can at least say that the way that it spreads is consistent with this idea of the way that it spreads both across space, as well as across time and in size.

 

Michael Garfield 16:46:

So, this elicits for me, links back to this sort of giant network model and building of this, these conversations on the show with all of the other conversations that hopefully we'll publish in the SFI press one day and make a poster. This reminds me a lot of, of two things. One is the conversations I just had with Jeffrey West a few episodes ago, where he was talking about biophysical scaling laws as being extremely coarse and like not actually capturing all of the variation that we would see in like specific evolutionary instances. The Primo example, being that human beings, modern humans use something like 30 times more energetic resources than we would expect from mammals of this size.

 

Michael Garfield 17:33:

Because as we've talked about with David and some of the transmission episodes, the human being is basically just the visible corner of this hyper object now. And then each of us is actually like a cyborg spanning the globe with all of these electronic augmentations. Right? So, something like that is going on in this work with conflict. And when you actually peer into the data with a little bit more granularity, you see a lot of variation in this, not enough to challenge the thesis, but enough to suggest that there are some interesting details about the regional and temporal variations of these conflicts.

 

Michael Garfield 18:14:

And I would love to hear you talk about that. And then also just to like make this kind of unwieldy let's pin onto that since you brought up COVID, Chris Morris piece for the transmission series, where you talking about super spreader and that's and how the, you know, our zero as a measure of the transmissibility of a disease, doesn't actually tell you what's going on in that particular church or in that particular campus building or, you know, so there's. So, what are the features about the data set that you've noticed that seemed to be providing insights into why there's variations in the size of these conflicts around that mean.

 

Eddie Lee 18:56:

No, that's, that's great. I mean, I think it's really important to point out that a lot of this work 's goal isn't to explain all this variation, which, you know, eventually we want to do, but it's really to talk about the shared common features across all of these phenomena. And then of course it's important to mention that, that's the only thing it explains. And so, the features that I'm talking about in terms of scaling are the averages, but you can have scaling in the variation as well.

 

Eddie Lee 19:29:

And that's, that's really interesting. And as you mentioned, right, you know, Chris Moore wrote that very nice piece where it's really important to mention that are not, is an average and not necessarily going to tell you what's happening in a very particular location. And that's also exactly true for theories about metabolic scaling in force or conflict. And in fact, it turns out that that sort of variation manifests directly in our work looking at conflict.

 

Eddie Lee 20:02:

And one of the things that we find is that it's not only essential to account for the similarities between how conflict grows in different locations, but there's also variation in terms of this sort of coefficient in front of it. So, the way that it grows the shape of the curve may be similar, but the offset of that curve may be different. And that might depend on factors such as prosperity or governance. So, here's one example, right? So, you look at Eastern Somalia, weak governance, a lot of widespread poverty and a huge amount of conflict.

 

Eddie Lee 20:36:

One of the hot zones in this and the status set. But if you look at South Africa, it's quite prosperous, they have a pretty strong government and its relatively low levels of conflict. So even if it's the case, that conflict evolves in a similar way, you just don't expect to see as much conflict and South Africa, as you see in Eastern Somalia. And so, as you also said, these things will vary in time. So hopefully Eastern Somalia is not a hotspot forever, but maybe somewhere else will become one. So, there may be patterns also in the diversity of hot zones.

 

Eddie Lee 21:09:

And we see a hint of this. We can't say that will be the case forever because we only look at 20 years. But one of the sort of fascinating things is that Richardson, who I mentioned earlier, looked at wars between, I think, early 1800s and mid-1900s. And he found basically the same statistics. Aaron?? looked at the statistics all the way from Richardson's asset up to today and found the same statistics. So, there's some sort of weird preservation of conflict despite changing technology, different countries. And, and so on that either, I don't know, is, should you be amazed or should you be afraid?

 

 

Michael Garfield 21:47:

Well, I mean, again, two things come up here, right? One is Jen Dunn's work on food webs and how the structure of these food webs that their patterns of connectivity have been amazingly well-preserved conserved in there, in their form for like the last 500 million years as we've gone through all of these regimes in what is actually living in the oceans. So that's part of it, you know, and that's, that's sort of begs the question that I asked Brian Arthur, which is, do you think that we can use these models as a way of predicting like basins for future innovation, you know, and like predicting areas where we can expect potentially larger conflicts than we're seeing.

 

Michael Garfield 22:30:

And then the other piece of it is, bringing in the geopolitical component, Peter Turchin's work. When we're talking about a quantitative study of history and his writing on what he called the double helix of inequality and social instability, where he saw that they were very strongly negatively correlated over the last a hundred or 150 years. But, you know, you make a point in this paper that there are other geopolitical and geographic features that have to do with the coastlines and national borders and locations of these conflicts relative to large urban centers' populations.

 

Michael Garfield 23:11:

Yeah. So, I, again, I know that this is sort of beyond the scope of the paper, but I'm really curious, you know, what other, what other factors you think are playing into this and how that might be guiding your follow-up research in this area?

 

Eddie Lee 23:23:

I think what you can say is despite the fact that there are a lot of these universal patterns that we think we find is that there's a lot of work trying to understand the mechanisms behind why these patterns might appear. And I think there is yet a connection to be made there. And that would be sort of, I think the Holy grail for, for this sort of work is really understanding those connections and flushing them out.

 

Eddie Lee 23:53:

And somehow, they lie in the intersection of maybe social mobility, social prosperity, economic prosperity, technology, geography, right. You can't really fight someone if there's no one there,

 

Michael Garfield 24:08:

Tell that to our Facebook group.

 

Eddie Lee 24:13:

Yeah. I guess you could make up stories and so on, but there are a lot of different factors that somehow mashed together to generate these patterns. But as you were mentioning, Jen Dunn's work. That's not necessarily to say that you can't get universal patterns from that. So, you know, this is maybe too simple of an example, but one sort of that comes to mind right now, which is the central limit theorem, right? There are many reasons for why you get random likes statistics, right?

 

Eddie Lee 24:46:

Many different ways of generating them, but by virtue of their randomness, you end up getting some sort of regularity. So, we'll see. Maybe it's the fact that there are so many factors that do influence conflict that ended up getting simplification at higher scales, or maybe not, maybe, maybe it is contingent. And we, we just don't have enough granularity in our understanding to access that contingency

 

Michael Garfield 25:15:

Later on, in this paper, you talked about there being, you know, getting into this granularity, you say: "Unlike canonical cascade models, conflict also includes lattice style dynamics that evolve with geographic spread. The suppression of these dynamics away from the core could reflect social processes or geography that impact conflict evolution. Furthermore, our model suggests conflict is not only the result of local correlations and activity, but also regional and temporal disorder, perhaps reflecting memory of the severity of initiating events".

 

Michael Garfield 25:47:

So, this links directly to another paper that you wrote with Brian, David and Jessica on primate conflicts and temporal scaling collapse. And the idea, you know, Jessica's work on primate conflicts as a form of collective computation and the severity of those conflicts having to do with various factors that are performing computations at the level of like an entire primate troupe, right? So, this notion that the patterns that you're seeing in conflict data, that they hinge on there being certain evolutionary reasons for the distribution of conflict, duration, and severity is really interesting. I'm curious to hear you talk a little bit about this other paper and how it links in.

 

Eddie Lee 26:36:

Right. No, that's, that's interesting. Right? What are the functional properties of conflict, right? Is there a reason why conflict plays out the way that it does? It's hard to say for human conflicts. Obviously, there are reasons why people give for starting wars. But in monkey conflict, there's potentially one reason which is, has to do with a social hierarchy, right? So, you need to have a little bit of disorder and conflict in order to establish people's roles and where they sit.

 

Eddie Lee 27:07:

And the severity of the conflict at the same time cannot be too, too big, right. If it's too severe, it's actually detrimental to society. So, there's this idea that maybe what conflicts does is it serves an information gathering acquisition role, as long as it's not too bad,

 

Michael Garfield 27:26:

People have to survive, remember.

 

Eddie Lee 27:28:

Exactly. If you don't, if you know, if it's nuclear Armageddon again and that's it, we should remember that. But, I think one of the things that we were trying to study in that work was characterizing the features of monkey conflict and perhaps extracting from those features a potential functional role that, that these things, or these observations could, could conserve.

 

Eddie Lee 27:53:

And actually, what's really interesting about the monkey conflict is we studied a group, a society of ??. What we found was that the duration that conflicts lasted well, beta distribution, it was not a parallel distribution like human conflict, but how long they lasted that distribution of duration seemed to be a distribution that looked the same, whether you looked at small conflicts or large complex. And it's very strange because conflicts with 2 monkeys are not the same thing as conflicts with 10 monkeys.

 

Eddie Lee 28:24:

There are many more ways they can interact. There are many other things that could happen, go wrong. And so, this was very curious. And so what we found was that if you could think of these conflicts as being sort of the time for all the pairs to resolve their differences, and each of those pairs, or took some amount of time to resolve their differences, then you could get something similar to what we found in the data with the condition that the pairs that resolve their differences later remembered the sort of intensity with which the pairs resolved their differences earlier.

 

Eddie Lee 29:03:

So, in other words, if the fight started off with some intense pairwise interactions, like, you know, you bit me, how could you bite me? And that's, that's actually very apparently a very aggressive interaction that the other monkeys will not tolerate.

 

Michael Garfield 29:18:

I remember preschool. Yeah.

 

Eddie Lee 29:19:

And so, you end up having these long correlations from the beginning to the end, and if that's true, then you end up getting these, these sort of long tails of conflict duration. So, you have many conflicts that tend to last longer. And for us, when we, that was an indication of what we call collective memory, right. This idea that the entire duration of the conflict from beginning to end, some remember at the beginning. And that's what we mean by we say in that paper.

 

Michael Garfield 30:00:

Well, okay. So, you know, this, I have the luxury here of kind of like going out on the plank and speculating. Yeah. But this seems to be possibly why certain human conflicts are just insane, like the Hatfield’s and McCoy’s, you know, that the feud has a sort of durability built into it because of the way that the history of a family is prioritized. And like the, you know, the bigger the conflict gets, it would seem like certainly there are factors that are keeping resolution that are like forestalling resolution, but on the other hand, the more abstract the conflict becomes for people, the easier it is to get over.

 

Michael Garfield 30:45:

Right. So, you know, this, this is sort of like, you know, I wonder what this approach to understanding this might have to say to, or link with other work on the way that conflict between human beings has changed as war has become industrialized and has been more and more about foreign conflicts rather than defending your own, your Athens or whatever. From there are certain things that would seem to be kind of holding us in the United States in a position of kind of unending foreign Wars.

 

Eddie Lee 31:20:

Right. Well, I think it's a little bit unfair to say that these Wars are external. Right. For many people, these Wars are personal, right? Yeah. And that's their life, but sort of touching on this, on this question of, you know, how is it that conflict persists or why is it that conflict persists? I think one thing we suspect from looking at these patterns is that these memories, these core correlations in a conflict, are expressed not only in the history that people maintain about themselves, but also in the geography of how it spreads.

 

Eddie Lee31:57:

So, it's quite possible that what encodes the history of the conflict is not just what we think right, the stories that we tell ourselves, but potentially also factors of the environment. Structures that we build. And some of these structures are really obvious, right. I mean, we built the nuclear arsenal, right. It's going to be there. Right. This is really interesting to me, it sort of connects to some of the work that David, Jess and I have been doing recently with adaptation and outsourcing a memory into the environment.

 

Eddie Lee 32:30:

Right. Which is this idea that there are biological organisms that intentionally or use and manipulate the environment to couple their bad memory or the behavior with longer timescales that they need in order to better adapt. So, ants do this by building these trails. It's unlikely that individual ants can really remember what they're doing, but over the collective activity of many, they can establish these very long-lived persistent trails. And they can use this trail to harness resources.

 

Michael Garfield: like Google calendar.

 

Eddie Lee 33:01:

Exactly. No, I mean, let me do exactly the same thing, right. I mean, David has that example of the notepad, right? Multiply three times, six times, five times, 2,374. I can't do it in my head, but give me a piece of paper. I can just, basically you short term memory expressed on paper to do that calculation very efficiently. And in addition to opening up new ways of doing that calculation, that aren't accessible in my head to myself. So, there is this question, right? Sort of going back to the functional properties of armed conflict, you know, what purpose does armed conflict solve?

 

Eddie Lee 33:32:

Are there ways that we are sort of driving our conflict ourselves by embedding into our environment that we don't know of, that we do know of. And I don't think those questions are necessarily answered. At least I haven't seen those answers, but would be really interesting to think about, I mean, right, this is just one other phenomenon in nature, right? It's not just us that fight. We do fight with certain technologies and armed human conflicts or especially cause a lot of fatalities compared to other organisms. So, we're quite brutal, but.

 

Michael Garfield 34:02:

You know, to, to bring up a marvelous work of science fiction, that seems to have some direct bearing on the insights that you just said. I've been in a book club recently discussing the science fiction trilogy Lilith's Brood by Octavia Butler. Should I read this? It's an extraordinary hit. It's a very, very relevant and kind of evergreen piece of work by the first science fiction author to ever become a MacArthur fellow. It just highly awarded a black female science fiction author, you know, just an amazing, amazing mind.

 

Michael Garfield 34:36:

And I've been on the tip of trying to read more black Sci-fi right now to get a better understanding of the world spaces disclosed thereby. And in this book, this book is about humans interacting kind of non-consensually with a race of aliens that comes to earth and finds us in the aftermath of a nuclear apocalypse. And once they're masters of bioengineering and they want to reboot our planet and nurse humankind and the biosphere back from extinction.

 

Michael Garfield 35:09:

But in order to do that, they kind of have to change the rules for human beings. And one of the things that they do is they refuse to give the humans that are going to repopulate earth, any writing materials, like it's part of their thing, that they all have genetically engineered eidetic memory. And so, they would rather engineer us to have perfect recall, then allow us to record history on paper.

 

Eddie Lee 35:37:

It sounds like a curse. You know, you know, that, I guess that was the thing for a long period of human history, right? Oral tradition, Homer, obviously, until it was written down. Yeah. What is it that keeps us fighting? Is it, is it literature? And that would be even, there'd be very tragic and somewhat ironic actually, because we're supposed to learn from history.

 

Michael Garfield 36:05:

But you know, in this paper, I think it's, like you said, it's the end of the conflict remembers the beginning. Yeah. There's another piece here, which is about the beginning. And I think that this might give us a hinge to get into some discussion about your other work on swing voting, which is really, you know, also very timely and interesting. You and your co authors on this piece, talk about policers or the police macaques. And they're being like thresholds at which they will or will not intervene in order to break up a conflict before it becomes a conflagration.

 

Michael Garfield 36:44:

And I'm curious to hear you talk a little bit about how thinking about conflict as a collective computation illuminates, why we may make a choice to mediate a conflict or to de escalate it or to allow it to run its course. Right? Yeah. So, what are your thoughts on all of that?

 

Eddie Lee 37:04:

Yeah. Well, first the disclaimer is that I'm not an expert on mechanics society. So, so that, you know, when I talk with Jess later, I can say that I said this, but yeah. So, these police or macaque or high-power individuals that, I mean, you know, in short other macaque don't want to mess with them. So, if they get in a fight, you just don't hit the policer or per usual. So, they're serving some sort of regulatory mechanism and maybe in the sort of conceptual abstract sense, regulatory control is very important in terms of outbreaks and contagion, right?

 

Eddie Lee 37:40:

So, for example, for example, and I'm not saying that this is macaque conflict, but you know, if you have a fire or a forest with a lot of dry wood that hasn't been cleared away by previous small conflagrations, you can have a major conflagration, right? So, you have this idea that by preventing all, but the largest of conflagrations you've sort of saved the system, allowing it to renew itself. I'm not saying that that is a macaque conflict, but you could imagine some kind of role, right. That, that a regulator would perform in this sort of society.

 

Eddie Lee 38:13:

I don't know what that role is, but definitely one of the ways that it manifests is by preventing these, the largest, you know, outbreaks from happening or an attempt to. And, you know, that is sort of interesting. I don't know what the computational rule of conflict in human society is, but you know, one of the ideas behind some of the models that have been proposed for conflict is actually similar to this forest fire thing, which is that the conditions for conflict will simmer, right?

 

Eddie Lee 38:47:

And then when it reaches that boiling point, you can have one and when you have one, it's this big thing. But it depletes the system in a way. And so, you are always poised between never being able to have these big conflicts. I mean, truly massive ones, you know, think world war five and, and having these sort of like smaller conflicts, but I don't know, I don't know what would be, right. Would it be good to prevent all conflicts or would it be bad?

 

Eddie Lee 39:18:

Would that sort of lead to some sort of special situation where you suddenly had, you know, the nuclear Armageddon and I don't know, but it is, it is sort of an interesting question to think about,

 

Michael Garfield 39:29:

You know, pull in a totally non-scientific perspective on this from a rather sketchy character. I remember Osho Rajneesh, the disgraced guru that brought his cult into Oregon, had a line in one of his books. He said, you know, be aware of the pacifist because they're sitting on a volcano and. And you know, race had, ?? D'Souza just gave one of the flash workshop talks last week on timescales and tradeoffs, where she was talking about this in terms of exactly what you just said about, you know, if you have small forest fires every once in a while, it prevents this big one.

 

Michael Garfield 40:09:

So, you know, the question is in our efforts to expunge conflict from the planet. Are we really doing ourselves a long range disservice, are we making a tradeoff That's actually making the system dramatically more brittle than it otherwise would be? And then this starts getting into sort of more archeological or historical perspectives on the role of organized sports as a way of, of, you know, channeling our aggression into something that's less likely to result in. I mean, you might flip a cop car if you win the championship.

 

Eddie Lee 40:42:

Well, you know, I don't even know if it has anything to do with human nature, right. It might just be the way that modern society is structured. So, who knows? I mean, I think there's a real question here about what are the conditions required for armed conflict? Why does that happen? And, you know, could you get, could you develop a system where it just doesn't and totally avoid this whole issue? And I think it's actually quite complicated because conflict is associated with so many things. You know what I mean? It's not just like conflict by itself.

 

Eddie Lee 41:13:

Conflict comes with famine, comes with disease, comes with poverty, comes with deaths. I mean, you're, you're basically tearing a lot of things apart. So, it's just one side of the coin. And so maybe, you know, it's not pleasers you need, right. Maybe it's something else. Maybe you just need to change how people see one another.

 

Michael Garfield 41:34:

Yeah. At risk of being misunderstood as making a political statement here. I've been thinking a lot after talking with Jeff West about his work on, you know, the branching of the circulatory system as being a way of minimizing friction, minimizing turbulence in the bloodstream. And so, if we're going to apply that kind of biophysical scaling law insight to the way that we have established our systems of governance, then the question is, well, what does a truly fractal policing system look like?

 

Michael Garfield 42:06:

Because I used to work in festivals and, you know, there is this question of, if someone is trying to get into a fight at a festival, is it really appropriate to call the police? Can this be handled like burning man does with volunteer Rangers that are trained in de-escalation before it becomes a legal issue? And so, you know, this is, I think that your work is really inspiring for anyone who's interested in the conversation around police reform and for, you know, the kind of dynamic governance that I talked about with David in our transmission series about how many layers of modular decision making are necessary in order to get something that is both robust and isn't going to just implode periodically.

 

Eddie Lee 42:54:

Right, right. Yeah. I think, I think there were some big questions here, right. That goes far beyond the political issues at play today. Right. I mean, the questions that we're asking about complex systems are big questions about the system and might even question the sort of principles of the system. If you can get sort of a similar outcome with a completely or sorry, a different outcome with a completely different setup, that'd be really interesting. You know, I mean, especially in the context of armed conflict, because it really does seem for at least a couple of hundred years of innate features of the system.

 

Eddie Lee 43:31:

I mean what is it about the system that you can change? Can you change it? Is it human nature? Is it our construction and modern society? I think those are questions that haven't been answered and somehow, they're connected to, by the fact that conflict is associated with many of these other things, they're really connected with serious social problems that still have played to human society for a while. So, I think there is a chance here really by digging into one of these aspects, potentially being able to get an understanding of some of these broader social issues.

 

Michael Garfield 44:07:

So, this is the place for us to pivot, forgive the pun here into a discussion of the paper that you lead authored for Royal society interface on this is sensitivity of collective outcomes, identifies pivotal components. You know, we're coming up on the presidential election, the conversation around swing voters. And everybody wants to figure out who are the people to target that ended up sending the sand pile down one side of the Hill versus the other, right.

 

Michael Garfield 44:40:

So, could you unpack this particular paper and a little bit of the, you know, the sort of fundamental framing that you're doing here and, you know, get into the details on that place. Yeah.

 

Eddie Lee 44:52:

So, what we did here was we started with this idea of the median voter. So, the median voters, basically the person who sits in the middle, right, that's a swing better. So, imagine you had a bunch of people, let's imagine the simplest situation where you could just label them left and right. But then there's an odd number of people you decide by majority. And there's the person sending the middle, that's the median voter. Now that's a very simplified situation. And I think many situations are often simplified into this binary choice, which is not the case because there were so many things that actually come into play in a vote and across many votes, actually, right?

 

Eddie Lee 45:31:

Right? So, you never just have a median voter. The voter might be immediate on one issue, but not on any others and you might have variation, right? So, imagine that this bill is an obvious left, right? But you want it to pass, you know, what their median is going to do, then you change it. You make a more complex bill with many amendments and so on, right? So, you can play with this in many ways. And so, as a result, you're not really asking, or we weren't really asking what is the median voter, but once we account for all these sorts of potential complexities that manifest in the statistics, how could we identify that person who sort of plays that role more consistently than others?

 

Eddie Lee 46:11:

So, a little bit more of a subtle definition of median. And we call these pivotal voters who pivotal components. And this is a very general way of asking the problem, right? So, it's to say that if I were to push and tug on each of these people in different ways, how would the outcome change in the voting system? The outcome is the majority votes or different ways that people couldn't divide up into groups. But again, you can think about this and in different contexts, right? So, Twitter is an example. So, you imagine everyone sort of voting in a different way by choosing to use certain words.

 

Eddie Lee 46:43:

So, the vote is not just left or right, but it's whether I used it, I talked about this K-pop idol or this K-pop idol right. So, you make a choice about which community you start to belong to in terms of the words you use. And then we asked, like, if some people were to change their words, would that really substantially change the collective groupings of, of other people? And you can ask a similar question about stocks in that case. It's a little bit less about what you can change, but more of a sort of identification principle, right?

 

Eddie Lee 47:14:

Because it's very hard to change. For example, the, what was it, the SNP spider indices, you don't, you don't just go in and change the data. Those are an aggregation of many different things happening. So, it's more, just a sort of a way of understanding how they move with respect to one another. So, we took this sort of generalized idea of a median and try to understand, for example, and in congressional voting or Supreme court voting, could we identify these people that you would want to tug or push?

 

Eddie Lee 47:46:

And we actually asked quite a bit, sort of a general question. We didn't ask if there's one person, what if you could control anybody? Right. So, I could pay everybody a bribe, but to go in different directions. And the goal would be to change the outcome as much as possible. And what we found was that in some cases we found certain individuals that seem to matter a lot. And some cases we found that it actually wasn't one individual actually you'd have to sort of control the entire system to change things.

 

Eddie Lee 48:17:

And that's sort of interesting because you know, it sorts of hints at this idea of manipulation, right? So, a system where you have to push a lot of different things at once, you have to push all the buttons in different ways, all at the same time, that's hard, that's complex. Whereas if you just paid one guy and you get all the votes your way. Well, great. I know exactly what to do. So, we sort of look for that signal. And I, I guess maybe I'll point out one of the examples that we looked at, which was the Supreme court, and this is not the modern court, this is the Rehnquist, the second run core score, which was from 94 to 2005 when William Rehnquist was a chief justice.

 

Eddie Lee 48:55:

And what we found was that the primary signal was not about O'Connor and Kennedy, who are presumably the median voters that matter the most and that's, that's conventional wisdom. And that was really interesting. Right? So, suggest that there's something more complicated going on. I think given some of the work I've done, that's not so surprising, but as it turns out that it's a little bit more subtle, right? It's not that power is totally diffused throughout the court.

 

Eddie Lee 49:26:

If you ask whether courts, whether a vote was liberal conservative, and there's some ambiguity here in how you determine whether something is liberal conservative, but sort of pushing those aside for the moment. If you just ask whether the decision was conservative or liberal, and you asked who was the sort of most influential, according to this measure, then you do find O'Connor and Kennedy. So somehow the fact that you can identify them as being important voters hinges on your interpretation of their votes as partisan.

 

Eddie Lee 49:58:

If you don't, if you don't, if you forget about the partisan, you just say they voted in these ways, then they don't, they don't seem to be important. So somehow, right, the statistics are telling us something potentially very interesting, which is that influence is not only a measure of the changes that someone can impose, but your interpretation of those changes as well.

 

Michael Garfield 50:23:

Gosh, you know, I'm just thinking about this in light of research that was done right before the 2016 us presidential election on search result ranking. I forget who it was that published this, but they were talking about the difference in a news item, making it to the top search result versus the second search result and how they were able to swing by modulating that across the entire population of people actually issuing these searches.

 

Michael Garfield 50:53:

They were able to swing electoral results by up to 25% in one direction or the other. And they actually tested this on some mayoral elections in India, where they were able to demonstrate that they were, they were able to throw, and they didn't actually do it. Oh, I was okay with it, it was a retrodiction of results that had happened, like a matter of weeks before. Yeah. But they were, I mean, they basically said, look, you know, the search companies need to be aware that the structure of accountability within these organizations, when you have an opaque algorithm for search results and the company, they can throw somebody under the bus who was working under nondisclosure on the algorithm and say, Oh, this was a rogue agent.

 

Michael Garfield 51:40:

We lack effective safeguards in society to prevent this kind of thing from happening because the kind of manipulation at scale that you're talking about does seem possible.

 

Eddie Lee 51:51:

Right. So, what you're saying is that it's not just the facts that you give people, but the salience of those facts. Yeah. And that's, that's actually very subtle because, you know, I certainly actually, I, you know, it was just last, was it this past week I heard a talk about this, but I just can't for the gut of me remember, but something about salient being really important for determining how people interpret events around them.

 

Michael Garfield 52:18:

So, there's another dimension of this paper, in your discussion on this paper, you're talking about this, not only in terms of voting results, but in the susceptibility of population to disease or disinformation. And so, you know, David and Jeffrey and many other people at SFI have written about, and we, you know, we had Laura Hibbard, the friend and the scarp, you know, on the show also recently talking about cultural contagions. Yeah. And so, I'm trying to wrap my head around how this model could inform a strategy for fighting disinformation in this way. And like, what would that look like?

 

Eddie Lee 53:02:

Well, you know, it's, it's kind of hard to make that connection. I think formally because we're looking at the statistics. So, we look over many aggregate dynamics and so on. And I think probably for something like this information who really do want to understand the dynamics of how things are spreading. So, there is that sort of shortcoming, but you could still ask, I think in that context, you know, how changes in the system could result or what changes in the system could result in the largest changes in the outcomes.

 

Eddie Lee 53:35:

And I think sort of to do that, you sort of have to rely on the fact we're focusing on this mathematical technique for, for studying the response of a system to perturbation. So, we're asking, we're sort of thinking about that in a statistical sense. It doesn't have to be in a statistical sense, but actually it sort of connects to a really interesting set of ideas of sort of tangential to this. And what it should say is this, this field right, is relying on this idea of information geometry.

 

Eddie Lee 54:07:

And what information geometry is about is imagining that if you have some sort of mathematical description of a system, it has to be parameterized. You typically choose parameters in a way that makes sense relative to what you're studying, but you could imagine changing those parameters. And you could imagine an alternate universe where a different sort of dynamics range. And so, this idea is that these models are all connected to one another. If you change the parameters a little bit at a time, eventually you get to those crazy alternative universes, which is totally different.

 

Eddie Lee 54:42:

But the path that you take is described by information geometry. And what we say is the curvature of, of that geometry. So how quickly it changes tells us how sensitive the system is, right? So, if I change this parameter a tiny amount and all of a sudden, I am in this alternative universe, then it's highly sensitive. It's highly sensitive to perturbation, right? Whereas if it's totally insensitive, two changes, then this primary almost doesn't matter, right?

 

Eddie Lee 55:14:

You basically get the same universe. It's slightly different. Maybe the colors of the cars are not quite the same, but it's almost exactly the same and what it turned out to be the case in a lot of physical models of the world is that, and you have many parameters that actually don't matter. And you have a few parameters that really do. So, for example, in the biology of a cell, you have these sophisticated bottles with tons of parameters that you can never hope to measure.

 

Eddie Lee 55:44:

And in fact, you can't really measure them anyway. Okay. But if you plug those values into the mathematical model, you basically get a cell that works. So, this is incredible because you're sometimes off by a hundred percent, 200%, but it's the works. Why it's because the information geometry is actually multi-dimensional and in some directions of parameter space, it's totally flat. So, all of the universes look exactly the same along that dimension. And it's very, very sharp along some other dimensions. So, you have to get some things right. And those things matter a lot, but most things don't matter.

 

Eddie Lee 56:15:

And so, this idea that you have this hierarchy of sensitivity, right, is what allows us to understand that the world, because if everything mattered, everything's out the window, you can never do that. You can never understand nature, but the fact that there's a hierarchy means that I can first build a model that sort of right, right. Newton's model or Copernicus's model, it's sort of right. It's not exactly right, but it's good enough that people believe me. And then, you know, Newton comes by and then Einstein comes by. But all of those additional effects features are smaller and smaller.

 

Eddie Lee 56:45:

And so that's what I mean by hierarchy of parameters. Right? And so that's what allows us to, well, that's a claim, but you know, that's, that's what makes the universe easy to learn about. And hopefully that is true. Generally. It's not, it's not clear that's true for every complex system, right. Which is why the complex systems are hard, difficult, but what's sort of interesting in the context of pivotal components is if there is such a hierarchy, then it means that this system is easier to control, right? Because you don't need every degree of freedom.

 

Eddie Lee 57:16:

So perhaps in order to facilitate diffusion of power, you want to design systems that are completely impossible to control unless you have all the fingers on the right triggers at the right time. Whereas if you want it to be controllable, then you need to design it with this idea in mind. And you can imagine situations where you want A, and you don't want B or you want controllability and you don't want non-controllability. And that's sort of the connection. There's a long-winded way of getting to this connection with, you know, susceptibility in say, disinformation is one might want to ask, how are these systems designed and are they designed in a way that makes them easily manipulated or makes them hard to manipulate?

 

Eddie Lee 58:03:

And do you want one or the other? Because you know, it could be that disinformation is really hard to root out when things are very decentralized, right? So, it may actually be a clash of values, right? You may actually want one thing, but it just leads to the wrong thing.

 

Michael Garfield 58:19:

I'm reminded of a, was it around the 2004 presidential election? Someone had a video where they taught a chimpanzee to manipulate one of the Diebold voting machines. You know, you can like, you can teach a Chimp to hack this, which brings us back all the way back around to the primate stuff. Yeah. You know, as somebody who has a lifelong abiding interest in both the sort of philosophy around evolutionary theory, as well as time travel fiction, you know, these, what you just described are the two sort of worldviews that you see at war in the debate over evolutionary contingency versus inevitability, you know, it's, it's ultimately, it's a dispute between two different models.

 

Michael Garfield 58:56:

Like, do you have Ashton Kutcher's butterfly effect, time travel where every, you know, like you keep changing everything on accident or is it more like back to the future where it only matters who sleeps together? You know, that's, that's, you know, somehow that, you know, the timeline is like completely robust against these perturbations. And so, you know, that's, I guess until we have this figured out, my advice is to like, not get in the DeLorean. Right. I don't know. Well, dude, it's, it's been absolutely wonderful talking with you before we wrapped this.

 

Michael Garfield 59:31:

I just an opportunity to tell people a little bit more about what's on your plate right now as a researcher and what you're looking forward to, what are the burning questions for you right now? And, on the horizon, in the months of confinement to come here,

 

Eddie Lee 59:52:

Well, I'm, I'm writing the second Principia Mathematica, you know, because people keep saying, you know, this is the time when you really get some work done. And I'm sort of past that now. It's been so many months?

 

Michael Garfield:

And it's horrible.

 

Eddie Lee 1:00:12:

If you keep telling people you're working on this grand project, nothing comes out while you're still working on it. No, I have, I have several things on my plate right now that are sort of on and off. One thing is sort of extending this armed conflict stuff to start really digging into some of the, sort of maybe contingent factors, but trying to find patterns in the contingencies that unify them. We're also been looking at Jeffrey West Chris campus. And I have been looking at metabolic scaling theory or in the forest. So, trying to understand some of the dynamics that lead to scaling and forest populations and hopefully actually touch on armed conflict in some way.

 

Eddie Lee 1:00:52:

I've also been thinking about acquisition of information in firms. So, trying to see why is it that firm lifetimes are distributed in a very sort of regular way, surprisingly across sectors. And does it have to do with how they're learning, how they're acquiring information from around them. I'm finishing a project with Jess and David on adaptation learning and adaptation. And it actually, the idea is that we're thinking about different forms of memory embedded in either the organism in its behavior or in its environment, and trying to unify these various mechanisms or implementations of memory into a sort of optimal adaptation framework.

 

Eddie Lee 1:01:36:

I'm sure I'm forgetting something, but it's a lot and, um.

 

Michael Garfield 1:01:39:

You don't have to keep fighting it. Right. If you've forgotten it, if you've heard that ??.

 

Eddie Lee 1:01:47:

There's no glue, you know, at some point it will lead to some cascading failure.

 

Michael Garfield 1:01:52:

Fair enough. Well, yeah, thanks again for being on the show and folks check the show notes for a link to all of your research and the other stuff that we've discussed in here.

 

Eddie Lee 1:02:06:

Awesome. It was fun.

 

[Outro]

Michael Garfield 1:02:08:

Thank you for listening. Complexity is produced by the Santa Fe Institute. A non-profit hub of complex systems science located in the high desert of New Mexico. For more information including transcripts, research links and educational sources or to support our science and communication efforts, visit SantaFe.edu/podcast.