Peter Dodds on Text-Based Timeline Analysis & New Instruments for The Science of Stories

Episode Notes

"There are decades where nothing happens; and there are weeks where decades happen.”
– Vladimir Ilyich Lenin

When human beings saw the first pictures of the Earth from space, the impact was transformative. New instruments for taking in new vistas, for understanding our relationships and contexts at a different scale, have in some ways defined the history of not just science but the evolution of intelligence. And now, thanks to the surfeit of textual data offered up by social media, researchers can peer into the dynamics of human society and analyze the turbulent flows of stories that drive our collective behavior and twist time itself into nonlinear structures. As a species, we are on the cusp of a new epoch in which the body politic reveals itself to us in real-time like a single human body in an MRI. How will these tools change how we think about the world and what it means to be a person in it?

Welcome to COMPLEXITY, the official podcast of the Santa Fe Institute. I’m your host, Michael Garfield, and every other 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 we speak with Peter Dodds of the University of Vermont’s Complex Systems Center and Computational Story Lab about how to use Twitter data as a kind of satellite telescope observing the collective mentation of humankind — what it reveals, and what it doesn’t, opening a cornucopia of questions about how we measure sentiment and the power of narrative for social control.

Tis the season, so if you value our research and communication efforts, please consider making a donation at — and/or rating and reviewing us at Apple Podcasts. You can find numerous other ways to engage with us at

Avid readers take note that SFI Press’ latest volume, Complexity Economics: Proceedings of the Santa Fe Institute's 2019 Fall Symposium, is now available on Amazon in paperback and Kindle eBook formats.

Thank you for listening!

Follow Peter Dodds at Twitter and read the papers we discuss (and many more) at Google Scholar.

And then go play with Hedonometer & Story Wrangler.

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:
Twitter • YouTube • Facebook • Instagram • LinkedIn

Episode Transcription

Machine-generated transcript by and lightly edited by SFI's Mindi Madrid. If you would like to volunteer to help us edit podcast transcripts, we would love your help — please email


Michael (1m 34s):  

When human beings saw the first pictures of earth from space the impact was transformative, new instruments for taking in new vistas for understanding our relationships in contexts at a different scale have in some ways defined to the history of not just science but the evolution of intelligence. And now thanks to the surfeit of textual data offered up by social media researchers can peer into the dynamics of human society and analyze the turbulent flows of stories that drive our collective behavior and twist time itself into nonlinear structures. As a species we're on the cusp of a new epoch in which the body politic reveals itself to us in real time like a single human body in an MRI. How will these tools change how we think about the world and what it means to be a person in it? Welcome to complexity, the official podcast of the Santa Fe Institute. I'm your host, Michael, and every other 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, we speak with Peter of the University of Vermont's Complex System Center and Computational Story Lab, about how to use Twitter data as a kind of satellite telescope observing the collective mentation of humankind, what it reveals and what it doesn't.  


Opening a cornucopia of questions about how we measure sentiment and the power of narrative for social control. Tis the season so if you value our research and communication efforts, please please consider making a donation at You can find numerous other ways to engage with us at Avid readers, please take note that SFI presses latest volume, Complexity Economics, Proceedings of the Santa Fe Institute 2019 fall symposium is now available on Amazon in paperback and Kindle ebook formats. Thank you for listening.


(3m 45s):

Peter, welcome to complexity podcast.  



It's a delight to be here.



It's a pleasure to have you. So before we dive into the numbers and the sort of God's eye view of things that you have revealed through your research, I'd like to start with a bit of human background anchor this in a bit of story. You like wrangling stories so what story can we have about how you got into your life as a researcher, and what drew you into the kind of questions that animate and inspire you to do the kind of work that you?


Peter (4m 28s):

So that could be a long story. I guess in short, I'm sort of a meaning seeker, you know, I'm kind of one of those people, like what, why does anything exist? Kind of all this sort of stuff. And so that's drawn me to think about lots of things from my life, you know, religion, science, all these pieces, all these ideas of how you should live your life and how we are living our lives. So I'm just this curious person, but in terms of coming to think about complexity, you know, there's a friend of mine, I'm from Australia originally, that handed me a book and said "You should read this book." It was Mitchell Waldrop's Complexity book.


(5m 10s):

This was a long time ago when I was an undergrad and I read it and I thought this was fantastic. I ended up getting to the States as a result and since then it's just been, what can we understand about everything. So I could've end up in string theory but that's not where I am so this opening of thinking about human systems and systems that humans are attached to, so you know, climate change as part of that, this is kind of everything we touch on now. When I think of a socio-technical system it just opened up and up over the last 20 years.


(5m 50s): I've just been drawn along and there's always something new to think about. I will just sort of end by saying, we want to create things that are meaningful, that help the world and that matter.  I know that students over the years have been attracted to the work we do so that's good. You know we want to do meaningful things. We're not just aliens watching humanity, we're part of a system and we care.


Michael (6m 20s):

So yeah, you brought up aliens. We are the universe becoming aware of itself type sentiment that I think really sings through the work that you do and the work that you and your community of people around you are doing at the Vermont Complex Systems Center and elsewhere.  The right place to start here, I think, is just with the idea that the kind of work that you do is even possible. So I would like to talk a little bit about the background and the conceptual framing for Story Wrangler, because this is the kind of thing that I don't know, maybe when I was in high school would have been in like a minority report style science fiction films, and now here we are. In one sense, utterly prosaic and mundane, and in another sense, completely wondrous and numinous, and honestly kind of terrifying.

(7m 28s):

We'll put a pin in that and get back to that. But I'd love to hear you talk about how this project came to be and you know, how people started to thinking about how to ask and answer these kinds of questions.


Peter (7m 42s):

Well I think thinking of stories collectively and at the population level is still pretty primitive.  I've come to feel it's just a crucial, crucial thing in science for us to understand how people tell stories' how they believe stories, how they spread stories and how various cultures across many different fields have studied this for a long time. But I think there's an opening now in this sort of big data world to maybe start to create something that's really solid and endures. It goes back to us measuring happiness was this sort of first big data thing that we did. And before that, I had done experiments looking at how things spread, how things take off


Peter (8m 25s):

so I think we have an understanding of same from that. So there were experiments and that was enabled by finally were not studying 100 people in a psych one-on-one course or doing surveys. Those things are still important, right?  I would never pretend that our work wipes out other fields, it augments it and adds to it. So going back in time we started to see that we could do these experiments that involve hundreds of thousands of people. And then data came along, social data and to some extent it's the social media stuff but blogs before that, perhaps, so people are putting these expressions online and


Peter (9m 5s):

it's not being filtered through experts and a few anymore. You're starting to get the murmurings saturations of a populace, you know, and what do you do with that? I mean, we can see the stars, right? So we start off looking at them from thousands of thousands of views and eventually we get some telescopes together, pretty terrible to start with, but, you know, we keep getting better with that. And then we have a raise of telescopes and it switches completely to, you know, much more of a data driven kind of situation where you're not necessarily ever looking through a telescope. So I feel like I looked at that, that sort of history, the history of building thermometers, the history of building clocks, you know, and you can go back and think about how we just, we couldn't measure these things well.


Peter (9m 48s):

And in some cases like temperatures is an example, I guess there's some argument about this, but maybe we never thought that was a number that you could measure, right? Someone else could have make another instrument and measure the same numbers as you. That there was some sort of human element to it that made it fuzzy. It's different than measuring distance. We were so solid on distance that that becomes this substructure for the metaphors, for everything we use, even time, you know, distances is fundamental thing. So, you know, we've slowly as scientists gone out and measured everything. And I think of basic sciences describe and explain, right?


Peter (10m 30s):

And then after that, you get to create, you get the goals and predict all those things. We kind of race the prediction, but you know, the call for me is describe and explain, and you've got to describe. And so I feel like we try to do everything, but we're a bit in that game when we start to measure what we'll get to hear as stories, but it was, you know, what do we do with all of these inputs coming in from the outside world.  This is now through the internet, you know what can we sort of sense from that? We had this big idea to start with which was what about happiness? Right? So at least that gives us something we think is important for society. You know, it's a bit of a popular framing, but it's different than a GDP, right?


Peter (11m 12s):

You know, at the end of a news pieces like local news, the GDP went up on the market, it's like this really important an thing? Same with the weather here are these basic things of today. And what about something like happiness well being, these are harder things to measure. And in fact, when we went to look at how to measure happiness, and this is from texts and ultimately from Twitter and books and all sorts of things, what we look back on on psychological research, happiness is really the first dimension of meaning. And this is based on work, asking people and the semantic differentials. How do you feel about something on some semantic differential, rough to smooth, happy to sad, angry, powerful to week, do you feel out of control, excited to bored?


Peter (11m 59s):

And it turns out that these old studies at least came up with these sort of three main dimensions and their happiness balance, this kind of happy sad axis. And that's really the dominant one. And then excitement, do you feel excited for support and then power, dominance? So we've sort of stripped off that first dimension and we applied it to all sorts of things, including the count of Monte Cristo or Frankenstein, right? So you do those with books, try to measure happiness in this kind of big reading way, right. And we've done it with Twitter, which we'll get to a Story Wrangler. And that goes back to 2008, it's daily, we're going to have within a day stuff soon, I think. What we're trying to do


Peter (12m 42s):

there is just take all of twitter in different languages and you don't want to read 50 million tweets a today, right? We don't want anyone to do it. No one should do it, but know we can talk to our computers and they can sleep past them and they can get a feel for the whole thing. I'm only being able to show is that that correlates, let's say state level well with Gallup polls, it connects well with major events that happen in the world, particularly negative events. They really stand out and there's still a long way to go with this, but this is a kind of a, I don't know, it's a remarkable thing to be able to build. We're very worried about it from the start thinking of this leading to like brave new world or, you know, being part, you know, that sort of these foundational texts that we should worry about all the time or 1984.


Peter (13m 28s):

Right? So there's the big we're watching everyone kind of thing. Of course, we only look at public stuff and then, you know, what do you do? What's your public policy as a result? You know, what do you do? And our framing is, this is just something more to put on the dashboard, right? You're flying a plane. We understand with planes, we want lots of instruments in front of us. We don't want just one number that says plane okay, plane not okay. That's a bad dashboard. And I will say this more generally about complexity, we have a huge proclivity as human beings. And some scientists inclusive to really want to measure something with one number. Like he's this huge complicated thing and let's just get it down to one number because then we can talk about it, it'll be easy. It's a real disservice, especially from the scientific community.


Peter (14m 8s):

So we're very mindful of that. We're trying to just put another number on a dashboard, and then we want people to look at, I mean, if you think about the ecosystem collapse or something like that, there are a number of dimensions that you should be looking at. You know, you don't want to be just sort of thinking about some big number, you want to be looking at Keystone species. You know, all these different pods, sort of systems. So that's the happiness work that's ongoing and what we're trying to get to now is, you know, and many people sort of, I think, coming around to this, like, how do you measure those stories that have been told by a population. And we deliberately want to be away from individuals because we don't want to be invading their space.


Peter (14m 49s):

We don't want to say, look this person did this. So a lot of that stuff is very thermometer like, it's very distant. So it just by its nature will not function at a small scale. I mean, I think there's both a privacy concern there that's felt. And then the second is just also doesn't really work and people will tell you it works, but you know, you hand someone a tweet and put it through their machine and it says happy or sad maybe, but I think that's pretty dangerous. You know, individual sensors, these are hard things to base your whole decision of how to do something based on one statement, we can argue about that. So for instance, Parkland happens and now I'm well-prepared to understand what happened the next day, which was conspiracy theories emerged.


Peter (15m 33s):

And I remember going to YouTube just to see what the balance was, and that's sort of typical thing where you search and they give you 20 results. The 20 top hits, 18 of them were conspiracy theory videos about crisis actors and so on and two were just straight up news pieces. How do we measure that and how do we track them? Just, and you know not conspiracy theories aside, like reasonable narratives around something. It's hard to know what's going on. How do we do that? What are the stories that people hold in a population? And how is that trending over time? And again, you know, this is not to create a system of control, but it's partly to give people back information, you know, look, here's your world.


Peter (16m 13s):

People are not happy in this year compared to last year, or here are the stories everyone's talking about. So you may think because, and I think this is really a fundamental thing about social phenomenon, is what you might think about what everyone else is doing. Well, I don't know. Right? So you don't want to ever really project your own family onto the world, or, you know, what happened to you with your friend? You know, you can think about that, but it's, it's pretty fraught. And I think, you know, we need to have these sort of big picture views. So feeding back to individuals, feeding back to populations, but also to policymakers, I do think I, you know, they need to see this in the dashboard, not just money went up.


Michael (16m 49s):

So you're talking about the heatonominator. Yes, Which people should go look up online. It's, glorious. I mean, Glory is in a kind of a tragic way because you do actually see this who we've been slumping as a society as measured by this instrument over the last several years and that sort of begs its own question about the causes behind that. I think we might be able to poke at that question throughout the rest of this conversation, but then this other piece, which you and your coauthors on this piece laid by, is it Fairalshabi?


Michael (17m 29s):

An interesting article, "Story Wrangler, A Massive Exploratorium for Socio-linguistic,  Cultural, Socio-economic and Political Timelines using Twitter." Yes. Quite, quite a masterful mashup of all kinds of different interests in this one paper and you make a point here that, you know, people might be familiar with Google angrams, google angrams do not give you a sense, in spite of what you might think, of the popularity of a given word or phrase, because it's just how many books were published in that year.


Michael (18m 11s):

And it's not how many people were actually talking about those books. So like you say here actually 1984 or the Percy Jackson books of Ricky Arden, they're read and re-read, they're shared spread around. And so there's this question, not only of sort of the number of species, but of the interactions of species within this memetic ecosystem. If I can annoy some people with my flagrant use of the ecosystems and memes in this conversation. I mean, there's just so there's so much here, you know, and so I would just love to hear you talk a little bit about first of all, I think you're right,


Michael (18m 56s):

I think that this it's very clear that reading this work and looking at the diagrams in this paper. That what we're seeing is sort of like the earth from space, you know, like Stewart Brand petitioning to get NASA to declassify those blue marble photographs back in the sixties. And now it's like, well, here it is. You know, and like maybe Twitter was doing this. We know Facebook is looking at this stuff, the orbital view of people in this way that they, you know, they have massive teams of internal researchers, but it's awesome to actually have your hands on this kind of a dashboard and to be able to search.


Michael (19m 39s):

I was looking at, because I'm a dork, I was looking at dinosaurs on story Wrangler. And then I was looking at where the spikes were and then I cross-referenced it to Jurassic and sure enough, the major spikes were concurrent with the release of Jurassic park movies and video games. So, you know, I was like, Oh yeah, yeah, that makes sense. But it was nice to see that there is a sort of a, a hum of continuous interest in dinosaurs. I was like, yes. So the geographic features of the knowest sphere, right? Like VI Vernadsky's notion that what we have done is we've created this layer of mental activity around the planet and you and your colleagues have added this extremely important dimension, which again is the relative population sizes and rates of transfer and the relationships between all of these different ideas as they're flying around in this, this meme space.


Michael (20m 44s):

So I'd love to hear you talk a little bit about how you actually went about partitioning this, doing the research, organizing this data, and then maybe perhaps, you know, what were some of the things that surprised you in your findings in this piece?



Peter (21m 1s):

Well, so I want to say overall, we've set up this site, it's and it's just an opening, I think, to an enormous amount of research. So we're really excited. So in part, we're putting this out as a massive sort of data resource for people and we're kind of excited about what that will build, but it's something that we've wanted to do for years and we've made little sort of pokes in this direction, but this is going back to 2008 is when we first wrote to Twitter and my colleague, Chris Danforth wrote to them and just out of the blue and said, could we maybe get some tweets? And there were only four people working there.  It was a small thing at a time and they said, Oh, sure, so we have this little sort of feed.


Peter (21m 42s):

So we've been getting 10% of the tweets since then. So in terms of, you know, from the data source point point of view, that's how it started, just from an email and largely at no cost which has been a tremendous service. Twitter has gone through various stages and it will keep going through them being knocked as being frivolous and silly, and it will go away. Of course, it's been rather important I think in the last five or six years. It's also perhaps held a place that Facebook hasn't in terms of this, if people put quotes out, put statements out, it's the stuff that gets embedded in other news stories. So they've really kind of held that space quite well, but it is remarkable that so much is on Twitter.


Peter (22m 24s):

The dominant thing I'll just say from the start is it's kind of K-pop right. So K-pop is enormous and and we just have to acknowledge that, but music is, sports is enormous, you know, these movies, as you pointed out, right? These, when some of those advertising, some of it is just, people are excited about it, TV shows. So that's sort of, I feel, like just to kind of talk about what it is that's become to me like the resting state of the thing. Like when Twitter is going along, you know, when the world is sort of going along somewhat peacefully, nothing terrible has happened. That's the stuff that's going to be bubbling to the top, which seems, okay, it's sort of water cooler discussion writ large, right, At a population scale, but to really process this data well, it's taken many years graduate students to come along and try all sorts of things.


Peter (23m 9s):

And we use a supercomputer here at UVM, which seems maybe a lot for tweets, but it's a big database and we'll have to say that emojis have proved to be an incredible complication from a data point of view. I mean, they're just curious animals, right? I mean, if you're an ecologist, this will be like tyring to catch ghosts or something, you know, it's really, really, really a mess. So we've dealt with that and you know, they're all in there as well. You can see the rise of emojis in particular ones, instructions. So this is one of those things you will see, like you mentioned Google angrams, right? So that's from books. There's a nice little viewer and it's easy to go to. And as a user, you think, Oh, this is really easy,


Peter (23m 50s):

it's a thing but the backend of that is really, really hard to create because every day you have to look at the forest. I like the ecology thing. There's this forest of this ecology of words that's put out and we have to kind of take it all apart and count up all the little bits. And then we look at the next day, you know, how much did this forest turnover from day to day, to day? And for Twitter, it's pretty rapid, you know, a new species can come and go on in a day in terms of angram or whatever. There is just so much to look at that you can think about the in training of like really basic things like how will you function because we're in a solar system, right? So you see the patten of the months and the lunar cycles and all those things in years.


Peter (24m 34s):

I mean, that's going to sound trivial, but if, you know, it's there, you could kind of actually detect that from Twitter if you looked at it and you had nothing else, you could figure out for 65.2, it's, it's a curious thing. But then you start to see, you know, events that we celebrate every four years or something like the Olympics, you start to see those kinds of spikes and then it gets more complicated when you get into political events and turmoil and, and changes. I mean, I'm just looking at one now that we've been thinking about, which is the rise of the term fake news, right? So we have separate work on Trump and all the tweets that contain Trump. So just to connect to that later, you know, it's really, after Trump comes into that, he's elected and I can talk about that a little bit of dynamic that, but fake news goes from just the term that appears a little bit to a completely new state, but now it's a stably used term, it's just in the language, it becomes a species in the forest that was not there at all.


Peter (25m 32s):

And now it's just everywhere in the forest. And it's just a normal thing to encounter if you're walking into this forest of angram's. One of the next pieces we're trying to build now, this is kind of complicated, so you've got every day all these words are competing with each other and we'll rank them, right. We'll rank the words and I know there are a hundred languages so you can look through it in that way, but we also then want to be able to do this next step, which is to say, okay, you want to find all the tweets that continues an use and you want to then look at the story Wrangler for that. You want that first. So we're going to take that subset of tweets, which conceptually is easy to think about, but basically it means is universes within universes, you know, there's just this incredible sort of multi-verse linguistic, lexical multi-verse if you like, right.


Peter (26m 17s):

So you're going to say, I can go to this forest so I'm just going to sort of take all the Oak trees and everything else deletes. And then you're looking in that subset. So I think that's going to be powerful in terms of looking at how meaning changes around terms like news, for example. So fake news, you know, that that word is not attached to news and then suddenly it is, you know, what else has happened around a term, which is otherwise very stable. News is very stable over 10 years, the word news, fine. If just looked at you and say, Oh, you know, I guess it's just a fun thing, but it's the stuff that's, you know, code around it, rich complicated and you don't really, you know, in the world we live in really meaningful, right? This is a really powerful thing that's happening, the sort of transition to de-legitimizing the press and so on.


Peter (27m 0s):

So we have this online piece, you can play around with it. We have, I think, a really important part which is, and you mentioned it before with, with Google books. So we had an article, I guess it was published in 2015 where we brought out this critique of Google books, which at that time sort of remains now this kind of incredible lens into the last couple of hundred years, especially. And so that's taking old books and just counting up the woods. So each year it's at the year scale, the problem, yeah. Is it's the same as going to a library. And every book gets one book and you don't look at how much a book has been checked out or read if you could see that as well. And so that's obviously not right, right?


Peter (27m 41s):

It will give you something and it will look like popularity because you know the word, these huge ran the world Wars. Well, those words pick up, but it doesn't tell you exactly what people were reading. We do this all the time with corporate, I should say too, right? I mean, we do this all the time. People will take the New York times, study the New York times as it goes for time, but not index it by which articles were writ because in part, you know, it's easy to do that. First part you just have, the New York times is a copy of the paper kind of thing. And I think sometimes we know that, but sometimes we forget. So wiyh story Wrangler we have built into it, a little toggle, it's include retweets or not, which is very nice feature that we have for the system.


Peter (28m 22s):

We can just see what was amplified so we can kind of recreate the amplification that's in the system, or we can dial it down. We can pull that away. And it's been a really interesting thing to look at. So Twitter is deliberately recently tried to tame the amplification that's on its system and something I've talked about for years. So for instance, Instagram, for whatever problems it has, it doesn't have like a re Instagram. It doesn't have that kind of copy. If you're talking about memes, it doesn't have that kind of built in where you throw it out to everyone else, it's not built in, Twitter of course does, and Facebook as well. So if you remove all the friction and make that really easy, you know, amplification is really easy


Peter (29m 2s):

then, you know, you're basically lighting fires all the time or enabling it to spread through your forest, let's go back to that. So you can see, we can see that that's actually, we have these little contagious grams. You can see that that's being tamped down and that's not like the deleting accounts or deleting tweets. They're just simply making a little bit physically harder. Like you have to click on two or three things to retweet something purposefully, anyway, so that's a really important thing here. So some things really do get affected by whether they're amplified or not, Right? You know, when you turn that piece off. So actually I think maybe its was a simple thing to say, a lot of junk tweets


Peter (29m 42s):

They don't get amplified. So maybe astrology tweets, which are just sort of thrown out into the world, you'll see them if you turn off the retweets that kind of like they bubble up a little bit, but this is a forest that hasn't had the very imperfect and erroneous filter of humans. And I'm bots sitting on top amplifying things and spreading things. It hasn't had that filter and we don't always amplify the right things. We know that. And that's from some of that work and other people's work, right fame, the things that take off, they aren't always right at all.  


Michael (30m 17s):

It's funny that you mentioned that I was looking at the endgram history for complex systems. And like one of the all time high spikes was actually captured texts from the little preview that you get when you shared a horoscope. It was mentioned that an Aries horoscope in like 2016


Peter (30m 39s):

And I was like, Oh, the work that we have done to promote this thing and like, of course. So you really do get it, you get a clear idea of what is actually capturing people. It's interesting that, you know, to extend this forest fire analogy, certain species thrive in systems that are regularly cleansed by fire and other species take over and then they don't. And yeah. And so I think that's sort of where Twitter is with this whole thing, you know they're actually trying to sort of shape the ecosystem, trying to make it less habitable or inviting for invasive species that know how to capture attention.


Peter (31m 23s):

Yeah. And so you have tweets that pour kerosene all over themselves and then, you know, basically they just, you know, I mean the most, your average person they're tweeting about a cat or they're tweeting about something. And so they're not trying necessarily, but if you have a sophisticated outfit, you can make it happen. So your algorithms have a built in structure that you have to be very mindful of. And people are very cleavor.


Michael (31m 43s):

There's one question in here that, I mean, this should possibly be self-evident in retrospect, but you and your coauthors mentioned in this paper, we see the time series for scientific advances, generally show shock like responses with little anticipation or memory to think about what you're actually reading as a kind of data visualization of again, the anticipation or memory of society. Then I don't know that's telling, that says something about the way that we habituate, I guess, to discovery, you know, like obviously we're not going to necessarily expect a particular, I mean, that's the whole point, right? We're not going to expect a particular discovery, but the fact that it just seems to fade from collective memory so quickly seems to say something about, perhaps it's best understood through the counterexample that you give, which is crisper, which is something that seems to really have captured the public imagination and is landed in a way with people that they understand it has, you know, ramifications that something like the discovery of gravity waves does not. But I'm curious what your thoughts are on what it takes for something to persist in cultural memory, or, you know, what this reveals about the mechanisms of cultural memory and its relationship to anticipation.


Peter (33m 12s):

I mean, this is such a profound question. When I mentioned fake news before, that's really actually what we're trying to get out there. We're trying to find all of the species that invaded and became new durable elements of the system. So I think we can actually do that at least, you know, in this second, you know, for this Twitter example, we can do it in kind of an exhaustive way, right? And so we'll have this, here's a list of things that became part of what we talk about. They became sort of the backbone of our conversation, normal, everyday durable things. And then we can look at the taxonomy of them, right, So fake news is a political thing. Like inception, for example, that movie really blew people's minds and still confuses, you know, it's still difficult to do with, but that became a thing.


Peter (33m 56s):

But for something to really take off, it has to get bound up in stories. It has to end up in stories, it has to end up in movies, your average person talking about it has to become something that they will use. And so when we see something become durable that's what's happened, you know, gravity waves was a cool story. We're kind of, there's a predisposition perhaps by going that science will produce cool things now, and then, and we'll get to be excited about it, especially like space, you know, that's really cool, but there's a huge amount of competition for attention and the stories that are bubbling and so on, the best producers of that, you know, the most durable ones, as I was saying before movies and sports and music, because there's this constant sort of creation of things and you know, where do you get fandoms?


Peter (34m 43s):

Where do you get fandoms? Fandoms built around those things? You have a sense of fandom around Trump, right? I mean, so you could kind of build them out as sort of analogous kind of groups. So when does that happen? This is the first sort of cultural products for terms that's, something else, you know, something becomes just something people, you know, like I dunno someone in a family system, someone else, you know, that's fake news. Like it becomes kind of a joke, a term that you can use at any point. I think we can do that. Of course we can't go back. This is Twitter, right? So we have this incredible resolution, this temporary resolution. That's something where we're sort of enjoying, of course it doesn't go back thousands of years.


Peter (35m 27s):

So that's a loss for us. It would be amazing to have tweets from whatever period you like, but yeah. Yeah. So, no, I do think scientific things kind of fraught in that sense. They do need, so science does need storytellers, of course. And it's really important for that to become part of popular culture for movies and these things to excite people and to make it kind of stick. And it will be interesting to see the vaccine, right. That vaccine is moving along, but there's been a jump in 2020, because it's just being talked about all the time. Essentially, just the jump with coronavirus coronavirus pandemic becomes real on March 12th.


Peter (36m 10s):

And then you see on our Story Wrangle viewer, that vaccine just sort of goes into a different kind of regime. It's just being talked about a lot, but it's been talked about a lot very recently, of course, with Pfizer and the other company claiming that the very close to asking for, I guess, permission to sort of build and distribute. So, you know, that's a scientific achievement that I think, you know, we may see more of a durable response to it. I've always wondered what would happen if that was just a straight up clear cut cure for cancer or something like that. What would happen with that? How would people respond around that? I will say people, people do tend to forget things pretty quickly.


Peter (36m 54s):

There's a lot of stuff passes away. And so that's been of great interest to kind of find what becomes durable. But yeah, there's a lot of stuff you think, Oh, you know, that's a big deal, but it's just a blip. Yeah.  


Michael (37m 6s):

It sounds like the right foothold to pivot or leverage ourselves into this other paper that you lead authored on Computational Timeline, reconstructions of the stories surrounding Trump, you know, using Trump as a case study to explore these concepts of story turbulence, narrative control, you and your coauthors call out, I love this, collective chronopathy. This is so so true. You know, this idea that the perception of time is subjective, and we know that perception of time is linked to the pace at which a person is experiencing and has to process information.


Michael (37m 50s):

So this gets at something that you were talking about in this paper, we were just discussing, which is competition for attention. And I would like to hear you introduce this piece in context of the question of, have you noticed any changes in the dynamics of attention over the history of these data sets? Because it would seem that, for example, like almost everyone knows the Beatles, but now a musician can sell more music than the Beatles ever sold and be known by a smaller fraction of the population. And so there's something going on


Michael (38m 30s):

it would seem with all of that. This gets to something that you were talking about in that first paper about the way that you can use this kind of analysis to forecast social fragmentation. You know, this is aligned with the work that David Krakauer and I talked about in the transmission series by Miguel Fuentes, he was looking at the fragmentation of social graphs, proceeding, social revolutions. At any rate, that's a whole basket of questions, really, but these are the thoughts on my mind when I'm reading this paper that you and your team wrote about Trump and what you found in terms of like reconstructing timelines and how you use this, you know, contagious programs to reconstruct narrative control


Michael (39m 12s):

and then what that says about the way that we perceive time and then this, quote, unquote, as you mentioned earlier at the beginning of the call, the distances between moments, and then what that means in terms of how these different histories, as they are held by different groups, how these different narrative framings are sheering against one another to create what you call story turbulence. So, yeah, it's a fascinating thing. I'd love to hear you riff on this for a while.


Peter (39m 40s):

Yeah. So this, this has been a couple of years of my life just really dedicated to this. I mean, I just really wanted to understand particularly this, this idea of turbulence, right? The turnover of story, and that led to these other pieces, which is, one is specific to Trump, but could be used for others, which is narrative control. And I'll get to that. But because we made up this word, I sort of use the old pronunciation, which is chronopathing. That's what happens with these things. So, but yeah, so the feeling of time passing and the idea that's collected that there is there an individual one, but there's also maybe something we can measure at a collective scale, but it was about story turbulence, which is just, you know, how much churn is there and is that changing over time?


Peter (40m 28s):

Is it always, even if there are these major events and you're kind of like put everything together, there's just sort of this churn. But we can also look at other things like baby names. We can look at the churn of baby names or just, and you could go to Google books and you could look at the words being used from year to year what's the turnover. Now somewhat very famously in complex systems there's this idea of heavy tail distributions and distributions in particular, right. Which goes back to George Kingsley. So it's this idea that if you rank things from biggest to smallest, and there's this heavy tail distribution of them, and this is a power of size distribution, right? So lots of rare things or small things, some really dominant ones. And there's this idea that those distributions don't change much as you go through time


Peter (41m 11s):

maybe. If you, look at book to book and so on, but of course they're categorical distributions, you know, there's a label, you know, whatever it is, you know, Los Angeles, New York city, if we think about cities, they're actual labels, right? So if you look at this Ziff distribution, this ranking of the biggest things, and you can think of music rankings, you can think of soccer rankings, whatever you like, as you look at that through timed maybe the distribution of their sizes, it doesn't change much, but of course there's a lot of this turbulence. And if we look at language, the words the and of say in English, yeah, they're pretty much at the top. So they're really pretty stable over time. There's not much change in the top 10, over a hundreds of years in languages, but as you go down, you start to get more and more, and I've talked about some frame, this is turbulence, and also sort of invoking this idea of influence of scaling, right.


Peter (41m 60s):

That there is a scaling that, that turbulence increases the further you go down and rank, and this there's quite beautiful scaling that you can see. All right. So that's sort of the, to sort of put it into complex systems in general markets. I said, baby names, ecological systems like numbers of species and forest and so on, but text is amazing, right? There's just so much that we have. All right. So we took, if you, like, we made story Wrangler for Trump, we took all the tweets that just matched Trump. And so that includes his handle. It includes trumpet. It just, we just did a very simple thing and it was enough to work because most of the tweets really about Trump and includes his retweets, you know, people re-tweeting, so that's built into it.


Peter (42m 45s):

And I will just back up one step and look, we have another paper where we've just simply looked at how much major political figures in the U S have been talked about since 2008 on Twitter. And that's it just very simple and there's nothing like Trump, right? So there's this big buildup around Obama. And it's talked about a lot on the election day in 2008, then it starts to fade away and then it's kind of up and down. You know, things happen, people talk about other stuff and the president matters, but it's not, you know, it's not always in view. And if you look at over time, these sort of like maybe in the thousands in terms of rank, and that's kind of like the UK, you know, sort of like dominated by the U S right. Twitter. So Obama in the UK kind of have a similar sort of name recognition, you know, like amplification.


Peter (43m 31s):

And if you think about marketing, right? Like how much do you want Pepsi or Coke talked about how, you know, what would you say? Like, you know, when you get your brand talks about, okay, so sure you want to be further up the ranks, but what's crazy about Trump. Trump has, since he took off and won, the election has been basically in the top 300 words every day, which puts him in the realm of function words. So these are just functioned words. So he is in the last three years, three or four years, he's like the word would, or man, like just, or say, just the most basic words that we'll use over and over again. He leveled up enormously and he was kind of fading after the, it was sort of going down and there's all these details, you know, what spiked and so on, early on, but you know, when he gets out, he's just, it's really kind of extraordinary to look at.


Peter (44m 21s):

And of course it's been talked about more than ever recently. So he's an unusual figure. There's an enormous amount of texts about him, you know, everywhere. But in terms of Twitter, we recreated this Story Wrangler thing for all the tweets that contain Trump. So now we've got this forest and every tweet, if you read it, we'll have Trump in it. And what we want to see is, you know, what are the sort of what's being talked about? And so we have a way of sort of trying to find the most narratively dominant word, which is to take a forest of words around Trump on a day, and look back a year ago. So that kind of takes out the year signal. And, you know, what's a surprising word because every day, you know, you still going to see if the awe of it, right?


Peter (45m 2s):

The basic words are still there, but you want to see one of the big movers, right? The surprising ones. And so as you look through time, you know, we've done this at day scale, week scale, month scale to kind of like, renormalize it, of course, 2016 is about the election and it's usually his opponents, right? So he's sort of, they're being talked about when you get in 2017, suddenly things kind of really take off, this year inauguration. Right. That makes sense. But then pretty quickly you get to Flynn, Russia, Komi, Mala, all those things come along. And then it can be very hard to remember all of this after say, what's their favorite, but North Korea, that's in August. And really, and that leads into the Charlottesville weekend.


Peter (45m 44s):

So North Korea, rocket man, all that sort of stuff, that kind of posturing, suddenly you have Charlottesville. Then you have hurricanes, Maria of course being devastating. All of these events 2017 in particular was all. And then the Mueller report is developing the background. So we kind of get that out. So it's like creating a, kind of a competently enabling historical analysis because it sort of produces the keywords, the dominant terms, one grams and two grams around a figure in this case, Trump and some of the frozen up and says, look, here's the backbone of what was talked about. It's true. It's through Twitter, but Twitter, you know, this is not an opinion thing. This is much more about what was the dominant stuff that was being talked about.


Peter (46m 25s):

Opinion is sort of more buried in here. So, you know, when you go into 2018, there's all this turnover. Again, Kavanaugh is a big piece there, for example, but towards the end of 2019, it becomes much more about impeachment. And then 2020 is really just this extraordinary year. I mean, we've never seen anything like this in, in all these years that we've looked at for hedonameter or for story Wrangler. So just to jump back to Hedonameter, they're the two events of coronavirus becoming widely accepted to be a really, you know, we talked about January and so on, but it's March 12th, MBA suspends the season, Tom Hanks announces he has coronavirus and Trump gives a speech. It doesn't go well in the oval office and the markets tank, not that all sort of happens within 10 minutes in the evening of March.


Peter (47m 12s):

Well, it's all happening. And it's just, it's kind of crazy. You know, it's, it's a crazy confluence of events. Hedonameter it just dropped. And then it took a long time to come back and we've never seen a collective response. We've always seen like something bad happens. You know, there'll be memories of those words that we certainly be talked about, but it gets washed over, you know, it gets washed over. And I look, I'll mention a particularly awful one, which was the Las Vegas shooting. This was horrific, but you know, now Hedonameter, there's a big drop and then it's gone. It doesn't mean the Las Vegas isn't still being talked about, but the wash of stories kind of wipes it out. When we see coronavirus that takes months to sort of get back to some kind of normal.


Peter (47m 53s):

And then George Ford's murder pushes it down to the lowest we've ever seen. And then again, that takes sort of a month to come out of those sort of signals are much more like what an individual might have. And so it's, that's real collective trauma. And these, these are real signals of collective trauma, ive never said it before. So, you know, throughout our instruments. So, but for Trump, what happened in 2020 is if you go back to the start of this year, the first thing that happened was Soleimani was assassinated, right? That was the start of it. Then the impeachment comes back into view that becomes dominant and then Corona virus takes off. And that becomes, he can't break that story. That story dominates Trump it's around Trump for really months on end, as you might expect.


Peter (48m 35s):

But then actually George Floyd's motor comes along and that is that him to talk, but that is an in for him. So when we look at narrative control, what we try to sort out there is these dominant terms around Trump. How much of it is actually just true to retweets of him? Know what fraction? And if you go back in time, you're crooked Hillary that's him, but also fake news, all the uses of fake news that you see connected to Trump and that kind of, that this virus we've made that's just the Trump virus, you know, whenever you see fake news, all of those fake news animals that you find in that forest, they're all retweets of him, basically, which hunters like that. So there are pieces that he's putting out, but he could not break coronavirus. He didn't use that term.


Peter (49m 16s):

He used other terms, some of them were derogatory, but he couldn't break it. And you can look in there and you can see like Obama gate, he's trying to throw in other things hydroxychloroquine or something. But he's trying to talk about other things altogether. He's trying to talk about corruption in the Democrats and you can see they kind of get into the top 10 words around him, but they never break this kind of I'll use the word wall of coronavirus that just stands in front of him and sort of what he's trying to get out publicly, but it's been much more jumbled since then. Pandemic has come back. COVID has come back. Those words have come back. So when we go to stories, let me say this too. We have our work studying pandemics after SARS, right? We looked at pandemics, we looked at historical pandemics and you know, what we're able to see there is that they're the most unpredictable of natural disasters.


Peter (50m 5s):

Let me, let me say it like that. You don't know how far they're going to go. They're very difficult to contend with this deep uncertainty and also the potential for resurgence. And we're seeing that sort of now, right? We're in this third major wave in the U S but this idea that you think the thing is going away, but it sparks up again. So the classic, if you go to simple narratives that, you know, the sort of core stories we have, right, there's romance, there's journey, one could argue, let's kill the monster, kill the monster is a story we tell over and over again, Beowulf, you know, stories where you overcome cancer. It's such an important Trump, and it's such a fundamental part of survival, right? This terrible thing happens. And it's not like you want to get out the other end of being, you know, 10 times better.


Peter (50m 45s):

You just want to get back to where you were and that's what pandemics are, we just want it to go away. That's why the magic bullet of a vaccine will, you know, it is something that people look to otherwise, it's this big collective action to try and like stop, limit the spread of it. Anyway. So in terms of the story around Trump, it's been a killer monster story in the sense of a horror movie story, because it keeps coming back, right. You know, this is one of the tropes of horror movies is the monster is never dead and you know, it, you're waiting for it to pop back up. Right. And so it's done that. It's done that. And more recently, of course, you know, he got COVID the narrative Ginsburg's death, and then, you know, the white house event and you know, that kind of bunch of stories there.


Peter (51m 28s):

And I will say one thing that isn't in our finished work is as this is going along, but we're watching these stories unfold. And if there are points and we haven't quantified this, you know, and this is sort of a great challenge, I think what is the uncertainty of now socially, right. And, and what is the incentive now? And I, I just, and this is just my gut feeling like when Ginsburg died, there was just all these sort of possibilities. And then when Trump got infected with COVID again, there was just this enormous range of possibilities. And I'm just saying what, you know, he could have died, or he could have, you know, not, not being sick at all. He kind of ended up with this almost resurrection kind of story, right? He's narrative there is, is a pretty good one for him.


Peter (52m 9s):

You know, he got him, he was definitely sick, but I mean, you know, he looked pretty unwell, but he recovers and he's, you know, he gets to portray strength and sort of to tie it all together. I mean, he's a storyteller, you know, that's my view of him. And he's a storyteller and this has been story versus the virus. This has been the game it's story versus the virus it's been pretty successful. And let me just tell you he's fundamental. We have this other work on the overall sort of emotional arcs within books within fictional works and what they're like, and this connects to Kurt Vonnegut, who said that you could do this, right? He thought you could do this. And he was upset that Michigan, I know was university of Chicago denied him the ability to do this as a thesis.


Peter (52m 49s):

He said, I, you know, I want to be able to measure the emotional locks of stories. So we did it, and it was kind of a, it's an homage to him really. But he had this one that he would talk about Cinderella and so on. He had the most basic one, which is, I need to describe it like this, you know, there is a sort of this wellbeing access a man and he portrayed it like this, you know, there's a guy he starts out okay and in time, goes this way and something bad happens and he gets out of it. And that's the story. And that's like, you know, like there are many stories like that. It gets back to normal and he called this the man and the whole story. And the thing is actually, that's not a good framing because it doesn't have any dynamic to it. You know, metamorphosis a deepening whole story, for example, but person in a whole, shall we say, there's no dynamic to it.


Peter (53m 30s):

It doesn't tell you where it goes, but make America great again, is a story in four words. It is really spectacular. It tells you about it indicates something about the past, the present and the future. And of course, it's not, you know, it's not who it's from. It's from Reagan and Bushes. That's as far as I know, it goes back to, from their campaigns in 1980, it was making, it was actually let's make America great again. So it was a little more collaborative, but you know, really powerful template and it's been reused, you know, bill Clinton used it and many other people have used it. So in terms of, you know, that's what I see. And I feel like, let me just add the turbulence part because I need you to say that. So there was a lot of turbulence in that sort of alluding to before in 2017, lots of scrambling, right. There was just so, and what that meant was there was no memory.


Peter (54m 13s):

Those are crazy things happening, and we're not talking about what happened the last month, because it's just fallen off the lip. You know, there's only so much you can handle, there's this new thing that's just erupted 2018, 2019, it's a little more stable and you start to get more stable things like impeachment, you know, it becomes more of a remembered thing and it's sticking, but you know, there are still things like Greenland trying to buy Greenland in August in 2019, right. That's sort of dominant story, which is pretty funny to reflect on, I think, but what you see in 2020 is a story that absolutely sticks and it's coronavirus, and that becomes the slowest turnover around Trump that we've ever seen until George Ford's murder and then there's this just really, cause that's the end of may, the month of June is, you know, just massive turmoil.


Peter (54m 60s):

You're broken from the previous story, but then both of those stories basically just persist. And so you see something like August, August starts to look like February, because coronavirus is the thing again, in terms of that nonlinearity of time, there's this strange thing where you can feel like it's like six months ago, last week can seem different. But now, because you know, right now, as we're recording this, the pandemic is really raging and there's a sense of panic buying feels like March again, right? That kind of lockdowns are kind of coming. So that time feels close again. And you know, maybe if you think back to some things you did in the summer, that actually may feel like years ago, but what happened then of course it was October, the start of that is Trump getting sick with COVID or infected with it in November, the rush and the of stories has just been enormous.


Peter (55m 49s):

And we quantify that, right? So we have, we have these kinds of measurements for doing this, but we try to actually really put a number on it. So let me just say that 14 days in April, which was so thinking about what happened two weeks ago, April's is really slow month because nothing seems to have changed. That feels like two days now. So November, like two days change in the forest, right? The forest is changing in two days in a way that it took 14 days to do back in April. But if you'd go to something like the summer, the summer, because if was starting to look like this early part of the infection, that's like six months in August starts to feel like 13 days. Now, if you think about how much turnover in the forest you had over six months back in August, that's happening in two weeks now, it's not linear.


Peter (56m 35s):

It's kind of a messy thing, but it's just, this is a powerful thing to look at. And then we have this crown apathic equivalency heat map, you know, we have these kinds of things and it just shows you, are we untethered to the past right now? Have we lost memory of what's going on? And you know, how much is sticking? And we'll see, I mean every month I kind of watch this thing for when I'm wonder where it will end up. And it's just been sort of, I was trying to say this before, just you just see the story in front of you just flapping in the wind. Like, you're not sure where it could go and what's the uncertainty of now. And right now, again, we're back in the deep uncertainty of these last couple of months.


Michael Garfied (57m 13s):

Well, this is, there's so much here so much, obviously it's worth bringing up Vladimir Elliott Lennon's quote, there are decades where nothing happens. And then there are weeks when decades happen. And, you know, when I had a SFI Miller scholar, Lawrence Gonzalez on the show to talk about trauma, this is what we're thinking about. You know, we're thinking about those moments where it's seared into your memory for the rest of your life, like the Chicksalube meteor impact, right? Or in general, you know, golden Eldridge talk about punctuated equilibrium. You know, that this notion that there's this sort of a, royaling micro evolutionary flux going on at all times, and then something happens and, and it changes everything.


Michael (57m 55s):

And then we move on and that's what we're seeing here. But I love that you described in this paper, the idea that a slow down in story turnover, a drop in story turbulence is when quote unquote and you've now created a rigorous quantitative way of describing this term, the plot thickens.


1 Peter Dobbs (58m 21s):

It's true. I almost put that in the title of the paper, but yeah,


Michael (58m 23s):

That's great. I mean, I'm fascinated about this because, you know, I remember years ago, reading Doug Rushkoff's book, present shock when everything happens now and he was talking about, you know, the way that the internet in a way that I came to understand after reading his work resembles the functional MRI scans that I think it was Imperial college, London did on the brain under the influence of LSD and showing how that you get this massive increase in functional connectivity between brain regions that are normally inhibiting signals between one another. And that that's like what Twitter is doing for the collective brain of humankind. And when that happens, you get this huge boost to the noise in the system, and it makes it difficult for things to, if we're going to use your forest metaphor, for things to take root, when everything is going on all at once, then an event that would have seemed epical and defining of a generation is just lost in that wash that you were talking about.


Michael (59m 28s):

And so part one is, you know, what are your thoughts on, again, as in terms of like memory dynamics, the way that information is held in society. When we see things from space in this way, in the same way that, you know, astronaut rusty Schweikert turns around, or Edward Mitchell turns around and says, at least in the sixties and seventies, there were no definitive visible from orbit country lines. Now there are because of development and deforestation, but you get to see it in a unity that you didn't have. And it becomes clear that there is a self similarity across scales in which the collective dynamics of the entire human race and our collective cognitive activity resemble really much this kind of dynamic at the individual level, but where that gets kind of weird and creepy, cause I promised you, I was going to get into the dark side.


Michael (1h 0m 21s):

Like the weaponization of this kind of a tool, is again, to call back to the paper on story Wrangler. You mentioned changes in word popularity, predicting future changes in geopolitical risk. Like you were just talking about here. You know, you define it here as a decline in real activity, lower stock returns, movements and capital flows away from emerging economies. Following the federal reserve, you mentioned being able to track the words crack down in protest associated with changes in geopolitical risk index. And again, when you think about the fine line between genius and madness, right, between the increase in functional connectivity, leading to a brilliant creative innovation versus a schizophrenia break in personality, we've talked on this show with Carl Bergstrom, Jevon West with Eric Olsen and Vicky Yang about polarization and divergent reality tunnels.


Michael (1h 1m 16s):

And you know, what happens when the information is coming in so quickly and it's being integrated, the systems that were in place to integrate information at the collective scale are no longer satisfactory in their function. They can't keep pace with these things. And so, I mean, I guess part of the, the ominous question here is just how insane do you think the emerging collective mentation of our Noah's sphere is right now? Like how crazy are we collectively? And then the other is, you know, once you know how to Gaslight someone, then you can do that.


Michael (1h 1m 57s):

And in a way, these tools, I'm sure. I mean, I know that you and your colleagues have thought about the ethical implications of being able to offer this kind of an insight to people. So what are your thoughts on the evolutionary arms, race in information asymmetry and access and the literacy around this kind of a tool and what it means in terms of the agents that are undoubtedly already using this information as leverage to divide people. Because I mean, even Gregory Bateson back in the forties, when he was part of the OSS, pioneering skidmagennis in the Korean war was thinking about this stuff, you know, I mean, it's refined now.


Michael (1h 2m 43s):

So this is not where I necessarily want to land this conversation, but just to take it out of sort of the starry-eyed wow, we can do this to the, Oh my God, that means that we have orbital lasers and we could theoretically take someone out. What do you think?


Peter (1h 2m 59s):

So a lot of things, yeah, no I've worried deeply, deeply about, and this is my interest in stories, like what is happening in terms of the stories that are watching for populations that are being manufactured to spread and so on, or are being found sort of in the wild and then amplified by, you know, certain agents or whatever, maybe a little too a little bit before, but I mean, the kinds of stories we have in the U S now are not great on there. It's very divided. You really do have quite divided populous. And where does that lead in time? I'm pretty limited in my view of predictability for a little bit, we do have some stuff that you showed, but I came up with a correlelesation the other day. I mean, it's sort of, it's a little bit along those lines of some predictability, but those numbers for geopolitical risk there, and they're pretty rough.


Peter (1h 3m 44s):

They're pretty broad. It's not going to say there's going to be a protest in that town. There's nowhere near that kind of thing. But even if it did, I think there's, this would be good stuff to have out in the open so people can understand that this is what is being known about you. I mean, potentially about your population you live in, we're going to eventually figure it out, I suppose, but there's going to be huge limits to it. You know, we know, we understand now for the weather, right? That's a sort of a two week horizon. We get as much data as we want, you know, computational power and all that sort of stuff. But we understand from dynamical systems, we just can't tell in two weeks, like in detail, that's gonna be a limit I'm really interested in where we'll get to with that for social phenomenon. But I think it's going to be much more limited. I mean, there so many shocks that come along, their just shocks and we then talk about them in history, but people don't anticipate certain things, you know, they really don't.


Peter (1h 4m 33s):

So from the sort of the whole sort of the story space that's emerged. I mean, how do you help people defend themselves against all these stories coming at them? And there are different approaches that have been kind of understood, You know, I think from certain countries and cultures, like one is flood the world with stories, you know, just put out an enormous number of stories that essentially are all competing with each other. So you can't figure out what's going on. It's like denial of service from a story point of view. And so that's a powerful thing. So if we had this sort of sensing thing that could tell you, here are the stories they're coming from this place they're being fed into the system, they appeared on this TV show, but here's where they nucleated from.


Peter (1h 5m 13s):

No good. Right? So it's absolutely an arms race. And I think of story was you have to have stories to get people to do all sorts of things. There's any number of quotes from history, from people who sort of figured this out, I suppose. There's diversion by telling people, you know, happy stories all the time, like the American dream, right? That's a good story. Rags to riches is one of the most basic stories, you know, in terms of fundamental stories, that's it rags the riches. I know it's just like, come here and you go up the train. So cultures tell stories about themselves, you know, and I don't know what the U.S. will tell about itself in, in years to come, but it's certainly dividing it stories now that they're kind of broken. I will say, I think  


**we're terrible at understand stories of collectives and there are a few reasons.  This is a systems problem, right?  This is why we struggle with systems, we're really good at individual narratives. Journalists will do this all the time. People writing books will do this all the time. They'll frame a major event around an individual so you can travel through in their shoes and get a sense for it and then maybe step out. Religious sermons do this as well, right with sort of parables and things. They tell you about individuals, we're individuals. So, but it's much harder to understand 300 million people like the collective behavior, you know, and this is why we've created all these simulations from the simple time model things to, you know, football and things. We don't have a good understanding for how people behave in large collectives. They can seem strange because it is a different thing, but we do this mistake of imprinting individuals on top, and many people will be sort of comfortable if one individual was running the whole thing, right? That manifested various ways as religions. You know, the idea that Trump is in charge or Obama was in charge. You know, there's a hand at the wheel, Leviathan, cause that fits into our limited cognitive capacity. So the stories of the many are outside of our mental frame.  What we sort of walk around with actually. We have to really train ourselves to understand that.  


So one other piece here to kind of go with that is we live in a really different age from the 17 hundreds when in the U S it takes months to get information from one side of the country to the other. And of course, it's only from one sort of few individuals to a few other individuals. That's not where we are now.


Peter (1h 7m 22s):

Now we have live video and people in all around the world principle can watch the same thing at the same time unfold, we have media that's spread out. We have some of that media has say big, like Fox or CBS or something. You know, the big things that many people will watch, but this local news that might be local, but it's also tethered together because it's owned by the same group. You know, USA today owns a lot of different newspapers stuff to say they're a good or a bad actor, but they're not. So there's a correlation that's come about in what people see in the world, which is very different. So I come from rural Australia from a farm and we're always sort of at odds with cities, right? There's always like, Oh, the cities get all the resources that no one cares about.


Peter (1h 8m 2s):

They don't know what sheep or food is. You know, this is sort of that kind of battle. So, and when I sort of came to U.S. and understood this effort with the electoral college to distribute power and to put capitals in small places and so on. There's part of me that thought I can see that right. From a systems point of view, it's clever, right? It's like spreading things out a little bit and allows for innovation. John Stewart years ago called it the meth labs of democracy. But, you know, so can work in different ways, but you need them to be acting independently. So, but if these different States in the U S are acting in a correlated way in terms of the information that they consume, then you've got a really different game.


Peter (1h 8m 44s):

And that's really dangerous, right? The correlation of people's intakes, even though we're in this world where there is an infinite amount of information, and there's all this stuff I talk about science is going from data scarce, the data rich, right. This is a transition that if it's really a science, we'll make, you know, we talked about astrophysics early on, right. So looking with your eye, trying to figure out, you know, when people figured out amazing things in the last couple of thousand years, right. But it took us a long time to get the whole ellipse stuff sorted out, you know, and then telescopes, and then John Res you know, data scarce to data rich. So that transition, I think, you would think there's sort of this naive idea that I think a lot of people have was like, Oh, you know, we'll be able to sort of sort everything out, but of course, what emerges on top of that, you have to have these managers of information because it becomes incredibly hard, right, now you're looking for a needle in a stack of needles or one piece of hay in an enormous amount of hay, right?


Peter (1h 9m 41s):

Like it's just really hard. So you go to authority figures or entertainment or whatever and you can see from that. That starts to lead to many people being correlated. Hundreds of years ago, there's no way information transpose across countries, it just doesn't. So people are making their own judgements that are, it's really the Telegraph, I think, which is kind of a switch in terms of suddenly information going across at a distance. But I, you know, I see that as a grave problem, but I guess I have this hope, if we can kind of show it, look here is the tapestry of stories that are kind of floating around in the world and here's, what's being consumed.


Peter (1h 10m 22s):

Here's, what's being spread. Here's an origin, you know, that that could help. And I think it's really important as well, just from a public policy point of view, what what's going on, what is being discussed, what is happening? And you've tried to do a good thing. We've tried to Institute, you know, wearing a mask. We tried to Institute people doing certain things, you know, how are they talking about this kind of thing? How do I talk about this city now? Is this a better place? You know, so there's a good side. And of course the terrible side is something we have to explore. And of course think about very deeply because there are many ways for things to go wrong and people are very clever. They'll find fun ways that you don't anticipate. So we're very mindful of that, but, you know in terms of the control of people and that sort of thing, which is, you know, awful, you know, it's underway, it's happening just the rags to riches stories in your American dream.


Peter (1h 11m 7s):

That is, you know, it can be pretty soft. It's not like there's some, you know, person, you know, creating, you know, whatever it is, you know, sort of inserting this information to every, every person. But, you know, there is these stories that we're bound to that are for better or worse things that we, that, that got out lives. So I do kind of at times, think stories are everything right. That why we get out of bed when you kind of draw down, a lot of them are about, and I don't want to be too, like, about survival in like kill the monster, right? That's better survival rags to riches. That's flourishing, right? This is the things are great. You know, that that's an arc. We want to see, we're very interested in the ox where things go wrong as well, because they're, you know, they're the cautionary tales, right.


Peter (1h 11m 50s):

But we are very preoccupied with survival and I, it gets very elaborate and we get a long way away from like the sort of roar idea of survival. But I have this way of talking about what many stories of sort of revolvers hatchings matchings and dispatching, right? These are the three main events. And so we're, you know, there's some aspect of that that's spreading around and you think of the stories that are dominating the U S now say, you know what Trump talks about, For example, a lot of them are pretty dire, right? There are some times where like the country will end many different sort of groups will sort of frame it that way. Like it's going to be the, the end. So there's that sort of, that kind of frame, which is very different to the time where things kind of building and we're kind of creating something fantastic.


Peter (1h 12m 33s):

And we're flourishing. So, yeah.


Michael (1h 12m 36s):

Thoughts on that, you know, since one of the things that you point to in this paper at the end on, like, what could we be doing better? What could we do? You know, what, where is the research to expand is gesturing at what you're talking about, which is the story taxonomy in press. If it bleeds, it leads sounds a whole lot like, well, you know, the killer whale is the Keystone species of the ecosystem and, you know, work by professor John Hart and others, you know, I've talked about the way that ecosystems are organized according to, they seem to follow the principle of maximum entropy production, right? Like a river basin, it's like the network is going to get as fractal as possible in order to minimize the turbulence at its branch points.


Peter  (1h 13m 24s):

I did my PhD on river networks, by the way.  


Michael (1h 13m 29s):

Yeah, so this, you know, again, to, you know, shout out, like we seek to, at every episode to the conversation we had with Jeff West on this particular topic. And, you know, in a way it makes it seem like it would be a danger. What I hear is for there to be too much correlation, for there to be, you know, monocropping


(1h 13m 51s):

The diversity of perspectives, we want a diversity of stories. And then the question is, how do we actually study this rigorously? So, you know, talking about you being from Australia, I can't help but think of the story that Aldous Huxley opens his essay, heaven and hell. Talking about how when the first European naturalists came back from Australia, and nobody believed them when they came back with a specimen of like a Platypus and they all, they had to rip it apart, autopsy it, to see if they could find riches.


Peter (1h 14m 19s):

So many of the others had done that. They'd made a few, you know, kind of a bad kind of a bad history there. Yeah. I'm a huge fan of the total champion animal, but then you have for predictability, you know, let's stop with a bunch of quarks, right? Hey, you've got a super corks. You're going to get a Platypus out of that.


Michael (1h 14m 42s):

Yeah. I think that actually, there's a Platypus in the taxonomy of the cartoon animals of the Vermont complex system center, as I recall.


Michael (1h 14m 52s):

Yeah. So actually suggest in that riff of his that what the world needs is a new cadre of naturalists to perform the same duty of creating a taxonomy for the contents of the human mind. And it seems like this is really what you're getting at here. And that one of the things that stands in your way that you and your coauthors address in the story Wrangler paper is the data that you do not have, which as you alluded to earlier in this conversation is like, somebody's Twitter accounts are corporate entities. Many of them are bots. They're not even human beings.


Michael (1h 15m 34s):

And so, you know, to get a read on the Noahphere from this is skewed, you know, like you've got a blind spot. And then also with geographical data, like presumably much of this data is available. Maybe Twitter just doesn't want to give it to you.


Peter(1h 15m 52s):

We actually dialed that down, you know, used to be more, I mean, not great, but it used to be there and you could do something with it, but it's actually been doubted, which I look, I think that's to their credit, to their credit. Like researchers might be like upset, but no, I mean, they're making a service and I think that's actually smart of them to do that. So you're right. So geo is a little less is hazy and we've stepped away from, we've gone for language. So there's a hundred languages.


Michael (1h 16m 18s):

Right? So that's not exactly a perfect fit because it's kind of two different questions, but like one is about the lacunae in your data and what kind of questions that bags. And then the other is how you're starting to tease apart this panoply of fantastical things that you're discovering and what a taxonomy of story is starting to look like in a rigorous or can you even start that?


Peter (1h 16m 49s):

No taxonomy of story is, is this, I don't know. I see that as a grand, a grand achievement of science, and look, it would vary from culture to culture and over time, but just being able to do that well, there've been efforts for a long time and there was an early efforts with them folk tales, which are, you know, and so that more bespoke by hand, you know, magical animals and the magical trees, you know, there's sort of this list of things, but trying to do that in a data driven way, I think is this great, great challenge for us? So it's kind of coming, but I would say, you know, yes, Twitter is this limited thing also it's enormous. So we're sort of, you know, it takes a lot of work to contend with it, but we've really tried to look at other different corporates, sort of see, you know, spread out.


Peter (1h 17m 32s):

So Reddit is another one, books are another one, of course there's video and pictures, or there's just enormous other array. It is painstaking work. You have to be incredibly careful with this. Know what I mean? Maybe we've messed things up, but we've tried so hard to count everything and be really careful with this because yeah, it's tough. So I think that the errors are not always really in the analysis they're just in the corporate themselves, that you've chosen. Of course, I mean just to sort of say that ups quite obviously but then they are also in your instruments as well. So the hedonometer for example, the way that works is we have a list of words that have scores attached to them on the happiness sadness spectrum, and what changes things is not really the scores, you know, the scores could be moved around a little bit.


Peter (1h 18m 24s):

It's just which words are in the list. It's kind of like, are you looking through a red filter or a green filter? You know, if you're looking at colors or is it a bad lens, you know, that changes the story. But I will say one of the things that's helped me particularly kind of move along with this is when we've thought to improve something by getting more data and improving an instrument in what feels like a way that you might improve a physical instrument. Right? So we, we made telescopes, we made better lenses, you know, and it took a lot of time. And then 10 years later we make a better one with more, you know, the curvature is, you know, more accurate and so on.


Peter (1h 19m 4s):

We've been able to do that with language. And so that's what we've done say with the hedonometer, a larger list of words that fits natural language better and having scores for it. It didn't give us a completely different picture, right. So there was this concern that I had early on that we would change the instrument, like try to improve it and everything we saw before would go away, you know, there'll be all kinds of wrong, but actually we could still see that, but we started to see much more definition. So it had that kind of improvement, which is, Oh, now we can see more finely in this photograph in front of us. And so that was a good experience.


Peter (1h 19m 44s):

Right. It makes me feel like some of what we're doing is a bit more like physical instruments, that they were charitable parameters associated with them, which is kind of like, you know, having a focal thing on the side anyway.


Peter (1h 20m 3s):

How do we tell these stories.  Theres a whole bunch of conspiracy stuff about flat earth or, you know, Q Anon kind of things like how do we start to kind of tease them apart and, and see where they started? You know, time is this enormous thing for us, right? It's the big, the big sorting index. So just seeing when a word or a term comes into being is pretty powerful so even if we're using Twitter with something else, that's going to be this pretty good background information source. So Q Anon, for example, when does that emerge? And like hashtag you know, where we go one we go all and that sort of thing.


Peter (1h 20m 44s):

Like you can just see what pops up. So then you could go to other corporate as well, or media sources and kind of try to analyze things there.  One of the stories that people tell, do people give up on stories at some point, you know, there's the rags to riches stories we've talked about for the American dream, you know, that can be broken, you know, things go wrong for you and maybe that breaks, maybe you move to a different story. Maybe you become more tied to a religious group that's more community oriented and helping and so on. I mean, how do people move around and kind of the stories of their lives and their communities, you know, the stories just about themselves versus everyone else.


Peter (1h 21m 28s):

And you know, there are the, of course, the stories that we like to read, you know, there's that whole world as well, you know, which ones animators, you know, and there's marketing on top of that. There are there sort of just standard stories that people will come back for. Right. I mean the Avengers or something like that.


Michael (1h 21m 49s):

Thinking of Simon today, I would just coauthored that piece from probability to consilience. You know, the base therom. And like you said, there's the different kinds of simplicity that we seek. So it's that sort of speaks to your, you know, which stories are persistent or the stories that, you know, like insectivores survive a mass extinction. Anyway.


Peter (1h 22m 19s):

Yeah. Which stories do you miss completely? You know, we're pretty, we're kind of funny when it comes to probabilistic stuff and randomness. Right. Sometimes we're good with it, but often it's completely bemusing to us or we miss it. Or we see some sort of random behavior and we tell a story about it. And I think sometimes that's a pretty good survival trait of sorts because, you know, there's this old sort of notion of it's good to make the mistake of seeing a rock or some sort of twig in the woods as a tiger. Right. You know, cause we think that thing has agency, but I would sort of, I would reframe that as saying what we see as a story, right?


Peter (1h 23m 3s):

We mistake this and because stories are dynamic things, right. They're there there's time and something happens and then they can be just a little plots and they told them different ways. So this is the plot, which is the mechanic. And the story is the telling of it, you know, Proverbs are stories, right. The algorithms are stories, they, lots of things can be seen in that way. But yeah, we see this tiger is going to come out and eat us and we imagine that, right. We run that little scenario. We run that little story. If we see a rock, we don't run from it.  You can talk about agency and all those things, but what we're running all this all the time are these little possible paths and which one do we want to be on?


Peter (1h 23m 47s):

And we're trying to guess that about other people, what paths are they taking? What paths is your community taking, what paths are the world taking?  



I feel like we could continue to delve into this and I would love to for a while, but I think that we've already kind of consigned ourselves to this being a two-parter. So I don't know, Peter, this has been truly fascinating. I wonder whether this is now seared into my memory or whether I'm going to just be, you know, washed aside by the next sort of piece of sensational news.


Peter (1h 24m 28s):

But I suspect that given all the editing, this will require, I'll be thinking about this conversation for some time. Do you, do you have any parting thoughts for people? You know, I mean, we've covered so much ground here, but are there trails through this wood of connected ideas that we have not traversed that you find worth identifying, pointing to people as they walk out of this conversation perhaps, or where would you direct people if they want to think about these things more deeply beyond of course your work. The dashboards and the papers we'll link to in the show notes.


Peter  (1h 25m 11s):

Yeah. It's interesting you bring up the path through the woods. Phillip Pullman has that when he talks about sort of a notion that he, actually I had this book, it was just on writing and stories and I was kind of delighted to hear him talk about that. Right. But his view of that was you have to find that path through as a storyteller. Right. And I have my own separate stuff there and I think about adjacent stories in adjacent narratives. Right. And this is the problem I think for stories and for science in general. So let me just say this, this is I think an important thing, because we haven't really talked about the truth.


Peter (1h 25m 52s):

Right. You know, I don't think there are alternate worlds at all. There's what happened and there are true things, that's why the lights stay on. Right. And all this sort of stuff. Right. So, but the truth has been de-legitimized in so many ways. So somehow we have to remore ourselves collectively to the truth. So that's the great challenge, I think, now it's an ongoing challenge forever. And I will say, I think the problem we have as scientists and as journalists, my wife's a journalist and I think of journalists as scientists with a deadline and scientist are journalist without one, trying to tell what happened and try to explain it.


Peter (1h 26m 39s):

Right. So there is what happened and adjacent to that will be a better story. It will be a story that might be better just simply to tell, right. You know, there's everything that happened in this event. Even if you want to be a good journalist, you've got, you know, you're gonna write three paragraphs, so you have to distill it and break it down. Or even if you have video, you got to break it down so that it's already going to be removed a little bit from exactly what happened. But then if you're a bad actor, you can work to find a story adjacent to this one that serves your posts that fits your overall narrative. Or you could find many and start to put them out. And just so they seeds out.


Peter (1h 27m 19s):

So I think we have to protect the truth. We have to keep developing the truth, like reinforcing it, protecting it. This is the great challenge. I think the really hard problem for us is that there are beautiful stories adjacent to true stories that spread better, that people will believe much more easily because they didn't have the weird random thing that happened in the real story. You know, real stories can have, you know, stuff that we'd want to iron out. So that is a challenge for us. So not to fall too far into beautiful stories. True stories, they're gonna have some messiness in it. That stuff


Peter (1h 27m 59s):

is hard to explain perhaps, but we need to figure out how we can, you know, this is what science does, but we've lost it culturally. We need to get that back or reclaim it for the first time ever.


Michael (1h 28m 13s):

Yeah. I think about with Sabine Hoston Felder gave that talk. The SFI community lectures about Does Beauty Lead Mathematics Astray. And again, the thing about this is when our, our desire for simplicity goes from being a virtue to a vice where we pursue the beautiful story at the cost of truth.


Peter (1h 28m 32s):

So truth and beauty, not connected. I do not presume them to be connected at all. The truth can be told beautifully and that's our challenge, right? We have to tell scientific truth beautifully so that they can compete, you know, and its still the truth. But we have to tell them in beautiful ways, ways that people want to spread them so that they can compete with, you know, these mouse stories, these wrong stories.


Michael (1h 28m 58s):

Well, that's an awesome place to end it. I feel very validated in my science communication career by this conversation. Thank you, Peter it's was just awesome to talk to you. I love these, these wandering discursive ambles through the landscape of the mind, you are an excellent guide through that landscape and I hope that people found this as interesting as I did and that they go, and they twiddle around with the dashboards that you and your team have created, because it is immensely curious to turn the telescope back down upon the human species and to gawk at what strange creatures we are and what we look like in aggregate.


Michael (1h 29m 47s):

Thank you so much for being here today.  



Absolutely wonderful. I enjoyed it greatly. So thank you so much.


Michael (1h 29m 57s):

Thank you for listening. Complexities produced by the Santa Fe Institute, a nonprofit hub for complex systems science located in the high desert of New Mexico. For more information, including transcripts, research links and educational resources, or to support our science and communication efforts, Visit