Skip main navigation

£199.99 £139.99 for one year of Unlimited learning. Offer ends on 28 February 2023 at 23:59 (UTC). T&Cs apply

Find out more

Social networks in conflict

Neil Johnson describes the dynamics of protests and explains how data from Facebook can be used to anticipate protest activity.
So Suzy mentioned– let me talk about the first– Suzy mentioned that we were on this grant. We went to Malibu, et cetera. The point of the grant was the following– if you knew everything that was on the internet, and you knew about past events, et cetera, et cetera, could you, in my case, or our particular case, could you predict future civil unrest events? Could you make a prediction that in seven days, people on the streets of Brazil are going to be out in Sao Paulo– they’re going to be out walking around because you sort of saw it coming? You saw it coming from the information.
So we were tasked with that area in this grant. Every day we had to send off warnings of future events to the city level and also to the issue level. It had to be whether it was labour or it was housing-related, or education-related. So these are very specific warnings, sent in every day to the government. Government scored them, it came back whether you were right or wrong. And I have to say, when we first went into this– I mean, you think about it. Civil unrest– you hear nothing, nothing, nothing, nothing, nothing, nothing on the news. Sao Paulo– suddenly Sao Paulo, there were events, events, events, and then there’s nothing.
Then it’s on to the turtle that escaped from somebody’s garden and the news has kind of moved on. So it comes in bursts, these events come in bursts. So we thought like, I think, probably a lot of you think, maybe a lot of people in the literature were thinking– this is a contagion problem. So we busily went about building a contagion model. And something Suzy didn’t mention– there were three teams. We were competing. So the other teams went off and did what they were doing. They also built contagion-like models. And it turns out that civil unrest isn’t a contagion process. And I want to show you and share with you, what it actually turns out to be.
Or what we think it turns out to be.
I mentioned the bursts. So you’ve got nothing, nothing, nothing, nothing, nothing. There’s just time going in this direction. Nothing of anything– bursts, bursts of activity. This bit is like a contagion process. I’m not going to bother going through the detail of the model we came up, but suffice to say, that a contagion model modified can explain the types of shapes of bursts that you get. In fact, we found a kind of universality in the shape. The bursts tend to be longer than in a normal kind of contagion process. They tend to have a longer tail. They can have recurrent peaks, and they also tend to have a very delayed onset.
But, you know, more or less, you look across countries, you look across cities, all of Latin America we had to do– they classify, they fall down in these three classes, these three shapes that I’ve shown at the bottom. A very simple model that reproduces these three shapes is one where people, as individuals, are accessing some common space which may be me switching on my phone and looking at some chat room. And in that time I have my phone and I’m looking, I’m susceptible to the information from it. And then I maybe, oh, crikey, there’s something going. They’re three streets away. I’m going to go and get involved. I turn my phone off. I go over there.
Suddenly I’m out of that common space. And if you have people going in and out of this common space, it can give these three shapes. But that’s not what I want to talk to you about. Because I want to talk about, what about that dark part? What about before the burst? What is going on there? And the story of that turns out to be– it’s a good job we’re in the business school because it turns out to be something to do with innovation and also organisations.
What we found is the following. This is a complicated chart. It’s meant to be complicated. On the bottom you see the horizontal line going to the right is the time series of civil unrest events. Now, three teams tried for two years to look at time series analysis to see if this burst that you see on the bottom right, with that triangle, where that burst was coming. We tried every type of time series analysis we could think of, looking at previous events to try and predict the future. And for everything we and the other teams could do, this burst came out of nowhere. Just came out of nowhere.
Until we started to think about, well, it’s going to involve some kind of organisation. And that was the key. So first of all, we were looking at contagion, we were looking at tweets. Tweets had nothing in them. They had something on the day, and two-day, and three-day time scale. But they didn’t have this kind of month to year timescale in them. It turns out that Facebook had the answer. And it isn’t Facebook individual accounts. As all you know, you’ve got Facebook accounts– you can set up a page on Facebook. And when you set up a page, you have the choice of organisations and causes.
I mean, you can put all sorts of other things, you know, bakery this, or– but you can have organisations and causes. And what we found is that people were creating these pages. Even though there may be one or two of them creating them, and people would collect and organise around these pages. These pages might come and go. They may have many likes and then people stop liking them. But it’s that organisational development that we found to be the long-term buildup, then to the tipping point of this burst.
So on the top of this plot, in tiny, tiny print, on the left is the list of organisations whose pages appeared during a two-year period prior to the major burst that was a year ago in Brazil. I’m just showing you Brazil. We had to do all the other countries as well. It’s a similar story for the other countries. So you have a list of the organisations that appear. The red dot is where they appeared in time, their first activity. And what you begin to see, even just visually, is that there’s some kind of development as you get towards the right-hand side. And the right-hand side where the burst appears then suddenly there’s a large number of organisations.
So what we’re seeing is not a contagion process involving individuals just passing things on. Its people, maybe even a common pool of people, aggregating around certain causes and certain pages, and then disappearing back and then forming again. It’s like the typical kind of fission/ fusion, social dynamics process, but now on the internet. And now behind these, in some sense, nameless organisations. So we found– in a minute I’ll tell you what the order parameter is that we found for the transition. But let me give you a clue just so that we understand what’s going on with this transition. It’s not like a contagion process. A contagion would be, you know, I’ve got a bunch of sticks here.
If this was a contagion process, the more I shake it, the more they each shake each other and everybody transfers energy to everybody and suddenly all these sticks are flying out. That would be a contagion process. But what you can see is if you steadily drive a system– you drive it in a reasonable way, not too fast, not too slow, what happens is very different. See, you never thought you’d see an experiment here, but we’re going to do it. You know, it’s computational, but in the end, everything’s experimental, right? So let’s see what happens. These are, if you like, these could be memes, these could be ideas, these could be pages, around which interest is waxing and waning.
It’s coming and going. So you’ll see sticks come up, sticks come down. Let’s see what comes out at the end. Does one issue come out? Do 20 issues come out? Well it turns out that only one ever comes out. In fact, I don’t know if anyone knows what these are. These are actually Buddhist prayer sticks. There’s a whole reason that only one comes out. That’s why they exist, because then you take it off to one side. There’s a few Chinese characters on here and you get them interpreted by an interpreter on the side in the temple when you go in with the sticks. So one only ever comes out.
It’s not like I’ve done it once and I got lucky this morning. So that’s the type of transition we’re seeing. We’re seeing this transition where one issue is emerging from this sea of issues, not through contagion. Contagion would be a mess– entropy increases. This is more of a kind of ordering of the issue and outcomes. The issue outcomes– the day, bang, up comes the burst.

How do protests evolve? Can we anticipate a burst of protests?

In this talk, Neil describes the dynamics of protests, and explains in more detail how data from Facebook can be used to anticipate protest activity.

Neil Johnson directs the Complex Systems Initiative at the University of Miami, Florida, where he is also Professor of Physics. He is the author of two books and a previous presenter of the Royal Institution lectures for the BBC.

You can watch the whole of Neil’s presentation “The Dark Side: Dynamics of Clandestine Social Networks” on YouTube (33:27).

This article is from the free online

Big Data: Measuring And Predicting Human Behaviour

Created by
FutureLearn - Learning For Life

Our purpose is to transform access to education.

We offer a diverse selection of courses from leading universities and cultural institutions from around the world. These are delivered one step at a time, and are accessible on mobile, tablet and desktop, so you can fit learning around your life.

We believe learning should be an enjoyable, social experience, so our courses offer the opportunity to discuss what you’re learning with others as you go, helping you make fresh discoveries and form new ideas.
You can unlock new opportunities with unlimited access to hundreds of online short courses for a year by subscribing to our Unlimited package. Build your knowledge with top universities and organisations.

Learn more about how FutureLearn is transforming access to education