Ah, thank you very much. Perfect. Okay, so thank you very much. I'm Hannah Frye. I'm a mathematician from London.

And the title of my talk today is I Predict a Riot. But I think I was perhaps being a tiny bit misleading when I chose this title because perhaps a more accurate title that I could go with is Can I Predict a Riot? And actually, while we're thinking about it, I'm not really the only one involved here. So we could refine that further by saying,

can anyone predict a riot? And you know what? In for a penny, in for a pound. Let's actually explore, can anyone actually predict anything at all to do with human behavior? Now, I think this is kind of an important and interesting question because surely all of you are very smart, intelligent,

independent people acting under your own free will. So how is it possible to predict anything about your behavior at all? Now, I'm going to come back to that and I'm going to come back to riots in a little while. But first off, I thought I'd give you the opportunity to win a jar of sweets. So this jar of sweets,

all you have to do to win the jar of sweets is I want you to tell me how many sweets you think are in the jar. Any guesses from 16? Very low, I'd say, on the low side. Now, I should say, actually, because you're all independent and free willed people acting under your own free will, rather,

and I want you to make up your own mind. Don't let anybody else influence you. Okay, anyone else like to make a guess how many sweets you're going to throw? Yes, please. 83, okay, yeah, go ahead. 150, okay, and 60, okay. 160, okay. Okay. 99, we've got there.

Go ahead. 92, okay. How many? 115, okay. The actual answer to how many sweets are in this jar is 117. So you, sir, you win yourself a jar of sweets. I'd like to come and get you a jar of sweets. Round of applause if we could. Thank you very much.

Oh, there you go. A mean trick, I'm sorry. Okay, now, why did I do this? Well, so I also, that jar of sweets, 117, this is a picture of another jar of sweets, sadly didn't make it to Berlin, and I ate them all before we got here,

but this other jar of sweets, I put this picture up on Twitter, and I asked people on Twitter to do exactly the same thing as you've just done here, make a guess as to how many sweets are in the jar, and here are the answers that they gave me. I've got exactly the same number of sweets in that jar of 117, and you can see that there's a range of answers just as we had here in the room,

everything from 50 plus one paper clip, and I can guarantee there were no paper clips in that jar, all the way up to a frankly stupid 350. But as I said, there were 117 sweets in the jar, but something really interesting happens when you take the mean, when you take the average of all of the guesses of people on Twitter,

and it comes up with, the average of the answer is 117.9, which I think we can all agree is astonishingly accurate. Now, if we'd done that in the room and asked you all to write down your answers, we would have found a similar phenomenon. You would have got very, very close to guessing the number of sweets in the jar. Now, this idea was first discovered

by a chap called Francis Galton, and he noticed, he was at a fairground, and he was watching people guess the weight of an ox, and he noticed that the number of people who overestimate tends to balance out the number of people who underestimate, giving you an average that's astonishingly accurate, and it's exactly the same idea with a jar of sweets.

And I wanted to include this because I think it's a very simple example of how you can predict some types of human behavior. I'd have no chance at all of making a prediction of what any one individual would make, what they would guess for the number of sweets in the jar. If you look at you all together, then suddenly patterns start to emerge

that are much easier to get a grip on. Okay, so in the interest then of working all together, I noticed there when you gave our lovely winner a round of applause, you're a bit raucous, to be honest, and you're all clapping at different volumes, different pitches. I mean, it didn't even occur to you to clap in sync, which is a bit rude, quite frankly.

So in the interests of acting together and seeing what human behavior looks like as a group, I want to see, just a tiny little experiment, if you can get it together to clap in sync for me. Okay, go. Fabulous.

Fantastic, thank you. Okay, now I did actually make a prediction there of what would happen, and you are the first audience ever who has not done it, so thanks a lot for that. Quite mad, it's actually true. Because what audiences normally do, and a couple of you started it, what audiences normally do, but not you,

is synchronize very quickly in under two seconds, which you did manage, but then they normally increase in tempo and build to a raucous end. So thanks for ruining that part of the talk for me, I'll probably be, thanks a lot. But here's another example then, of how human behavior at a larger scale has characteristics which is not immediately obvious when you look at the level of an individual.

Because if you think about how is it possible that a room full of people can fall into synchronicity in under two seconds, it's actually quite, I think anyway, an interesting question. Because it can't possibly be that you're only listening to your neighbor and changing your clap rate according to what your neighbor's doing. Because if that was the case, the clap would be much more like a Mexican wave throughout the room.

But equally, it can't be possible that each one of you is listening to what every other single person is doing and adjusting your clap rate to everyone in the room, it's just not possible. There has to be something strange going on that a group of people can act in a particular way that's not immediately obvious from how they act as individuals. Now, you might think that this is because you're humans

and therefore incredibly intelligent and also very good looking and so on. But actually, this appears in the natural world as well. So fireflies, for example, have been known to do this, to synchronize in groups. And this is an example that was studied by a mathematician called Stephen Strogatz,

whose work is amazing, by the way, or check it out. Fireflies, very gorgeous. This is a picture of them in a wood in North America. Slightly less attractive close up, it's gotta be said. But fireflies have been known in Asia to synchronize their flashing. So they flash to attract a mate, essentially.

But this, along a riverbank, this is along a riverbank in Malaysia, this video, and you can see that all of the fireflies are doing this exact same synchronized behavior. Now, if you take a group of fireflies and you put them in a darkened room and leave them on their own, then they'll flash much in the same way as people clap. They start off flashing indiscriminately,

but they very, very quickly fall into synchronization with one another. So whatever's going on with humans is also going on with fireflies. You also see this with crickets, if you can imagine taking a lovely stroll in the countryside. Crickets chirping to attract a mate also very quickly fall into synchronization

with one another and chirp simultaneously. Now, maybe it's something to do then with the way that living things' brains work. Maybe they're capable of doing this kind of strange thing that's going on, the difference between the individual and the group behavior. But it may surprise you to know

that actually even inanimate objects have these properties. And so here is an example, I hope, yeah. Okay, so if you take a series of metronomes like this and set them off, all at different ticking rates, now what this guy's gonna do in a second is they're all sitting on a board of wood.

He's gonna pick up the board of wood and place them on top of two Coke cans, as you can see here. And what happens very, very quickly is that just the physics of the setup as it is allows each metronome to influence every other metronome

and be influenced by every other metronome. And you can see very quickly that they become in sync. This one here is being a bit naughty, but it does fall back in sync in a second. There we go. And I think this is astonishing that even inanimate objects can display group-like properties, which is different from how you would expect an individual metronome to act.

Now just before you start, you know, get carried away with thinking that metronomes or inanimate objects, humans are much cleverer than inanimate objects, there's a very similar example of this, exactly this behavior, exactly this property, that happened in London over the millennium. So let me just come out.

Okay. So in London, there's a bridge, the Millennium Bridge built surprisingly for the millennium. And within a few days, the bridge had to be closed because it was wobbling so dramatically that people couldn't cross it safely. Now, the thing is, is that there was nothing

really wrong with the bridge, it'd been tested and so on, and people could walk over it without it wobbling. But what happened on the bridge was very similar to the metronomes in that the humans, the behavior of the humans was what was causing it to wobble. Okay, so if you can imagine that you have a group of people walking in the same direction,

there is some probability that two people will be planting their right foot at exactly the right moment. Now, because the shock absorbers on the bridge, they weren't quite right, the shock absorbers on the bridge, what that means is the bridge moved ever so slightly, which would knock more people onto that same path

and more people and more people so that eventually, as you can see here, the entire crowd of people are moving backwards and forwards perfectly in sync with one another. And they're the thing that's driving the bridge and driving the wobble of the bridge. Now, I wanted to show you all of these examples, these early examples about synchronization,

because they're really, I think, beautiful examples of how the behavior of a group can differ from the behavior of an individual. And they're all quite simple. I mean, it's all about the interaction between the individual and the whole. Now, the study of these types of problems and these types of systems

is what's known as complexity, the study of complexity. And that's really my area of research. And as a mathematician, that's what I spend my time looking into. Now, complexity is something that has completely exploded in the past few years. And I'll show you a little later on, I should say, how I've applied complexity to look at riots. But I just wanna tell you a little bit more

about what complexity is first, if I may. Now, even though complexity has gone kind of crazy in the last decade or so, academics still don't have a completely agreed upon definition of what it means for something to be complex. But my favorite one, or the one that I think captures it the best

is by a mathematician called Mark Newman, an English mathematician who lives in America. And he says, a complex system is a system composed of many interacting parts, which displays collective behavior that does not follow trivially from the behavior of individuals. It's a bit wordy, but that's the best we've got so far of actually defining what it means to be complex.

Now, if you want something a bit less wordy, you could alternatively use a definition which one of my colleagues found in a paper. I don't know if you can quite read this footnote here, but it says complexity is a bit like pornography, hard to define, but you know it when you see it. So, a nice alternative.

Okay, so the definition aside, you can get quite close to understanding what it is and what complexity means by thinking of it in comparison to other bits of science. And although, as I said,

complexity has just exploded in the last few years, the ideas, these ideas have actually been around for a while. And there's a paper by a chap called Warren Weaver, an American scientist, which he wrote in the 40s actually in 1948. And I think he absolutely nails what complexity is and what it's about.

And in this paper, what he does is he does a kind of a review of all science, every scientific discovery, and every scientific application and technique up until that point in history. And he puts them, puts everything into three neat categories. In the first category, which he calls problems of simplicity, he gives the classic example of a snooker ball

or a billiard ball on a table. Now, if you can imagine a physicist or a mathematician observing that system, it would be very easy for them to write down a series of very simple equations which could track the movement of the ball on the table, how it rebounds from the sides,

how it interacts with other balls, and so on. If you, slightly less simply, if you scale those balls up to the size of planets, the problem is still the same, right? You can write very simple equations which describe what's going on very neatly. So problems of simplicity are essentially very few objects interacting with each other in a very simple way.

Now, planets is a good example because as soon as you get rather than two planets interacting with each other, if you have three or 10 or 15, suddenly the problem becomes completely unmanageable, and the traditional techniques of problems of simplicity no longer work. But strangely, if instead of just 10 or 15 items

or objects that are interacting, you suddenly have millions or billions of them, the problem strangely then becomes a lot simpler because you no longer care about trying to track an individual object, but you can start talking about properties of the system as a whole. Fluid dynamics is a really good example of this because essentially what you're doing

is you're tracking the properties of a system of billions of particles, right? And when they behave altogether, they behave in a way that is easily quantifiable, easy to understand, and that you can make predictions from. And this is the second group that Rowan Weaver was talking about, problems of disorganized complexity.

And it's disorganized because the particles, for example, in fluid dynamics are moving around in a completely random way, in an erratic way. So it's just sort of all disorganized. But at this end of the spectrum, at this point in history, statistical physics have been discovered, fluid dynamics have been discovered, and you could deal with problems of this,

problems of disorganized complexity where there are billions of objects that are still interacting in a very simple way. Now, Rowan Weaver made the point that at that point in history, scientific methodology had completely gone from one extreme to the other, from a very few number of variables to an astronomical number, and left untouched a great middle region.

Now, this middle region is exactly where all of these problems of complexity lie. And it's also where any problem that you can think of to do with human behavior and understanding and predicting human behavior, this is exactly where all of these problems sit in what Weaver calls problems of disorganized complexity.

And that's, I think, in a shortened version, the very best way that I can describe to you what complexity is. Now, in terms of looking at systems from this perspective, looking at the kind of disconnection between the individual and the whole, there are a few examples where people have made a lot of headway and be able to understand a system differently.

And one of those is in traffic shock waves. Okay, so imagine you're driving down the motorway, and sometimes you find yourself in a traffic jam, which lasts for a short while and then disappears,

and there was never any obstruction. There was never any reason for the traffic to slow down. So to explore this idea, a group of Japanese scientists, got a group of cars to drive around in a circle, as you can see here, just to try and explore this very idea.

Now, what they did is they got all of the cars, all of the drivers, they told them to drive exactly 30 miles an hour. But because there are people driving these cars rather than robots, some people would naturally be driving at, say, 29 miles an hour, while other people would be driving at 31 miles an hour. So what that means is that the faster cars will catch up to the cars in front

and have to put on their brakes, meaning the car behind will have to put on their brakes and the car behind and the car behind and the car behind. What you end up with is these traffic shock waves, as you can see here, that move backwards through the field of traffic. Now, in this experiment, these shock waves, which you'll see for me in a second, so they start off all equally spaced.

Where is it? It's gonna go here, I think, yeah. You can see this traffic shock wave beginning and moving backwards through the field of traffic. Now, the shock wave moved backwards through the traffic at about 20 miles an hour, which is very similar speeds to the shock waves that you see on motorways that have been observed in real life. And this is just an example, again, then,

of how looking at the bigger picture rather than just an individual car tells you something that you otherwise wouldn't be able to see. And this is something that has since been applied in trying to understand traffic and improving, well, improving traffic systems. Another really nice example is looking at how people move around

in pedestrian dynamics, is what it's known. Now, okay, if you imagine you have a corridor with lots of people moving in opposite directions down a corridor, right? Now, it would be possible for each individual person to just act completely as an individual,

completely of their own independence and free will, and just barge straight through and not care about anybody else, right? But it's much easier, as I'm sure you all have experience of, of picking somebody who is traveling in the same direction as you and then just following them. If you do that, if everybody does that, as we all do, what you find is that people naturally form these lanes

in pedestrian traffic. And these lanes have been really well observed. I think this experiment by Physics World, I think, the YouTube channel, just shows really clearly how these lanes form. Now, understanding how pedestrians move and the sort of macro level properties of pedestrians

have been used to look at evacuation procedures, evacuation design, and help architects and engineers to create buildings that are much more efficient by understanding how people move and behave as pedestrians. Now, the rules that pedestrians are using,

so if you, let me restart that, actually. Okay, so you know this macro level behavior, you observe these macro level characteristics, but the thing about complexity is that you want to then shrink that down and understand what individual properties lead to that big level behavior, the top level behavior and the individual properties. And it turns out that the way that people behave as pedestrians has a lot of similarities

to the way that birds behave when they flock in the sky. So these, this is memoration, starting memoration, flocking birds, making these incredible moving and evolving patterns. Now, you could be forgiven, I think, when you first see this, for thinking that there was,

for seeing this kind of macro level property, for thinking that perhaps there was a couple of birds who were in charge of the movement of these flocks and for the incredible shapes that they deform and make in the sky. But actually, something much more interesting is going on, just in the same ways with the pedestrian dynamics, because all these incredible shapes are created

just by a very simple set of rules for each individual, each individual bird, that is. And the rules are, first off, basically don't fly into other birds, but also to match the speed and direction of your neighbor. And it can be shown, you can do computer simulations of this bird movement that come up

with exactly the same types of patterns, the same characteristic patterns that you see in Starlands, making all of these movements just by those really simple rules of an individual. Okay, now, I kind of wanted to show you just in the same way as the pedestrian and the traffic, how these ideas can be exploited to make things better.

And the example that I've chosen, which I hope is the right, is Pep Guardiola's footballing style, to show you how he uses, well, how looking at it from a complex systems perspective can explain why they're so effective.

Now, okay, for those of you who are not football fans, let me just quickly and like a butcher, distill all footballing tactics into one sentence. Generally speaking, the traditional way of footballing tactics is 4-4-2 or thereabouts in this kind of structure,

push forward wherever possible and try and score goals. That's essentially, massively distilling. Thank you. So the general idea then with football tactics is that each player has a really well-defined role and position within the team. That's the traditional way of thinking about things. Now, Pep Guardiola's football style

or his tactic style has some differences and it thinks much more about the macro level of the team. Now, I have to say, this stuff is more relevant to the Barcelona than the current buying stuff. I'll show you how that links in in a minute. But while at Barcelona, who were widely regarded to be one of the best teams ever,

Pep Guardiola gave the players three very simple rules, just in the same way as the birds, had very simple rules as individuals and then the macro level behavior came from that. Pep Guardiola gave three very simple rules to his players. The first was to make triangles across the pitch. Now, if you make triangles rather than sticking to those straight lines, you give yourself always two passing opportunities.

But you also ask the players to think ahead. So to think of the future triangles and create these shapes all across the pitch. Now, here's a video where you can see this happen. So Barcelona are in the darker color and you can see the really straight lines that the other team are making while you can see how the players are moving around

to constantly create these constantly evolving triangles. So now Barcelona are in blue. Again, look at the straight lines of the other team and look how they're constantly moving around to make triangles. And then a final example, they're not worrying about holding position anymore. They're just using simple rules to create a macro level behavior of fluid football that's not the same as the traditional style.

Now, the second rule that Barcelona would use is to pass and be patient. Now, if you hold on to possession of the ball by keeping these triangles, if you constantly keep making these triangles and constantly move the ball around and don't let the opposition have any possession, then inevitably they're gonna be moving around you

and you're gonna force them into making mistakes. So the second rule was to pass and be patient, allow the other team to make a mistake and then exploit their mistake when they make it. So here's a really nice example here. So Barcelona are again in the dark. I'm sorry, it's terrible quality video but you might just be able to see. So essentially a gap opens up up here and you're gonna see Abbedale react.

He spots the gap up there and runs straight forward into the gap. So now he's in the perfect position to create a goal scoring opportunity but they need the gap to open up again. So what the players do is they carry on making these triangles, again, look at the straight lines of the other team, continually passing the ball around and being patient

until finally the gap opens up and they can exploit the opportunity, leaving him free on goal. Now I had to pause it there because he misses. But I think it explains the idea very well. Now the third rule that Barcelona used was the famous one of press for six seconds. This is the one that people knew about.

Now essentially the idea is if you ever do lose possession of the ball, what you should do is all run at the opposition player, the one with the ball and press him for six seconds. Now the idea behind this is that you narrow the field of play, leaving him with little opportunity other than to just boot the ball as far as he can.

You don't give him any passing opportunities. Now this video, I really love this video, which shows it really nicely. So Barcelona are in dark again and they kick the ball, lose possession and then they all run towards him. The guy has no choice but to kick the ball out and they regain possession.

Now it's these sorts of rules which Pep Guardiola employs, perhaps not from the kind of mathematical point of view, but it's these very, very simple rules that mean that the team Barcelona played as one moving beast, one fluid object, rather than as individuals. And it's these ideas of complexity. Now I just wanted to say

because Pep Guardiola has now moved to Bayern Munich, who unfortunately lost the Champions League semi-final rather last week, which makes this extra bit of saying how great they are, look at it. Thanks. But what I will say is that since he's been at Bayern Munich, he's only been there for a season, I think.

So they've been exploring lots of different styles, including some of the ideas that he used at Barcelona, but some others. But one thing that they have been doing is using Philipp Lahm as an anchor in the triangle system. As you can see really nicely here, the way that the players are structured around him. And Philipp Lahm now holds the record for the most successful passes in a single game of 134,

which is just astonishing. But the main point of showing all of these examples and the main reason why I wanted to include it was just to demonstrate how looking at things from the macro level can give you the chance to exploit weaknesses and to spot patterns that otherwise wouldn't be immediately obvious to you.

And looking at things from the complexity science point of view really gives you this opportunity. And this is particularly pertinent when it comes to looking at something like riots and in particular, the example that I've studied of the London riots of 2011. So just to tell you a little bit

about the riots themselves. I'm gonna miss order, hold on. So in 2011, between the 6th and the 11th of August in London, essentially the city exploded with rioting. It started initially after a very peaceful protest

after the tragic police shooting of Mark Duggan turned violent in North London. And because of the sensitive nature of the protest, police initially tried to just contain rioters rather than to move in heavy handedly and arrest people. Now over the course of that first evening,

rioting, a bus got set on fire and rioting and violence escalated with arson kind of beginning to come into play until finally they moved towards a local shopping center and started looting a local shopping center. Now, as things died down that evening

and the next day copycat riots sprang up all over London which no longer had any direct connection to the original protest. And strangely, I think, didn't really have a strong political motivation. People during the rise of 2011 were not particularly politically motivated.

I might just try and restart that, sorry. Where's my, I'll go that way. Where was that? What was I gonna say? But living in the city at the time, I think it was, well, it's important to say

just how much the city was affected by these riots. It was really unexpected to have just this complete widespread rioting across the city. On the third day was when the events reached a peak with riots, and you'll see this coming up here in a second, so this coming into the third day now with riots completely exploding across the city

and across the rest of the UK. Now, immediately after the events, the police, there you go, you can see how widespread these riots were. I should say actually, sorry, I should've said this earlier, each one of these dots relates to an arrest made in connection with an event. So each one of these is basically a person committing an offense that they were then arrested for.

Now, just to give you a few little facts about how dramatic these events were in London, there were over 4,000 arrests, 4,000 people were arrested in the riots, five fatalities, five people lost their lives during these events. Now, one thing that makes these riots sort of original

in some ways or unique in some ways is that it was a really strong emphasis on looting, much more than you've seen in previous riots because there wasn't a strong political motivation. People were really using this as an opportunity, especially in the copycat things that happened in subsequent weeks, or subsequent days, sorry. People really were using this as an opportunity

to raid shops, essentially. And there's a 250 million pound is the latest estimates of the cost to the taxpayer. Now, when these riots happened, I was working in London, and at the time I was working on looking at retail behavior, so how people shop,

people shopping habits from a complexity science point of view. And there were two maps that were printed in The Guardian, which is a very big national newspaper in the UK. These two maps were the thing that really sparked our interest and made us, or suggested, that perhaps we could look at the riots

using complexity science and mathematics. So this is the first of the maps. Now, immediately after the events, when people who were arrested appeared in court, The Guardian newspaper had sent reporters and recorded where they had offended, but also where they lived. So we have two pieces of information for every rioter,

and this is on this map here. So the riot locations are the dark, the white circles with the dark dots in the middle, and the red dots, which would be hard to see on that map, but that's where people lived. Now, this London map, it just looks a bit sort of all over the place, right? It really reflects the polycentric nature of the city

with lots and lots of shopping centers, lots of residential areas all over the place. But if you compare that to what the map of Manchester, sorry, looks like, you can see a real contrast in the signature patterns. So in Manchester, rather than everything being all over the place, there was a really, well, a violent center, essentially.

Everything happened right in the center of the city, and all of the suspects lived in a ring around the city. Now, if I had drawn, because we were studying retail at the time, if I had drawn maps which showed you where people lived to where they shopped, you would end up with something really similar to these two maps.

And the contrast between London and Manchester is two different cities, the signatures of those cities. So this was the thing that really sparked our interest and made us think that we could look at this using similar techniques to the retail side of things. So UCL, the university I work for, may have some really good connections with the police.

And as I mentioned briefly a moment ago, the police during or after these riots, they were really keen to understand how things got out of control so quickly, and if there was anything that they could have done to have minimized the damage or to bring about a quicker resolution to the unrest.

And so the questions in particular they were interested in asking were, why was it that the riots, the riots were widespread across the city, but there were some areas which were really badly affected and other areas right next door which had hardly anything at all.

And some of these areas like Brixton, for example, which is in South London has a history of rioting. There's been riots there before, but there were some other areas like Croydon or Ealing, which it just seemed really unusual that they would be so badly affected during the riots. The police really wanted to know what was it about these places that made them so susceptible to rioting.

They were also interested, as I mentioned, to know whether they had sufficient resources to deal with this and whether they could have done anything better, whether there was anything that they could have done better. Now, the way that the police managed to get a handle on the riots in the end was just really by recalling everybody from holiday, it was this in August, so everybody was away, well, a lot of people were away,

canceling all annual leave and bringing in police from all over the country and to London to try and suppress things. And you can see here how the numbers changed across the few days from three and a half thousand, roughly on the first night when things happened, all the way on Tuesday night, when they really managed to get a handle on things

of 16,000 police officers, almost four times, more than four times as many police officers were on the streets and that's how they managed to get a handle on things. And this, I think, was something that they were really interested in. Did they really need 16,000 police officers to manage to cross things or could they have done it with fewer police?

So the police gave us data on everybody who had been arrested in connection with the riots then. So this isn't, I guess you have to be aware, I suppose, that this doesn't mean that we have perfect information on the riots at all. We only know the people who got caught, really, or who were arrested,

rather, suspects who were arrested. But we do still have over 4,000 records. And in particular, the important thing, again, is we have where they committed their offense, but also where they live. So we can track how far people moved across the city and begin to try and tease out some of the patterns in why they behaved in that way as a group.

I should say again, actually, just because it's in my mind, the main thing here, as I tried to emphasize with the early stuff, is we're not trying here to make any predictions about the behavior of an individual person. And we can't talk, using these techniques, we can't talk about the motivations of an individual. This really is about looking at the big, wide-scale patterns of the city.

Now, once we got the data, we did this visualization. Now, the smaller circles are where people committed their offense. The larger circles tell you how far they traveled, essentially. So we can't plot suspect addresses on a map for obvious reasons. So it just gives you an idea of how far people were traveling. Now, your eye is quite naturally drawn

to the really big circles. Let me restart that one again, actually. Your eye is naturally drawn to the biggest circles like that one there. But actually, the vast majority of these events have very, very small circles where people traveled not very far at all. But there are a few things that immediately become quite obvious

from the patterns in the data. And the first one is the temporal signature, so the ebb and flow of things. So how things, how riots built up to a peak in the late evening or early morning and then died down overnight as police managed to regain control over the city. And if you look at just that particular thing,

if you look at how these events kind of rose and fell and rose and fell, this is the graph that you get. And you can see quite clearly there how the Monday night was just a huge thing across the city. Now, if I drew a similar graph to show you cases of seasonal flu in a country,

you'd end up with something that looks very similar. So lots of cases of seasonal flu in the winter and then that dies down over summer and then lots of cases and then dies down. And this pattern, this signature of kind of the ebb and flow of it is really reminiscent then of the way that diseases spread.

And so this, which we've subsequently done more work on, was the first thing that suggested that perhaps there was a contagious idea to riot that spread through the city. And it spread through the city in the same way that a case of the common cold might, for example. So that was the key observation, the first key observation. Now, the thing is that the way that ideas spread

or the way that viruses spread is a really old and well-studied problem, not from complexity science. It's been around for a long while and people really do understand how to look at the process of contagion. So that was kind of our first observation from the data. The second observation from the data

was where people came from. Now, in some ways, I suppose it's not a complete surprise to say that the people who were involved in the riots came from some of the most deprived areas of the city. So in the background here, the red and blue shows you

the index of multiple deprivation. Essentially, it takes into account things like income, things like how many, school qualifications, the number of unemployed people, the quality of the housing, all of those different kind of things are taken into account in this measure. So red is most deprived and you can see quite clearly that a lot of the pins,

which is where the suspects live, came from the most deprived areas of the city. But what surprised me anyway was just how stark this relationship was. It really was, this is a similar idea, but with deprivation along the side and the number of offenders at the top. And it really is, I mean, basically a straight line.

And the people who were involved really did come from some of the absolute worst areas of the city, places with the worst schools, the highest crime rates, all of these different kind of things. So that was the second really important observation that we made from the data. The third observation was the thing that brings us back to the idea of looking at shoppers

and it was where people, or how far people traveled rather. So this here is Brixton, you can see just that there. And around the outside is how far people traveled or where people came from. And you can kind of see that there's a bit of a speckled effect, which is where the deprivation feature comes in.

But there's also a bit of a radial thing, right? People generally didn't travel that far to go to the riots. It's certainly not people kind of coming from the other side of the city. And if you distill all this data down and look at a bar chart essentially of how far people traveled, you can see that it's, you know, most people in fact, really didn't travel very far at all

to go to the riots. People were rioting in their own neighborhoods. And in fact, 80% of people traveled less than three kilometers to get to the riots. Now with the work that we've been doing on retail, looking at people's shopping behavior, this is the thing which really confirmed to us that there was something that we could do here.

Because essentially the process by which the rioters were behaving is very, very similar to the way in which shoppers behave, okay? So if you're going shopping, you would prefer to shop local to where you live. You'd prefer not to travel very far to shop, but you're prepared to go a bit further for a really big retail center, right?

Now, essentially rioters were behaving in the same way. They were choosing riot locations that were close to where they live, but they were prepared to travel further for a really big riot site. And this is very particular to the London riots because looting was so important.

People were essentially choosing retail centers in the same mindset as they would if they were buying stuff, but they were going there and looting instead. And immediately after the events, a lot of the UK papers ran with the headline, shopping with violence, which is a bit flippant, but it does really capture what we found in our data.

And in fact, actually, you can see here, this straight line is what you would expect if you were looking at the behavior of people going shopping. And the dotted line is how the rioters behave. So it's really, really similar, the characteristics. Okay, so what we wanted to do then is we wanted to try and come up with a mathematical representation of the events.

And we wanted something that couldn't exactly predict when a riot was gonna happen or when a riot was gonna start, but could replicate the general patterns that you saw in the riots. There's no way that we could actually predict exactly where the next riot site was gonna be because we didn't have complete information.

There's just no way possible. We only had the people who were arrested. So we wanted to create something that was capable of replicating these patterns. And so using these things that we found out from the data and also all of the ideas from complexity science, what we did is we came up with a three stages, a three-stage model.

Now, in the first stage, as people decide whether or not to participate, inactive residents decide whether or not to get involved in the riots, this is very similar to the idea of infection that we were talking about. As soon as people have decided that they're going to riot, they then choose where they're gonna go based on the way that people behave in a retail setting,

which is also a very old and well-studied problem. And finally, when they get to a riot site and they're there with police, they interact with police according to a model of civil violence, which again is a very old and well-studied problem. So let me just try and explain to you how it works with the diagram. Essentially, you have a network of homes and shops, right?

A rioter decides whether or a person, a resident, somebody who lives in the city decides whether or not they want to be involved in the riots. When they decide that they want to be involved, they choose where they're gonna go as though they were going shopping, essentially. And when they get there, they interact with the police

and there's a deterrent effect of the police presence, which is important. Now, I'm not gonna show you the equations properly, you can read the academic paper if you want to, but just to say that this is, I mean, it's got a sound theoretical basis and it's highly interactive as well. There's lots of bits which kind of feed into each other.

So that's where the real complexity of it comes in. Okay, so I guess how did the model do? How well were we at generating these patterns? And actually we didn't do too badly. So if you squash all five days worth of data onto a single event, all five days worth of the arrest data onto a single event and say that it all happened at exactly the same time,

and then compare that to a model that starts everything off altogether at the same time, you'd never expect them to be the same, right? You'd never expect them to be identical, but it does because there were different characteristics across the days, which is another academic paper, which you can read if you like.

But still we do all right, basically. In 26 of the 32 boroughs, we're in the same or the neighboring category of events, but where the model is really powerful, I think, and really gives us something is in the way that it tells you how different areas of the city were susceptible or whether or not different areas of the city were susceptible.

So essentially the four worst hit areas of the city were Brixton, Croydon, Clapham Junction, and Ealing. These are the four worst hit areas. The retail centers had the biggest damage and the biggest problems. Now what we do in the model is we pair off Brixton with its closest retail center.

So there are two neighboring retail centers and we pair them off in the model. And we start off a very, very small riot in both of those areas, and then we wait to see what happens. Now, in some cases, the riot then explodes, draws in more people and explodes. And in other cases, it dies away into nothing.

And we record what happens in that experiment, in that experiment in the model. And we repeat that experiment by pairing Brixton off against Clapham, another neighboring retail site, and then pairing off Brixton against Camberwell and then Brixton against West Norwood. And we take the average of what happens in Brixton and compare it to what happened

in all of those four areas, the four neighboring retail centers. And in every single case, Brixton explodes while the others die away into nothing, just as happened in the real events themselves. And we repeat that experiment for Croydon, Clapham Junction and Ealing. And in almost all of the cases,

even at this, well, a slight different thing with West Ealing retail core, but in almost all of the cases, it's the places in reality which exploded are the ones that our model picks out as really susceptible areas. Now, this is important because it means that you can use it to inform the police about where is likely to be somewhere

that is likely to be susceptible, I should say. And in theory, you could have a surface of the city where you say to police, these are the areas that you've got to focus on. And even areas like Croydon, which was a bit surprising that they came out, our model's able to predict that they were susceptible. There's also some stuff that we can do about looking at,

because rioters interact with police in the model, you can also use this to set up sort of an imaginary riot and then use it to explore different policing strategies and see how important different policing strategies are because our rioters react to police. And the two things very simply that we explored are how many police officers you need to quash things

and also how important police response time is. Now, you have to be really careful because a person in the model doesn't necessarily relate to an exact person. One person and one person is not exactly the same thing, so you have to be a bit careful.

But with a huge bucket load of salt, you could very loosely say that perhaps 16,000 police officers weren't necessary to quash things and perhaps there was something a bit smaller that could have been used rather than bringing everyone in before they got hands-on events. But the thing is, we're mathematicians, right? It doesn't make sense for us to sit in an office

and say, oh, I think the police would act in this way or to make up a policing strategy to conjure one up. What's much more important is if the police themselves can use these tools to try out different things because we can create an imaginary riot that has the same characteristics as a real riot. The police can use this to try out different things.

So we created a game and at this stage it is still a game, I should say, and quite silly, I'll be honest. But we created a game where it's a touch table with a map of London where riots kind of crop up and down across the city. And at the moment, this is a very silly version,

you place Lego cars, Lego police cars on the table. The table can detect where you've placed your police resources and the model in the background reacts and yeah, it gives you a score and so on. But the idea in general is to wrap up all of the maps put it into an interface that is actually useful

that police can use to explore different strategies and inform them using all of the ideas of complexity into, inform them about what the macro level effects of their individual behaviors are. Yes, should I leave it there? Yes, I'll leave it there.

Thank you very much. Thank you very much. Are there any questions? We have a few minutes left, yes. Hello, thank you very much.

I have a question. I'm sorry, I'm a bit late to the talk so I didn't get the beginning and I'm not sure maybe if something I'm asking now was already covered there. But my question is, what do you think or what do others think about the ethical implications of this kind of research?

What you talked about was that we look at some sort of scientific technology that enables us to look at rioting. But what if we exchange rioting for demonstration? What about the way we already look at rioting? Like you also said that the rioters came

from not very privileged neighborhoods and that the rioting itself also started with a specific event. So, and I'm also looking back towards the talk we had by Sasha Lobo, who said that technologies

are not just technologies in a very neutral objective sense, but they also have purpose built and they have a social purpose built. And I wonder, also looking back in history, if we think if this technology continues to advance, gets better, gets more complex,

even better in predicting things. If we look back, for example, to the French Revolution, where would it be today if we back then had technology to prevent a French Revolution? No, of course, I agree. What would Foucault have said about this kind of biopower? Okay, so actually, well, I had one more slide

which I thought was probably in the interest of time worth shortening, which sort of addresses what you're trying to say. But essentially, I think that there's a couple of really important points to make. The first here is that this work in particular is very specific to London and the London riots

in particular. And the reason for that is because of how much looting there was. And the second thing is because there was no political motivation for people, they were very easily deterred by police. And the thing is, is that that is really integral to the way that this particular thing works. Because putting people off, going and looting

a pair of trainers is much, much easier than it is to stop somebody who's legitimately fighting for their freedom. Sorry, I can't hear you. Okay, all right, fair point. I think in this particular instance, there wasn't a united political aim.

So I think that's kind of the difference. And I think deterring people from doing something like an opportunistic chance to steal a pair of trainers, which essentially is what a lot of the evidence showed in the London riots, is very, very different from something like Syria or something like that. I don't think that you can use these ideas in something where people are so much more motivated

to show their opinion. I also think, though, that it's also really important to just make the point that this is not about making predictions. This is not about saying in 15 minutes time, there will be a riot on this street at this time. It's not about that at all.

This is about just understanding and exploring the general characteristics in the way that we behave. And I personally think that while data privacy and handing over too much control to technology is a really scary thing, and I completely agree with all the points that have been made several times over at Republica,

I still think that there is something positive to be gained by looking at the macro level behavior of people in the way that we can design our society. You looked at how to suppress the riots.

Are there also any hooks to make the riots worse? For instance, manipulating the infection process or the decision process of where to go rioting? I confess I didn't look at that. I'm just trying to think if it would be possible.

Certainly not with what we've done, because it would be like saying, how do you optimize where you go shopping? I mean, essentially that would be the analogy. And really it's about what you as an individual are looking for. And I think maybe that's the point then,

is that because we're looking at this from a macro level, those kind of individual choices don't really come into it that much. Sorry. Hello. Hello. That was a very interesting speech, though still, as far as I understood, in the long run, it came to the result that the most riots or the strongest riots

came from underprivileged areas, which had also traditioned those rights. So a bit provoking ask, haven't you not only proved mathematically what common sense or experience what most people are, and especially policemen tell already? Yeah. So this was something that I was really interested in.

I mean, the mathematics that we were doing was really about looking at the patterns, the macro level patterns, right? But I personally was really interested in exactly that story, that people were coming from really deprived neighborhoods. And so I also made a short film, which you can watch online, where I went and interviewed lots of different writers,

people who were actually involved in the events and tried to understand from their perspective why they had got involved. And I think that that is a really important point, that it's essential to add and complement this story. My talk went on a little long. It was gonna be contained in it. But I think that really we should use these ideas

of looking at riots as much to kind of try and understand why, for example, our young people in London were in a situation where that was what they were doing, as much as trying to help the police to stop the city from turning into a riot zone. I think you have to kind of have that coupled approach.

Have I understood you correctly, that your data set or your analysis is based on a data set of police arrests? It's, we use patterns that we found in the data set to inform a data-free model. But aren't police arrests, or hasn't sociological research shown that police arrests are anything but neutral?

Can they really represent who the rioters were? People, particularly, especially underprivileged groups of society tend to be arrested more, and hence your model is based on a screwed data set, now reinforces that screwed data set by telling the police that they should go

to these areas even more? Well, okay, so I think there's a couple of important things that you said there. The first is that you're absolutely right, that you have to be aware of the flaws in the data. I think that's really important. And I think that, yeah, understanding that you don't have a complete data set, you don't know everybody who was involved in the riots,

only the people that were arrested, I think that I'd add to that that the data set that we have is so large that we're picking out macro-level patterns that completely appear within that data and using that to inform assumptions about the way that people behaved. So while I agree about you have to be careful

about understanding the limitations of your data, I think you have to be careful about the limitations of this kind of mathematical modeling in total. You have to understand what it can offer you and what it can't offer you. I forgot the second part of your question. I'm so sorry.

There you are, let's see. Hi. Hello. Well, that was a really interesting formula that you came up with, and I hope one day you actually are able to predict with more precision than you are. But owing to the fact that the London government already has some really crazy surveillance methods for you guys living there,

don't you think that this is just too much power to the hands, in the hands of the London government against people that want to express themselves, maybe not in terms that will be beneficial for the government, but more for the people? Yeah, so I think, I mean, it's a similar point. I agree with you, I completely agree with you.

But I think I also really, really want to stress here, I think that with this maths and with this science, I think it's incredibly important that people, as in everybody, the layman, politicians, everybody, really understands what it can offer and what it can't offer. And what this can't offer is telling you

that you need to immediately go to this place and do this and so on. This doesn't give you control. What it does give you is an understanding for why people are doing things in the way that they can and highlights areas that you can improve, like, you know, focusing on education or other things for deprived areas or youth workers was a really big thing that came out

in the interviews that I gave, focusing on those kinds of ideas. It doesn't, this is not a control thing. And I think that people are generally and understandably wary of science and technology controlling their lives. And I think that there have been examples in the past where science and mathematical models like this

have been completely misused and been promised, you've been promised the world that they'll do things they just can't do. I mean, the global economic crisis is in some part, at least a result of people irresponsibly using mathematical models. And I think that it's really important, incredibly important that these types of things do exist because I really believe in them.

But I also think it's really important that people make the effort to understand what the limitations of mathematical models are, because I understand the limitations because I made it. But I think, you know, it's really important that people, you know, communicate where the limitations are, I guess. And yeah, well, hello.

Hi, first of all, I'd like to thank you, I guess, on behalf of the whole audience for a very insightful and interesting talk, thank you. Yes, that's right. So I happen to be in London.

When the riots broke out and being a sociologist and also being a little bit riot experienced as an observer, I went from Clapham Junction to Hikeme to Brixton, Croydon and so on just to observe what's happening. And I made some observations that I found quite interesting, I guess they are linked somehow. First of all, I saw that contrary to what the media gave us as a picture of the people in the third district,

two-thirds of the people, of the inhabitants of Brixton, Croydon and so on,

seemed to be against what was happening down on their streets, in their shops and so on. So they strongly opposed looting and stealing and everything. But interestingly enough, they were lining up on the street, observing it, and they were booing and they were applauding the police when they finally came in.

But they didn't stop the looters, although they were the clear majority. Whereas in East London, Turkish shop owners did the following. They just banded up, armed themselves and pushed the looters back within no time, like you don't mess with us.

No Turkish shop was being looted. And I found it interesting because if you have ever watched the first of my May riots, for example, in Istanbul, you clearly see that they have a whole different tradition in writing. It's way more like they got this riot pattern. Everybody knows what to do.

It's not organized chaos, but everybody knows where to go when the police rush towards you and so on. So I found that in democratic societies where you give up all the responsibility for what's happening in your neighbourhood and also the violence monopoly.

You give it up to the state. You've been taught from childhood that if something happens, call the police and they will solve it out.

Whereas in other colleges, you have to do it for yourself. And I found maybe there are also problematic indications for us. I don't know if it's the same here, but in England, the papers really make a big deal out of any story

where somebody tried to intervene and then there was backlash. So there was an example a little while ago where a father saw some teenagers doing something in the street. I can't remember exactly what it was. Went out to go and tell them off and then was stabbed, essentially. And papers make a really big deal out of any case where somebody who intervenes ends up being injured.

And I think that there is a general sense of fear across the UK that if you get involved and if you try and stop somebody from doing something, it'll end up coming back on you. And I would guess, obviously I haven't done a study of it or any kind of academic overview, but just as somebody who lives in London and somebody who lives in the UK,

that would certainly be the reason why I might hold back and wait for the police to get there. But you're right that it does seem strange when people are so clearly outnumbered why it would be possible for people to be able to take control. So thank you very much, Anna. And your applause, please. Thank you.

I think there are a lot of other questions. Perhaps you can do this.