James Ruse

Transcription

>> [Applause] So, yes. So, I'm going to talk about Social Physics in the Big City. What that means to me and what that means to the world. Soccer players [inaudible] myself, Jason's done a wonderful instruction but just as sort of a general backup which is me, with slightly less of a beard. And my name's Martin Zaltz Austwick and I did a physics undergrad, undergraduate degree and I took that and did a, a specializing in antisocial physics because, by default, physics in fairly antisocial, at least for the purposes of this, this slide, and if there are some physicists in the audience I apologize. And my [inaudible] was in solid state physics and -- in the materials department. I still call myself a physicist though. At that point I do. And went in quantum computing and then in technology and after that I decided that wasn't really sociable enough so I, I went into [inaudible] something a bit more involving human beings and I went to Medical Laser Physics for a while, still calling myself a physicist just about. And then I joined CASA Center for advanced special analysis in the bar that's two years ago. And I became a lecturer last year, and occasionally if people ask me I say that I'm a Social Physicist apart from this friendly aspect to it. This describes the discipline I work in to some degree. So what is Social Physics. Well, the term was coined by this man, August Comte and this man Adolph Quatelet. And they were working [inaudible] August Comte was doing work at applying mathematical methods from, from physics and to new social data sets, new information the state was releasing. And in fact, later stopped using applied social physics because this other chap started using it and the field that he founded became another sociologist and he was one of the founding fathers of sociology. To sort of trap Quatelet he was doing some very much similar work and the work that he now does -- the work that he does would now be called statistics, so he was one of the first statisticians along with the original statistical techniques came over from astronomy and physics and trying to understand datasets about human beings. In fact the name statistics comes from state as in the, at the country, the nation state. So, it doesn't exactly answer the question. this is all in the 1830s by the way. So this, this isn't anything [inaudible] this this is from that -- almost 200 years ago, now. He used to tell the people [inaudible] to be social physicists because they thought about using math to predict the future [inaudible] in the course of society. And the man on the left, as you're looking up there, is Carmont. He though that by understanding capitalism mathematically he would predict a standfall. He convinced them to seeds that would -- we would sow same destruction. The chap on the, on the right is Isaac Hassimoff. So it's fiction [inaudible] he came up with this idea [inaudible] history, and he said that -- what he conquered at the behavior of individuals maybe you can statistically aggregate the whole society and see the way that society was going. It was more Degrandier's plans, marks on [inaudible] successful at causing the downfall of capitalism at least not yet. And, and Isaac Casimow [inaudible] generally regarded as a serious academic. So, what is social physics? So much for the dead white guy theory of social physics. No one ever terribly successful. Those who were successful changed their name from physics to avoid the, the essential recriminations. So, I've got Social Physics applying mathematical techniques to social systems. And actually, lots of people are doing this already. This is not that new, again. It's statistics -- it's using mathematical techniques. It's using complete stimulations. It's not doing -- using that much physics. There are little bits of physics that have come over. No meaty statistics though come over to these sorts of studies which have been useful, but generally there's not that much physics going on. It, it's just a sort of a, a, a name, really. So you could just say this is quantitative social science and by which I mean things which deal with numbers and not feelings. But that's a sort of -- I'll put this aside so I could qualitate the social science and actually is kind of important what those numbers mean. And so I end up just, just, just to say that. I'm a, I'm a quantitative person and I, and I can do a list analysis. What do those things mean? I don't have them in context. Okay. So, if we fast-forward to the present day and we start to talk about the, the, the title of, of this lecture which is Social Physics and the Science of Cities. So while we're looking at the whole of society, I'm going to talk a bit more about what my department does and what people in the -- in my field do, which is more, generally more for cities focus. There was -- there are people in my department who don't explicitly focus on cities, but that's of now is the topic about modern -- talking about the whole of the world from the -- every human being in it. So what kind of -- how do you [inaudible] how do you think about science of cities? What kind of questions do you need to ask yourself? How do the [inaudible] make themselves useful in this, in this context in social science? What does a -- what would a science of cities look like or is there just one? Do you need to have more than one science of cities? That's probably yes, at the moment, certainly. What sort of theories do we need to develop? Or Cervantes doing it. What sort of people are doing it? Well, I can say -- I can tell you what sort of people are in CASA. It's a -- mostly an interdisciplinary effort and what we have is planners, computer scientists, there's a physicist, another drug -- there's a drug [inaudible], another computer scientist, photographers, network theorists, a psychologist, transport economics, another physicist, mathematicians, and another planner, more mathematicians. So lots of range of different -- this isn't all the people in my department by the way. This is just all the -- and just, just a few people to give you a flag on what they're currently -- interdisciplinary -- they're into interdisciplinary research, research, but we don't -- you can't just come in with one discipline for this new sort of -- to fill these new techniques. You need to have people who, who know math, seem to know -- people who know how to build databases. You need to have people actually understand -- be witnesses who are planners, who are architects, who are geographers. So the, the answer is -- the kind of people that you need are from bordering disciplines. You need them more to come together and to work together, to understand cities. Now one of the first things that I think you need to do is when you're looking at -- when you're looking at data source around a city -- and remember, when we start from this perspective if you've got data about the city -- one of the first things you have to do to try and understand that is to visualize it. Even before you get into mathematical answers, statistics, and graphing all that sort of stuff, is just to start to visualize the data to get sense of what is this data that we're looking at, what does it mean?, can we spot any patterns in it that will inform how we do the analysis and, and develop models around it. There's obviously -- this is an example of a visualization. It's a, it's a map which a visualization guru had [inaudible] he called -- I think he said it was the whitest visualization ever produced [inaudible]. And it shows the, the pattern Napoleon's army to and from Moscow. And the width for that line tells you how big his army is. That sort of stuff pretty large. By the time he gets to Moscow, it's not doing so well. By the time he gets all the way back it's, it's really doing to a tiny faction of its size. And there's some other information here as well if it tells you things like the temperature, which is obviously a failure to some [inaudible] factor in this, in this, the, the forced arm -- march of his army. But, with a very simple visualization there's an enormous amount of information communicated there. You've got vents, you've got geography, you've got, you've got a ruse, and you've got numbers of, of troops. So just by looking at that, I mean, we could talk, we could -- I could talk around this or I could produce graphs to show what was happening with his army, but that visualization really -- it captures a lot of the story. That case -- so, I've sort of skipped ahead though. I've said this is what we do when we get there to about how do we get data in the first place? How do we start to gather the information that we need to formulate theories of the cities into -- and to, and to understand the way a city works? A lot of data has already been collected. I mean you can go ahead and collect data, but there's a lot of stuff out there already. And you have things like Oyster Cards which are constantly collecting data about people's movements around the city. You have traffic cameras which are constantly collecting information about cars and trucks and, and, and obviously public transport vehicles on the roads. You have people's mobile devices and, and in some cases, people have had to volunteer information. You have things like smart meters which is meters for measuring energies in the home. So just starting to get into this we can start to think about all of these rich data sources that might be relevant to what we're interested in. So this is a, a visualization produced by John Reades at CASA. And it's showing oyster -- based on oyster card data which parts of the [inaudible] line are busy over the course of a, a Monday, in this case. So what you should see is a pulse -- there you go. That's the evening pulse as those parts of the city as -- those, those, those parts of the Central Line in this case. You can see the red centralized. It's particularly affected. But those parts achievement worker are flooded with commuters. Almost like a heartbeat. So, again, this is, this is just a assume provisionalization showing those raw numbers, and, and it's not the touching inter [inaudible] is working at from where people touch that to where the touchstone where were they on the [inaudible] at any particular time. So he's extrapolating back from that -- so terminus information. Starting to get a layer of insight into this system. Okay. [ Background Sounds ] Good. Some of sort -- that's a, a, a [inaudible] that we're in some sense volunteering anyway. Obviously, there're, there was some privacy issues that we need to think about for that. But there is actually data which is volunteered by us all the time, and actually with our knowledge, social media being of this one. People who use Twitter, who use Foursquare, they're broadcasting information about their lives. Maybe with Foursquare it's a lot of information about, about where they are. That can -- that happens with Twitter as well because people can geo-locate their tweets. And so not only are they providing information about what they're doing, they're providing information about where they are. And they're also -- there's also an enormous community of volunteer collectors, people that volunteer information about their lifestyle or, or, or where they are as well, which I'll -- I'll come into that later. And there's this concept of Internet of Things, as I did, that you can link objects, artifacts, places in the real world to an internet presence, and there's a big product in CASA which is involved in that coattails of things. And this is a, a, a photo taken in an [inaudible] where these objects have these QR Codes. You can't see them but I'll tell you but those are the two dimensional barcodes. There's little tags on each show these objects. You can scan it with your Smartphone with the, with the appropriate software, and it gives you a history of that object. Some people can conjugate to the story of that, of that item. They can contribute images, videos, text on the website. So when you scan that, you can see where this item has been and his habit -- for hanging on when you can start to build a, a sort of story around those individual items. And that's not quite at the stage where you're going to aggregate that up and, you know, you just have to sketch some things on it, but it starts to tie in a physical object's [inaudible] to a, an internet and a data presence. And a sample of volunteer data -- I'll just open it -- is GPS tracking as a big product down in CASA which is involved in that. This is actually my GPS track. I've got one of those little devices and it -- you can do this with a Smartphone -- but if you have a, a dedicated device, it doesn't drain your battery basically. So I have one of these GPS devices and when I, when I walk around London and I saw where I had gone. And, and, you can do this. Obviously for more than one person, you can aggregate over lots of people and you can start to ask questions about how long are people spending in their homes. It has a [inaudible] what mode of public transport are people taking? Especially if you combine that with an accelerometer so you can get some sense of acceleration and speed. Going back to the [inaudible] media, you can look at the geo-located tweets that, that, the tweets with geographical information and say, were people tweeting in London or any city for that matter, in a geographical region? And you -- and this is what Fabian and, and Steve did in CASA, and they formed this map of, of London and where people were tweeting, and surprisingly, there's a large tweet that's in the center of the city where there's a lot of people, a lot of people are working. But as you get further out, you see, isn't it, excuse me, there's an unexpected little islands of activity which were set above the plane. It's not homogeneous distribution. It's not little pockets of people tweeting a little bit more animatedly than in any other parts of the city. And you can go beyond that and you cannot only analyze the location of the tweets but you can start to look at the content of the tweets. I think this is a really nice piece of work that James and Ed did -- Ed and Geomatics. You can see he's not based in CASA but it works with us quite a lot. And [inaudible] is the language of the, of geo-located tweets. And what you have here is a map color-coded by the language that tweet was in. So you're seeing essentially a language map of London based on this sample of tweets. And I think that's lovely. So you can see that this is -- this is on the web and you can, you can look at this. You actually zoom in and explore it in a bit more detail. But you can see how different communities exist or form across London. You can see where they're located essentially. Obviously what you get [inaudible] is where that's where they live or whether they -- that's where they work or -- and [inaudible] have a letter [inaudible]. And on a larger scale, there's this initiative for open later. So this is not necessarily this individuals know this is, this is larger groups of people saying this data should be made public, it should be made accessible to academics, to developers, and, and to the public at large -- as those, as to the public as, as a whole. And local and central government groups increasingly making the data available to people like me. But to, but to everyone as well. And there is limited because of crime series and so there is limited transport for London data available that is available to everybody as well. And this is, this is in some kind of sort of combined one tariff. This is our constraint map. This is a sort of a, a, parallel if you like to Google maps except for the individuals or groups of individuals. They go in and they do the mapping themselves, and they upload information to this database. So this database has not only the, the map images that has the underlying data, so you can actually use that to create other maps and to manipulate it in a much more interesting way than Goggle itself allows because that underlying map data is, is, is proprietary and, and it's commercial. So the public is subject to public transport just by looking at timetable data and the locations of the places those public transport vehicles are supposed to be, you can create a map like this, an animation of public transport activity in London. So, hey, this isn't [inaudible] that Joe, Joan has done. You have trains, busses, coaches, the Tube, and I think you have trams and you have riv, rivet, river busses as well -- boats, essentially. And so what this does -- this is not very exciting, actually it's not getting -- it's not quite waking up. And as you see, we get into 6:00 a.m. and the whole city's waking up and the whole thing suddenly lights up like a Christmas tree as the Tube Network comes on line and the busses swap [inaudible] night busses and you see the whole thing wake up. And this is not based on any GPS data or about any sensing data, it's just timetables and the locations of the bus stops, and Tube stations, the, the docking points for the, for the boats, and the tram stops. So, this is how London should operate. Whether it actually does -- do that in practice is a slightly more complex question. And so, you could do -- we can do the same thing with bike data. This is something I worked on with Ollie O'Brien, which is to look at the locations of the Bikeshare, then have [inaudible] bike, the [inaudible] bikes, or penny farthings depending on your political persuasion. And again, we don't have GPS data. We can make reasonable inferences about their routes and we can produce animations of over this course of a day. Where are people going? So, again, not, not based on GPS data. there is some inference there. And this is Christmas Day, 2010. Actually, a very busy day for the screen because public transport doesn't run on Christmas Day. So that there is bitterly called cold and on Christmas Day is actually very active day for the screen [inaudible]. And I'm taking -- and you can go beyond this. You can go beyond these visualizations and you can start to extract analysis. I'll talk about that a little bit later. Okay. So then this other -- that, that -- we talked about open data to grow volunteer data and the data that maybe we've volunteer without quite realizing it, and that brings us on to sort of the sensitive data issues. I mean these things are still really important questions for cities and for society as a whole. And now amenable to analysis, we can add value and we can get value by doing that. But when you're talking about crime or health or people's incomes, and potentially even social media, do people want that information being shared? You might feel comfortable with your Twitter data being shared but maybe your Facebook data has some more personal service for example. Or maybe info [inaudible] with either. So these datasets -- there's still things we want to use and understand but we have to be more sensitive about the way that we use them. Now this is some data -- this is a visualization we did based on the London riots. This is over the five days of the, of incidents in 2011, and what you have is each one of those, those small spots is an incident. Each one of those larger circles indicates how far the, the person traveled to the incident or the crime when they're -- there and the point of their arrest, so the, the arrest point is the small circle, and the, and the large circle shows how, how far they travelled. So looking at that you can, you can start to see what happens that actually at the majority of people travelled a short distance but there were some larger distances to be travelled, and, and that leads on to the analysis of well, how to we quote for how do we understand that a bit more -- but more specifically to... Okay. So, so that -- so the outlines is one approach to understanding these datasets when a pictures starting from [inaudible]. This is not the only way to do it. But if, if you imagine you have a dataset and typically these data, datasets need some cleaning and some storage. You can't just take them and you need to tidy them up and make sure that there's no errors in there and wait for [inaudible]. And if you're not [inaudible] with data and your data scientists or quantitative people, they'll tell you that, that's the biggest part of the job is is fiddling around with data and making, making it look, look nice. And, and you can straightway visualize that as, as I showed in the previous slides. There's really straight visualization with very little analysis, but that means you can immediately start to share that with people. And, you can use that yourself to inform your research, but you can share it with other people and show, hey, isn't it really interesting what, you know, the pattern of, of public transport in London. Then you get on to the analysis that informs -- that can inform analysis. Obviously the analysis happens on the data itself. And that gives you some sort of outputs [inaudible]. Again, those are things that you can visualize and share. And then, the other aspect, of course, if theory and model development. Why these things behave as -- why do we think these things are happening? And that's informed by the analysis, but also you have to analyze the quality of your models and compare them to the real data. So, so there's this two-way process of, of building theories and testimony. So, that also produces outputs which you can then feed back through the analysis and feed through to the visualization, so that's the sort of slightly messy version of maybe how you can think about analyzing a dataset and it's -- suddenly another way to do is start with theory and, and, and [inaudible] but that doesn't always say -- it doesn't always happen that way in practice. Okay, so the systems we're dealing with, with cities, with social systems, that complex systems and what I mean by that is they -- there's many simple interacting parts producing a more complex whole. And you might argue that human beings are inherently complex if that's probably true as well, but in this particular circumstance maybe you can say that [inaudible] is quite simple. Now if this decision whether to get on a bus or not, yeah, that's, that, that doesn't necessarily have to incorporate that full range of experiences and, and, and emotions. So a complex system is [inaudible] so a complicated system is something where there's lots of components. It's very, very messy. And they all build together to form something which is simple. Okay. So like in a car engine. I don't know how a car engine works. I'm -- I don't own a car. I've never attempted to repair one. As far as I'm concerned it's a very complex thing for making a wheel turn. And it doesn't have -- and it's not supposed to blow up. Those are the two, two rules for a car engine. But the way that you get there is extremely complicated because you have to have gaskets and carburetors and injectors and things. Whereas opposed to, as opposed to a complex system where you end up -- you begin with simple elements and you create something really complex and that complex thing that you create is not necessarily a product of some design. I realize that's controversial as I'm showing there a beautiful snowflake. But, in this case, you can argue that -- you could start off with simple water like [inaudible]. And the way those interact produces patterns much more complex than you would expect from, from just those simple elements. And much more varied and, and potentially unpredictable as well. Okay. So how can we think about understanding these complex systems? I'll show you how to -- how, how we get the data, how we visualize that, how we start to get some insight into how we go a bit further and actually get some, some analysis. How do we -- how do we tackle these complex systems? Well, the first answer is data tools. 10 years ago, 15 years ago, you couldn't have stored a dataset which was the 11 million bike journeys that [inaudible] released, or if we could have, it would have been incredibly expensive. Now, now, that's at the reach of anybody with a desktop computer. And database tools are increasingly available and increasingly sophisticated. That leads on to sosh -- statistical analyses. This is the first graph of a lecture I'm quite pleased with. This is a [inaudible] graph and this is, this is actually to some work that a, a team in CASA doing one on cities trying to look at universal rules -- you know, that a [inaudible] rules in, in cities. That, that follows on some work, from some work in Santa Fe Institutes in, in, in the sights. And this is just looking at -- try and look at the correlation of the population of a city with certain other parameters. So this is the population vs the income. And it's showing a quite nice linear trend here, so that's interesting in itself. But you can start to do this with all kinds of different factors like number of patents a city produces or the, the infrastructure, the miles of roads that a city has vs its population, and you start to see some interesting things across multiple cities which show consistency. So you're not focusing on the uniqueness of a city. Obviously, every city is unique. You're focusing on factors which has in common the, the similarities which allay to [inaudible] as a similar object. So taxonomically. So you can go beyond that, maybe on statistical analysis and starts with the artificial intelligence. And artificial intelligence of having Skynet, so that mean [inaudible] of all powerful computer that rain nuclear destruction on our heads. I just mean software tools that try and find patterns, clusters, correlations within datasets. And that's a very much -- I mean, that's a relatively current field I would say. That's a data mining -- and it [inaudible] intelligence statistical analysis, but it goes to spec beyond the... Another aspect which is I think a really interesting area is network theory which explores the connections between objects. This is actually a, a sort of [inaudible]. You see on the middle one and there's green things of a -- they're different schools and incubate is a faculty and an off-the-faculty [inaudible]. So this is a very high arcical structure because that's, that's the way these cells organize but you can do that for any kind of system. You can do that for social systems, you can look at network academiological networks and disease-spreading. You can look at mobile phone networks. You can look at transport networks. And you can start to say things about the way these systems connect. Take easy -- is to move across this network, not in a geographical way but in terms of the number of connections going from A to B in terms of which nodes and ports and for, for traffic. Which nodes are important for transmitting or receiving entities, whatever those entities might be. Whether there's a bikes or, or cars or whatever they are. So that's a very interesting -- that's something that happens quite a lot at CASA and, and other parts of the field. You can also do a lot of old-fashioned mathematics. By which I mean pages of algebra like this, where you're not [inaudible] can be simulation. You're not doing statistics, you're just trying to create a mathematical model which explains how something behaves. But you can do the other option. You can also do things which are more like simulations, which are more like creating entities which are a bit, a bit like agent. We call them agents, but they're a bit like individuals, whatever that individual might be. They could be human beings or they could be cars queuing. And then we, we see what the output is of those simulations. We let them interact. We give for each individual set of rules and we let them interact with the -- some virtual environment and each other, and we see what happens. And this is an example of that. This is a -- I'd buy the pair of them Microsoft connect sensor. These bibles here which projects on the table are able to see these physical objects, these cardboard boxes and this Coke and that means you can actually move around space and we configure it and you can look at how these literal pedestrians can move through the space, how they're affected if you, if you put a wall to block them off or channel them down into a bottleneck what happens to their behavior. And, and it certain -- it's very hard to make -- to, to, to replicate human behavior. But, but this is a way to -- a step towards this sort of cartoonish version, if you like. It's a, it's a first step to trying to be like that. Okay. So, I'll, I'll, I'll just sort of finish with a few benefits from this. After we do all this, after doing this analysis, this visualization, we've got all this data. This is -- it's fantastic for us but who, who, who benefits from this brave new world of, of data measurements and analysis and, and so another way -- our lecturer probably said, hey, can we share this people? And there's a lot of people who, who, who benefit from this actually. So how can we share this with them? And there's lots of ways we can do that. This is, this is an example of a, a website that James and I set up. We just -- sort of boundary changes across, across England. And allows the user to slide between the current and past and the future proposed changes. Start seeing how that affects them. And this a , a project that Molly O'Brien has worked on which is city dashboard for London. It aggregates all of this information that we've been talking about and information about Tube lines, weather, radiation came weirdly, and it puts them all on one page and it allows people to still access all the information they need about their city. And this a sort of a prototype. What else could we do that would [inaudible] take us further? We could look at Smartphone apps and that's another way that you could -- you can deliver information but also collect information. It can be a two-way process. This is an app developed by George McCarran who is -- am [inaudible] fairly recently and he developed this -- his time [inaudible]. This is an app that asks you how you feel. Are you in a good mood? Where are you? And it allows you to map happiness, satisfaction, across the UK. And this is a slightly more fun -- well, mapping this is a lot of fun. This is a slightly more wacky example. It's got pigeons in it. It's easy to connect and allows people in a physical space -- it's likely -- this is a, a human here -- to fly around the city, literally fly around a Google map, through the Google map of -- at the Googler's view of, of London. And not only that it allow you to explore your city, it also allows you to put eggs onto that map. It's not very clear from this but there were geo-tech tweets appearing over that map. So as you fly through, you can see what people are saying and, and where they're saying it. So that, that's just a sort of different way to represent the same things, represent the same data in a way that's really accessible for both [inaudible]. Okay, so, to wrap up what have I learned? I've learned that "Social Physics' has been around for over 150 years. I'm happy to stop caring and using it really. Data is getting more and more abundant, so that, that's a fantastic time for people like me and I think for the world in general. Understanding is hard because it's complex data and we need to rise to the challenge. We need to develop new tools. We need to use computational techniques. They're the best techniques. Statistics, maps, piled all together and, and turn that into a, a, an interpretation in a, in a story. And we need to think about how we private -- provide value for people sharing their data. If people give you data, what are you going to give back to them? And the people who have [inaudible] systems as a whole. Public transport affects the majority of us. And that citizens -- that's people who use the service, that's people who provide the service, that's government and potentially other social institutions. And so I will end there and leave 10 minutes for questions. Thank you very much. [ Applause ] >> Okay. As indicated, we have a good eight, nine minutes for questions. Those of you who are veterans of Lunch Hour Lectures will know that when called upon to give a question you have to wait until one of these fine people brings you a microphone so that the many millions who are watching, not to give you any stress, the many millions who are watching online will be able to hear your very intelligible question -- intelligent and hopefully intelligible. Who has a question for Martin? All right. One over there. >> Thanks for an interesting lecture, first of all. So with all this scientific data and a lot of for chance of quite personal information in a way, what might happen if it gets in the wrong hands, and what's done to prevent it? >> Well, okay. I mean I think there are interesting things to know for people who on academics. I'm not that interested in individuals at all. I mean, I don't, I don't want to know your home address and how old you are and, you know, how much you earn. Marketing companies would love that information, and I think that -- I wouldn't say that I believe they're the great danger but that is where the privatization comes in, right? You can give me your data and I won't read it. I mean, I'm not going to send you letters. But I'm interested in the, the -- I might, I mean, I might be interested in, in, in your age group, so that might be useful to me, but not as an individual. So, I, I don't know. I think, I think that's evolving. I think that an evolving issue. I worry that there will be a serious backflash against this free data. I don't think people use Twitter, for example, -- necessary thing about the factor is a broadcast medium, and I'm -- you know - I don't think much people would be that geo-located but I think that's still an issue. [ Silence ] >> All right. One there. >> Hi. Thank you for the lecture and my question's about -- when you mention stuff like open street map, how do you validate the information? Say someone spams the data or say someone voluntarily ruins the data in some sense, so this is a lot depending on data. Everything's dependant on data. >> Yeah. >> Someone spams it. How do you get rid of it? >> I get -- I'm, I'm not personally involved with the action stream [inaudible] community, so I don't have a very specific answer to that. I, I, I would say that things like the Wikipedia model which is kind of what? Paper- stream [inaudible] I guess. Are pretty successful. I mean, Wikipedia itself is [inaudible] successful. Yes, people spam it. There is a lot moderation on sensitive topics. I mean, I guess that's one way to do it. I don't know what the backups are. I mean, I've seen there are, there are backups in case -- if, if, if someone destroys [inaudible] it does revision capability. So, nothing else is probably -- it shouldn't work but it does seem to. >> [Silence] We have one down here. Also, if, if multiple people want to ask questions, then we can line people up and get microphones to you, so don't hesitate to raise your hand while somebody else is speaking. >> First, thanks for the lecture. That graph of population vs income was incredibly linear, and I wondered if there were more instances like that that would give you an incredibly solid idea of what a city is. Like we, we, we wouldn't have realized that without that being -- plus that it was so solid. Is there, is there an emerging model of what exactly a city is in that kind of -- would more... >> Yeah. >> ...data like that, that, that surprises you say, that, that tells you more about what these big metropolises that we've -- that have emerged are... >> Yeah. >> ...that we might not understand all -- my surprises? >> That's a really good question, and I think there's a lot to say about that, really. The first thing is, this is on a log, log scale and that basically means that things probably look a bit more linear than they actually are, so there's probably more scatter if you put this on a linear scale. It would look a bit more messy. Another thing to say is that when this work was done in America, when they did it in the States, so Jeff Weston and Louie Bettencourt did this. It wasn't linear. They found that on average the wages increased for larger populations. The per capita wages increased for each population. So, this is from the UK and it was actually surprising. This is a work in progress to try and understand why the UK and the US are different. So then, the question is, is they -- is there a common city or does it vary on, you know, the country, the geographical features, the density, US population -- because in the UK, the cities are less separated, perhaps, than in the States, and questions like that. So, and that they're, they're trying to [inaudible] questions [inaudible]. And it's a very aggregate picture. It doesn't give you much detail, so you might start to get some classifications, but you won't [squeaking sound] maybe get all of the essential pictures of a city. So it's hard to, it's hard to say whether you could typify a city just from this data. But it's suddenly really interesting the [inaudible] >> If I'm coming from the middle here. >> Hi. I was wondering whether you see similar patterns across different datasets. Do you see what I mean? So do you -- can you recognize patterns in say, a two of today's set compared to a transport one? I mean, do you do something... >> Yeah. >> ...to merge? >> Typically, you're looking for really different things. There are -- I suppose the pen, for example, would be like if you have like really, really finely grained Twitter data, like, like over long, long periods of time, very fine traces, [inaudlbe] traces, and you compared that with transport data. Then you'd expect to see some of the patterns. But a lot of the time you're taking really different approaches. Some of these are like -- you've got this aggregate approach. Some of them mean -- you just, you just want to know geographical concentrations and another one... >> Sure. But say you're looking at, say, London. >> Yeah. >> And you look at the Twitter across London. Then you look at -- I mean, are you -- are there shortcuts is what I, I'm trying -- have you found anything of -- that's in -- well, [inaudible]. >> I rarely go to shortcuts. Now some of the techniques... >> Yeah. >> ...the techniques sometimes are used. And then what you attempts to do is rather than saying this is the same as another dataset, you would look for correlations and you'd try to prove that they were the same. So you'd look at correlation of Twitter with demographic data or something like that, and say... -- and that's, that, that sort of thing has been done about [inaudible]. >> Okay. We have one here in the -- one, two, three, fourth row, right side of the room. We're we're also holding out hope for you. >> Excuse me. >> Hi. Is there any good examples of data being analyzed and then fed back and being used for city planning and shaping the future? >> Hmm. Yes. In a broader sense, yes. So, for example, I, I -- took my something I've worked on the bike data. I know that the, the recent extension that they're doing to the southwest of London, it's based on, on services data. So that's a few years old, but they're based on information about who, who uses bikes, who's likely to use bikes, and they've planned the, the stands based around that information. So, yeah, it does certainly happen. I mean, in terms of -- I think you need to ask a planner. I think planners probably will say that they have evidence-based approach as -- by -- I, I'm not a, a, a planner, so probably, yes, they've been doing it for years. But I can't think of anything specific examples. >> [Background noise] You said you were a physicist to start with. Do you see patterns in sort of the collections of human data which reflect physical phenomenon like Boltzmann Distributions or nucleic growth or that sort of thing? >> Yeah. I mean actually so that Professor Batty who's this perfamed of CASA, has done a lot of work on factual patterns in cities and has used models and simulations which collect chemistry and physics to look at -- they're great for the cities or morphology, the shape of the city. And the way things will aggregate and, and then grow out in, in a sort, in a sort of, you know, -- he mentions like a, like a [inaudible] branch type structure. So he's done a lot of work on that. And there is a lot of the mathematical techniques that we use and -- come from statistical mechanics, so, so that, that, the maps of, of, of actions in the gas, essentially, in large numbers, that kind of, that kind of physics. So, there are correlations [inaudible] I think it's dangerous to form too strong an analogy because they are totally different systems. >> We have two final questions coming from the right side. First there and then that gentleman there. >> This is on. >> Okay. >> In psychology, we have statistical models that help predict academic future success, so you can look at, say, G [inaudible] grades and then use that to perhaps you, you look at past data and say, oh, people got these grades and you can look at the correlation between those grades and university grades and you sort of use that as a predictor of future outcomes of -- you have six neuros to get particular grades that [inaudible] because they are these ones are probably more and [inaudible] in this group, etc, etc. From the data you were showing about journeys -- you mentioned about marketing. I was wondering about purchasing so, if I go out shopping and, for example, at Christmas I might buy more toys for presents and I'd buy more music. I remember when I was 16, I used to buy more music when I was at Christmastime for example. Now I think that's kind of being demonstrated to be as a psychological [inaudible] certain things we buy -- we don't buy a particular times of the year... >> Mm-hmm. >> By wave analogy I was wondering how if you could use this to touch -- look at how people act, interact in the city, how they travel, how they move around in the city, you know, modes of transport perhaps for the, you know, used the car, bus, travel -- underground, and whether you can see, see classes and whether you can make a prediction about few -- you know, possible short-term or medium-term behavior patterns. So, perhaps, on a particular day of the week, particular month, year, people used the bus more, and people... >> Yeah. >> ...to do certain types of shopping. I wonder if you could use that as a predictor of future events or [inaudible]. >> Yeah. I mean, mm, I, I'm very skeptical about term prediction, but, yeah, you can -- the kind of statistical prediction you're talking about is, yeah, of a group, right? Yeah, you can do that. I mean, essentially, you can kind of cross-pollinate any datasets you want if you have the, the matching information. You could cross-pollinate how much [inaudible] how much people spend and how much they earn based on their, you know, their loyalty card. You know that that their [inaudible] card or whatever vs their transport patterns, in theory. Whether you get a fixed approval to do that -- I'm not sure. Yeah, in theory, that you could do all sorts of things, but it's a very complex system, and I think prediction's quite a strong word. >> Okay. On that note I'm afraid we're all out of time. I'm very sorry, but I'm sure Martin would be willing to talk with you briefly, but we have to clear out of the room to let the next class in. So, that -- let's thank Dr. Austwick for his time. [Applause] >> Thank you. [Applause] Silence.