 Kitchen is a professor in Manus University Social Science Institute and Department of Geography. He was a European Research Council Advanced Investigator on the Programmable City Project and a Principal Investigator for the Building City Dashboards Project and the Digital Repository of Ireland. He is the co-author and co-editor of 31 academic books and co-author of over 200 articles and book chapters. He has been an editor of Dialogues in Human Geography, Progress in Human Geography and Social and Cultural Geography and was the co-editor in chief of the International Encyclopedia of Human Geography. He was a 2013 recipient of the Royal Irish Academy's Gold Medal for the Social Sciences. Dr. Kitchen's talk today is entitled The Epsistemology, Praxis and Politics of Urban Science and City Dashboards, which examines the conceptual underpinnings and practices of urban science and its application to the creation of city dashboards, informed by the building of Dublin and Cork dashboards. The research makes the case for a more critical framing and application of urban science that aligns with approaches adopted in critical GIS, radical statistics and feminist data science. So Professor Kitchen, if you're ready, I'll pass things over to you now. Okay, great. Hopefully you can all hear me fine. I'm just going to share the screen. Okay, can you just confirm that you can see that okay? Yes. Yeah, okay, great. Okay, yeah, so thanks very much for the invitation to speak to you today. As Celina has already said, I'm going to talk really around urban science and city dashboards and talk through the kind of approach that we've been taking to these and that given some examples of the work that we've been doing and some of our arguments around how we think about how we think about how we approach kind of urban data, urban studies and understanding cities really through, I guess, the kind of the application of data, data science in one form or another. So I'm going to start with talking about what do I mean by urban science. So generally, this is a kind of a computational approach to city systems and the processes of urbanization. So it's typically using urban big data and data analytics. So things like data mining, statistics, visual analytics, modeling, simulation, and so on to try and identify causal relationships and predict how city systems will work and to kind of come up with technical solutions to how we might fix some of the urban problems that we face today. Now, typically, a lot of the work in urban science has been going on for people coming from kind of data science, computer science, maths, physics, engineering, and so on, but it actually builds on a much longer history of quantitative social science. So the work of kind of quantitative geography going back into the late 1950s onto geographic information science really from the 1980s onwards. Again, urban modeling going back into the 50s and 60s, social physics, urban cybernetics going back to the late 1960s and 70s, social ecology, urban informatics, location theory, urban regional economics. So these have all been taking kind of quantitative statistical kind of mapping kinds of approaches to understanding the urban. What's kind of new, if you like, at the minute is the use of urban big data and data analytics and machine learning and so on. So the aim is to conduct extensive analysis of urban systems to try and determine urban laws and produce new theoretical insights, kind of develop a synoptic and integrative science of cities, and to translate that knowledge produced into a practical application. And it's typically contrasted with urban studies, which is kind of portrayed as conceiving as cities as a constellation of places, as opposed to systems of systems that uses more quantitative data, but also much more qualitative data, typically on the quantitative is typically on the small data, and it adopts a more kind of contextual approach with respect to politics, culture, policy, and history. So it's often grounded in those. So it's grounded within a kind of more kind of social philosophical perspective, and is trying to contextualize within kind of political economy, cultural economy, what's been happening with various forms of policy, governance, legislation, urban history, and so on. Now the epistemology of urban science is typically rooted in a positivistic tradition, not unsurprisingly given where it's coming from. So it's applying, it's trying to apply scientific principles and methods drawn from the natural hard and computer sciences to social phenomenon in order to explain them. So statistically testing relationships between variables, building models to produce and verify laws, they explain and predict how systems work, and to formulate theories which can be tested and verified. And it's typically used in a realist epistemology. So this is, you know, there's an existence of an external reality, which operates independently of an observer, and it can be objectively and accurately measured, tracked, analyzed, modeled, visualized to reveal the world as it actually is. So it's taking the data to be representative of what, of phenomena in a kind of essential, natural, given way, as opposed to a kind of constructed, produced, generated kind of way. And there's typically three epistemological variations that kind of track what's going on with data science more and traditional science more generally. So the first is a kind of traditional hypothesis driven deductive scientific method with the questions and approaches guided by established theory. You know, so we, you know, we come up with a hypothesis, which we then test with the data that we've got. There's kind of an inductive empiricism in which data analytics are seen to be able to enable the data to speak for themselves, free of theory or human bias or framing. So this is, if you're familiar with the debates around this, this is kind of the Chris Anderson version of science where you don't need a theory, you kind of throw various kind of analytics at the data and that will reveal the inherent truth of what the data has to say about the world. And the last is a form of kind of data driven science, which maintains the tennis of the scientific method, but it generates the hypotheses from the data rather than from theory. So you kind of do some kind of preliminary data analytics, data mining, pattern recognition, so on on the data to identify interesting kind of relationships and patterns within it. And then you use those to, to generate your hypotheses, as opposed to starting with a body of theory, working out hypothesis and then using the data to test it. Okay, so, so they're the kind of the various ways in which urban science is kind of being practised. So just to say something about the urban, the urban big data because this is obviously key to what's going on here in terms of opening up new ways of finding out information about the city and being able to kind of monitor and track that. So one of the key things I guess about science is, it's really interested in dynamic processes. So it's interested in change and practice and process and how cities are kind of evolving and transforming over time. And urban big data is very useful for this because it's providing streams of typically real-time data that provide a kind of a longitudinal view of what's going on. Whereas in urban studies typically a lot of data will be snapshots at a particular point in time, often on surveys or other forms of data capture where you get a kind of a time slice on a small amount of on a sampled basis. So big data is kind of different from small data in the sense that it has velocity, it's been generated in real time and that it's exhaustive. So it's a sample based on an entire population within the system. So say for example within Twitter it's every single tweet as opposed to a sample of tweets is potentially what's there to be analysed and it's the same with the urban big data. So you potentially have every single reading off of a sound sensor on a continual basis rather than a sample of a selection of readings at a particular point and so on. So you have this kind of continual longitudinal view and it's urban big data is now being generated across all kinds of domains within the city. So within government we carry e-government systems, we have online transactions, we have city operating systems, performance management systems and so on. When security and emergency services, digital surveillance, predictive policing, coordinated emergency response system, transport, intelligent transport systems, things like integrated ticketing, smart travel cards, bike share, real-time passenger information, logistics management and energy. We've got smart grid, smart meters, smart lighting and so on. Across the environment we have smart, we have sensor networks so pulling in readings about pollution, noise, weather, land movement, flood management and so on. Within buildings we can have a array of different sensors that are pulling information back into a building management system. Within our homes we might have smart meters, various forms of app-controlled appliances and so on. So basically algorithmic systems are being used to monitor and track what's happening within a particular system or domain and that's given us kind of volumous data sets that have good granularity. So this is the other distinction is that we have very strong granularity here so the data is tied to individual sensors or individual swipe cards or individual phones or individual cameras or whatever it might be. So that's providing a massive amount of fine-grained details, information about various urban systems. So this is like some of the data that we would have access to. This is actually all open data in Ireland that we pull into our dashboards. So this is just some of the real-time mobility data. So I appreciate that the table's probably got quite small writing so I'll just kind of give some selection of this. So it's on the left-hand side we've got things like the transport, public transport GPS location. So being able to track the location of buses or trains or trams and so on, or it might be the information that's going to real-time passenger information that's been displayed on the boards in bus stops and train stations. We have travel time along roads, so along road segments. We have inductive loop counters that are tracking the number of cars that are going over the loops. We have numbers of car parking spaces, so the cars going in and out of the car parks and how many spaces are free. We have flight location, flight arrivals and parches. We have maritime information about where the boats are. We have bike share information about how many bikes are in the bike stations and so on, and we have some CCTV feeds. And the temporal granularity of that data is this column here. It's a minute, every few seconds, every two minutes, every few seconds, two minutes less than a second, a minute, a minute, five minutes and so on. So this data is feeding in a near real-time basis and it's allowing us to track what's going on with transport across the city. And we have another set of environment data. Now nearly all the data that we have access to is just transport and environment data. We don't have access into all those other domains that we were talking about, and that's one of the issues around some of this data is access. A lot of the data is held by private vendors rather than governments, and even when government has it, they're often reluctant to make it open and to give an API onto this. One of our ongoing issues actually is with the APIs because they're constantly failing. And so we're constantly having to update the system to try and keep it working. And all of that data then is feeding back into these kinds of control rooms and being outpotted in various ways. So the one in the top left here is probably the most famous in terms of the literature and since what's been written is the control room in Rio. It was effectively built for the Olympics and the World Cup and the Confederations Cup as a way of kind of monitoring what was going on in the city and so on. It's pulling in data from 32 different agencies and 12 private companies. The private companies being mobility, mainly mobility companies. And so it's pulling in data. There's about 400 people working in that control room on a 24-7 basis and they obviously can kind of monitor what's happening and going on. There's a media section at the back where journalists can kind of sit and they can see what's kind of happening in terms of various incidents and so on. So they can also kind of source stories from what's happening. City workers in the field also have devices where they can ping data up to the centre and also pull data and analytics down. The top middle one is a control centre for a smart district in Japan and Tokyo and you can see some of the dashboard kind of analytics on the screen and on the computer monitors in front of the bottom left one is the control room in Dublin and what basically what an operator sits in there sees. The middle one here I think is in Britain. I think this is a control room to keep a single road operating. This is to keep the M25 orbital motorway around London and flowing basically and to stop it getting congested all the time and then some of this information then flows back out to the public through these kinds of dashboard kinds of interfaces allowing people to kind of see what's going on and then make decisions in their own lives based on that. So this is if you like urban science creating data driven urbanism and kind of managing and governing cities in real time. And so one of the ways that we've been interested in this is pushing this data out to the public and allowing other groups other than those people managing the city to see what is going on. So these are just a sample of different dashboards but there's a whole load of them now in different places. The work that we've been doing is on Dublin and Cork. So this is work from initially the one on the left just from the programmable city project and then the ones on the right from the building city dashboard project and they're quite different types of dashboards in terms of kind of learning from the research that we've been doing. So the one on the left is the first one that we built. This is just the top level and effectively what we did was just pull together all the data that we could find for the city and various kinds of data analytics that other people were producing and that we were producing ourselves and putting them in one place. There's actually 56 modules so if you go down, this is now disappeared but when it was there you could go down underneath each one of these and you would find various kinds of various kinds of modules that would allow you to look at the city in various ways and it would also let you actually get the data. It would let you feedback data to this Dublin report and allow you to upload information and data back to the city and so on. So and you can kind of get a sense from the titles, you know, how's Dublin doing because it's kind of indicators. Dublin near to me is kind of, you know, what information is around me, whereas the nearest GP pharmacy, whatever it might be. Some of the planning information, some of the housing, some of the public administration data, this is not real-time data, this is a traditional kind of public administration kinds of data and so on. And then over the last number of years we've kind of transferred from this to a more streamlined, easier to understand, I guess, kind of dashboard and we've done quite a lot of work on this doing kind of user requirements and testing and feedback with people with different constituencies. So the general public through to kind of policy makers and people who use the data through to kind of advanced advanced users who maybe want to do their own analytics and design their own way of engaging with the data. And so in the top right you can kind of see how we've reorganized the site. So we have a single kind of landing into the site, but then we have different routes into it depending on what kind of user you are. So we have kind of tools, tasks and stories. The stories are aimed at the general public. They don't need to know anything about analytics. They get the data and they also get a narrative that translates and explains what the data is showing. So it kind of does some interpretive work for them. The tasks are you trying to answer a particular kind of question and then the tools are for the kind of advanced user who wants to build their own kind of queries. Now the reason for doing this is what we discovered is that people actually have very low levels of data literacy and what was happening on this site was although there was a massive amount of data and a whole load of different kinds of ways of displaying the data and so on is people would land on the site and very quickly leave because they were kind of bamboozled by all these various kinds of tools and having to learn how to use them and making sense of the data and not necessarily having the knowledge base of which to do that and so on. And what it made it very clear to us was if we actually want dashboards that are going to be used by the general public then we have to find a different way of displaying that data and making it accessible to them. So that's some of the work that we've been kind of trying to do at this portion of the project. We've also been looking at how to push the data into 3D and a different type of environment in which to explore the data. So I guess a kind of environment would be more familiar to people. So this isn't a kind of a 2D map base. This is where you can actually see the buildings and the roads and you can navigate them. So you can orientate yourself within the environment to an easier degree and you can also display the data in a different kind of way so you can take advantage of the third dimension to kind of do 3D visual analytics. So this is just some of the ways we've been playing with this. I should actually update this because some of these images are quite old now. So the top left up here is actually kind of Airbnb data. So I think this is bike stations. This is sound. The bottom left is kind of sound. The bottom middle is land use classification and so on. We've been doing this in VR. So this is kind of immersive AR where you can kind of see the real world at the same time as you can see the data and with WebGL. So this is kind of desktop kind of game space kinds of kind of view. The other kind of way that we've been looking at doing is using 3D printed models. So this is in the middle. I hope you can see it's a kind of a 3D printed model of the central Dublin. It's kind of seven kilometers by four kilometers in size. It's about 59,000 buildings on it. And then we're kind of hanging in a data projector over the top and projecting the data down onto the model. And it kind of allows people to kind of walk around the model and talk about the model and kind of do maybe more kind of collaborative planning kinds of conversations than the kind of individual immersive VR kind of experience. And then we've been projecting different types of data onto that model and looking at trying to create data stories that might lead people through. The model on the bottom right is the same model for Cork. The model is about four meters by two and a half meters in size. It takes up quite a lot of floor space. And it takes quite a lot of effort to align, given it's just a single projector going across that space to actually get the data to align directly onto the buildings is actually a little bit of a challenge, but is possible. So even playing around with those, now we were going to do the big public exhibitions and use the feedback on playing around with all this data. But of course, COVID kind of killed any of that kind of work happening. So we're still waiting to actually show this to the public and try and get feedback from them as to how they think about them and how well they work at communicating kinds of information about the city. Okay, so to think about these dashboards, what we've been trying to do is approach the dashboards in a kind of a critical kind of way and to kind of have a think about what they're actually doing and the kind of science behind them. So if I just start by giving a sense of what I mean by a dashboard, so for us, a city dashboard is a way of kind of being able to organize the data and allow interaction with the data. It's a kind of a cognitive tool that improves a kind of user span of control over what's actually quite a large set of volumous, varied and can be quickly transitioning data. Obviously the real-time data on our thing is updating as the data updates. So there's actually an issue of being able to cognitively track that. It kind of enables a user to explore the characteristics and structure of data sets and to interpret trends and the kind of the power and the utility and why cities are interested in them is their kind of claims to be able to show in detail and often in real time the state of playing cities that they kind of translate the messiness and complexities of cities into rational, detailed, systematic audit forms of knowledge and that they enable us to kind of know the cities that actually is through this objective trustworthy factual data. So this isn't data based necessarily on opinions or sentiments and so on, although it can be if we were to include social media data. It's data from sensors and from infrastructure and from systems and so on. And what we've been trying to do is critically reflect on this. So in our work making the dashboards, we're also kind of critically reflecting on kind of what's happening and we're doing that in relation to kind of six issues. The first is in relation to epistemology. The second is around scope and access. The third is around veracity and validity. The third is around usability and literacy, use and utility and ethics. And what I'm going to do is go through each of those and kind of pose them as six questions designed to expose the kind of the politics and proxies of city dashboards. And it can be used as a kind of heuristic to look at any other kind of urban technology as well as a kind of six questions kind of think about any other kind of urban technology system used to manage government cities. So if we translate the epistemology into a question about how our insights and value derived from city dashboards, we can kind of think about the kind of epistemological approach to how they kind of work. So these dashboards are relying on visual analytics. They're adopting a realist epistemology. They're typically posing urban data as being kind of neutral, value free, objective and essential in nature that they're representative and they show the world as it is. And when they're analyzed in a similarly objective way, they reveal the truth about cities. But from a critical data studies perspective, data are produced, they're not collected. So data are kind of constructed, they're generated, there's all kinds of decisions made about how we collect the data, how we process it, how we handle it, how we store it, how we share it. There's all kinds of issues around how we configure the various tools that we use. Even things like a sensor which seems like it would be quite objective, there's all kinds of decisions made about where to place the sensor. So if you place the sensor in the middle of a park or you place it next to a busy junction, you're making a kind of a political decision about what you're measuring and how you're measuring it. There's all kinds of decisions about how you calibrate the sensor and so on. These are full of politics and full of praxis in terms of how they're configured. They're not neutral, value free and objective, they're constructed. And they constitute a kind of a socio-technical assemblage which I'll show you in a second. And the dashboards and sounds are not simply mirrors of the world, they don't just simply show you what's going on, they kind of act as translators and as kind of engines. So they make things happen, like decisions are based on what they display, choices are made on what they display and so on. So they don't just display the world, they actively do work in the world. And through that work they change the world. So they're changing what you are measuring. So there is that kind of recursive iterative loop going on. They're quite reductive, so they kind of atomize what are very complex contingent relationships. So they pull out one measure or a couple of measures of things that are really quite complicated. So if you're looking at unemployment or something, you might get one or two single measures on employment, but unemployment is a kind of a multi-dimensional issue and has loads of different dimensions and there's loads of things that affect why people are unemployed, why people are unemployed, how people get jobs and so on. And what you're getting is a very surface kind of analysis which is decontextualized from the city. So what you get is these kinds of kind of analytics on the right hand side without any history of, you know, so this one is average house price and rent and so on. You get that without any history of housing within the city or any of the politics or policy around housing or any of the kinds of ongoing issues that are going on within that space. So you're just, you're given a very surface level kind of presentation but not the material necessarily to interpret that information in any kind of deep or meaningful kind of way. Just to kind of illustrate this notion of a kind of a digital socio-technical assemblage, your set of sensors or your infrastructure and so on, you can kind of think of them as having two kinds of stacks. The first one is a kind of a system process as this is kind of performing a stack and you know you at the bottom of the stack you have your infrastructure and your hardware, you have your operating system, you have your databases, you have your software, you have your interfaces and you have your kind of users and usage and so on. And decisions at any of these points changes what days are collected, how it's analyzed, how it's processed, what's displayed, how the algorithms work in terms of what they what they produce and so on. And all of that is shaped by a context, so a kind of a context stack that frames the system and tasks. So systems of thought, forms of knowledge, finance and political economies, governmentalities and legal requirements, how organizations and institutes work, subjectivities, communities, how marketplace works and so on. So even if we were just think about finance, a dashboard funded through philanthropy money is very different to a dashboard funded through venture capital in terms of what kinds of pressures are going to come onto the development team in terms of what's happening with the data. The philanthropy one is probably not looking to monetize the dashboard or the data in any kind of way, but the venture capital one will be and that will influence what kind of dashboard is produced, what kind of data is in it, how it's kind of presented and what what actually happens to the data and so on. And then there are various kinds of critical kind of social science technology studies that actually focus on different parts of this stack, you know, so platform studies, critical data studies, software studies, HCI and so on, all kind of making sense of the kind of the politics and praxis around this kind of this kind of assemblage. And so the kind of this is kind of led back into critiques around dashboards and critiques around urban science in general, that the approach is kind of it's reductionist, it's mechanistic, it's essentialist and just and it's deterministic is collapsing in a diverse, individual and complex, multi-dimensional social structures and relationships into abstract data points and kind of universal formulae and laws. And you know, and kind of overly simplifying all the kind of complications and complexities and wicked problems that actually exist in cities, all the kind of competing values and principles and objectives of different stakeholders and and so on. You know, critique again is like the data is an objective, neutral and value free, but it's kind of framed and situated in a kind of power geometries of knowledge and praxis. A lot of these dashboards willfully ignore metaphysical aspects of human life. So, you know, metaphysical aspects will be things like values, opinions, beliefs and so on. These are measuring typically measuring kind of factual information as opposed to urban lived experience and kind of sentiment and so on. And they're kind of ignoring all the kind of politics, ideology, social structures, capital, political economy, culture and so on that actually make cities what cities are. You know, if you're just going to measure the factual aspects of cities, then you miss out all of this stuff, then you're going to get a very surface level notion of what makes cities function. And the solutions you produce will not basically work because they're ignoring the things that make problems so intractable and difficult to deal with. So, promoting a kind of instrumental rationality, the posits that cities can be effectively steered and managed through scientific insights and technical instruments and that urban issues can be solved by technical solutions, you know, as opposed to other kinds of solutions such as political ones, civil society, fiscal, policy, legal interventions and so on, you know, rather than technical fixes. And in reality, urban science can contribute to solving problems, but it does so by working in collaboration with those other kinds of fixes. Like you're not going to fix congestion purely through a technical fix. It also needs social interventions around encouraging people to walk on to cycle, investments in public transit and so on. It's not going to be solved simply by optimizing the flow, trying to optimize the flow of traffic on the network. And so it kind of, you know, the critiques it produces are kind of limited and limited limiting understanding how cities work. And it also forecloses what kinds of questions can be asked and how they can be answered. Because it has a particular way of which it kind of looks at and thinks about the world. It closes off other ways of kind of thinking about the world. And I think part of that is to do, and this is a kind of critique that not just comes from kind of me, it's also coming from some people within urban science who have that longer history within, say, quantitative geography or quantitative planning and so on. That there's a kind of constituency that's maybe coming from data science or maths or computer science and so on, that just don't have the background deep knowledge of the kind of history of urban policy and planning and so on. So if you're coming at this from data science and you've never studied urban geography, urban studies, urban history, urban governance, urban policy, any of those kinds of stuff, then you're lacking critical domain knowledge that will help you make sense. So it's important that kind of urban science is done in a kind of interdisciplinary background where those kinds of knowledges can be added to the mix if people don't have that. I think that's a valid critique being made by people like Mike Batty, who's actually one of the leading lights in urban science. Then there's kind of questions around how kind of comprehensive and open are the dashboards. So these are questions about access and obviously access is important because it's also a question about power, about who has access to this kind of information and who can use it and what can they do with it and how that might change how the city is managed. So dashboards are kind of displaying quantitative data. There's a huge amount of information absent. So within a dashboard you might only be measuring 30 to 100 indicators and the city is massively diverse across every kind of domain you can think of. I've already said it's ignoring the metaphysical aspects and the kind of intangible parts of urban life, the significant gaps and sciences in the data displayed. A lot of the dashboards within those control rooms are limited to those control rooms. They're not available to the public and so on. There's various levels of openness in relation to administrative data and we've had ongoing issues around access. There's a certain amount of data made open and everything else has been these complicated negotiations to kind of leverage data out of institutions and then even when we do get the data then there's a whole series of other issues. I mean relating to data measurement, into data formats and media, into metadata, into data standards, into ways of sharing and so on and pretty much all the data that we host in the Dublin dashboard and in the Cork dashboard we can tell you nothing about in relation to things like error and bias and calibration, representativeness and so on. Now we've done a bit of work trying to look at that and I'll come on to that in a minute but it's difficult. The second is just assembling the data in places where there's a fracture of political geography. So in some cities you have very fractured data landscapes with respect to geography, so the kind of scalar organization between kind of local, county, regional, state and federal. You have kind of back-to-back services and planning happening across districts and you have a kind of a mismatch of functional territories versus administrative geographies. So the maps on the right kind of relate to Boston. The issue also relates to Dublin, so in Dublin there are four local authorities make up the city, they're all completely autonomous and there's no regional body over the top of them. So if you think about smart city technology it's quite possible for each of those local authorities to have a completely different bike share scheme and the bikes to be not interoperable going between. Now that's not the case but it's entirely possible that that could be the case which would mean four different data systems. This is Boston. Now Boston is interesting because Boston is made up of 101 towns and cities all of which are autonomous. So you have 101 planning departments, you have 101 school districts, you have 101 police departments, you have 100 you have 101 everything basically except for some infrastructure which is shared so some of the transport system, some of the water and sewage systems are shared between some of these towns and cities and so on but it also means you have 101 open data sites if they have open data sites and there's very little collaboration or sharing going on between these towns and cities. So what we have here on the top right is actually the city of Boston and you can see it's kind of got this loop out here and this is Brookline, Cambridge, Somerville and so on and the data for the city of Boston stops at the edge of Boston as does the evacuation plan from the city of Boston. It doesn't go all the way to the edge in the municipality once you get to Cambridge you're on your own. So trying to do a dashboard for metropolitan Boston is actually almost impossible in fact it is impossible and so that's a kind of this kind of fracturing effect is kind of interesting and then we also have that in respect to stakeholders so within municipalities and across municipalities you know you've got public sector agencies, you've got industry, you've got universities, you've got NGOs, you've got community organisers, you have to persuade all of them to share their data with you and so on and they all have different goals, resources, practices, institutional structures, funding models and interestingly in the Boston case they also have different data ontologies so how they collect and generate their data and categorise their data can be quite difficult so even if you had access to the data merging them together is not in any way straightforward. We have the same effect in Ireland between Northern Ireland and the Republic of Ireland it's extremely difficult to marry the data sets across that political border together to create all Ireland data sets and then there are questions around to what extent we can kind of trust the dashboards so this is questions around veracity and validity so this is questions around data quality as all kinds of issues here around human error, bias, abstraction, representation, generalisation so what you said the kind of the technical instruments have various specifications, various parameters, handled procedures, different scientific norms and standards and so on and a lot of the data is published without things relating to measurement, sampling frame, handling, veracity, uncertainty, error, bias, reliability and so on so we've done quite a bit of work on this so to go back to the public, the transport data that I showed you earlier on we've tried to assess that data in relation to kind of veracity across kind of the source, the data and the metadata so over here we kind of have sustainability so like do we think this data set will still be in existence in two years time, five years time, 10 years time, 50 years so on? How transparent or interpretable is it? Issues around privacy, issues around fidelity, cleanliness, completeness, coherence, metadata changes over time, how standardised the data is, whether there's good methodological transparency i.e. we know exactly how the data was generated, how it's been processed and handled and it's kind of relevance and the way we've done this is basically to systematically play around with the data and to go and interview the people who produce the data to go and go and talk to them and you can see we've kind of got good, fair, poor and not ready and there's very few of the datasets here that have good all the way across so there's some issue somewhere with the data that's not to say the data isn't usable and it doesn't produce useful insights it just means that there are some doubts about the data in some respects the other kind of issues around things like the appropriateness of the method and how the data is displayed so the kind of validity of the analysis and interpretation a lot of this stuff is black boxed you don't know how the calculations have been produced you just get the graph or the outputs from various kinds of analytics and then how the data is actually displayed in terms of the visual analytics you can get ecological fallacy effects so this is the map on the right is an example of this, this is exactly the same data presented at different scales the top the top scale is electoral districts the middle one is enumerator areas and the bottom one is what's called small areas so these are three different statistical geographies it's unemployment and red is over 40 percent unemployed and it looks on this map that the bottom the bottom part of the map is red at the top it's red in the middle and we can see some other kinds of changes going on some kind of orange appearing over on the right so and on the bottom when we get down to small areas so this is areas of 80 to 100 households this is 300 to 400 households and this is three to 4 000 households and really unemployment is really in this narrow strip that really bad areas and this area that was in red on the top here is in blue and dark blue is less than five percent unemployment now if you're doing a targeted area initiative if you're looking at this map you put resources in this area if you look at this map you put no resources in this area it's identical data it's just been displayed in a different way so how you display the data can can lead people to have a completely different interpretation of what's happening in the city and you can do exactly the same we've been looking at kind of this is modified by the area this is a spatial one we've also been looking at temporal so how you divide the data so if you divide the data and display it in kind of minute in 10 minutes in an hour in days and weeks and months does that change what's shown and it does basically and then there's questions around usability and literacy I'm going to skip over this it's basically that people find it quite difficult if you're not trained to interpret the information that's being displayed and that obviously has implications because how people interpret the data makes it makes a difference to how they act in relation to that data then there's questions around how how these are used and their utility so they are they used in a very strong direct way in terms of managing performance and so on or they're used in a contextual way to kind of inform thinking and policy in in US cities they're often quite used in a very direct managerial way whereas in European cities they're used in a contextual way alongside lots of other information so in some some US cities you know you have these purpose-built dashboard rooms this is one from Baltimore they hold a weekly dashboard meeting where every head of a service department has shown their data from a week before and asked to account for it and so on so this is a very direct performance management style of urban of urban management and so on so there's a whole politics around around that and what it what it means to manage your city off of metrics and key performance indicators and then there are another set of questions here around kind of ethics and so you know can can we assure that we can kind of use open science and dashboards in a way that's that's ethical so we have very fine-grained information on a kind of fine-grained longitudinal basis so we have individual level data on a longitudinal basis this raises all kinds of questions around kind of ethics and kind of ownership from control of the data how the data might end up in data markets how it might feed into surveillance capitalism how it might lead into social sorting redlining geodemographics how it's leading into targeted area initiatives and where finances invested or disinvested from parts of the city how some of the data might end up in predictive profiling into anticipatory forms of governance so this is kind of pre-emptive emergency management or in predictive policing and so on how the data might be used in other forms of governance so things like nudge so trying to nudge people's behavior to act in a in a different in a different way questions around data security and cyber security and the ability to hack into these systems and disrupt them but also the data to be stolen from them and then issues around control creeps so this is a system designed to do one thing starting to be used for another so an example of that would be things like the congestion charge cameras in London were were installed on the basis that they would only ever be used for for that purpose but now of course are used for for regular policing and security or the or the the Trace Together app from COVID in Singapore that was originally designed just to track movement for for COVID basis but a few months ago the data from that app was used in a in a murder trial to kind of illustrate where where the person was on a particular day at a particular time so data data collected for one purpose was starting to be used for for another for another purpose and this kind of raises kinds of questions around kind of technical procedural ways of thinking about this data versus kind of normative ideological ways of thinking about this data and I'm not really going to go into this but there's some really interesting work kind of going on within data feminism data justice and so on comparing how we try to regulate technology which is often to kind of locate the source of problems within individuals and technical systems so thinking about ethics bias consumer rights fairness accountability transparency and so on as opposed to kind of rooting them in kind of structural power and how society is configured so thinking about them in relation to justice oppression citizenship equity and so on and people people like Kate Rollinsen and Rachel Franklin have been doing some interesting work around sensor networks and urban sensor networks and moving from kind of data ethics to data justice and thinking about the justice kinds of issues around where those sensors are located what data they're producing who they're acting for and how they affect what's going on within a particular area and there probably needs to be more work done in urban science data science uh you know looking at at these two sides of the call of this column like the left-hand side is typically how we're looking at this and it's about compliance whereas this side is about kind of rethinking kind of social relations and political economy and and so on okay just so just to conclude thanks I'm kind of aware of the time um my my basic kind of argument really is there is a kind of a need to kind of reflect on the epistemology and principles and ethos of urban science and hopefully hopefully it's come across that I'm not anti urban science I mean obviously practicing some of its methods and techniques and so on and what I'm suggesting is a kind of a rethinking of the kind of the epistemology and ideology around um around kind of urban science I'm really learning from debates that are taking place in earlier decades in relation to GIS and statistics and quantitative geography sociology and so on so kind of learning from uh the critical GIS radical statistics and kind of feminist data science and really thinking around the politics and practice of what it is that we're doing when we do urban science and so all of those still employ quantitative techniques, inferential statistics, modeling, simulation, visual analytics but they're kind of mindful and open of their shortcomings and and they kind of are very clear around their kind of positionality and situatedness of the researcher and the kind of the power that's good that that is being exerted through them you know they're taking seriously questions around data veracity, data politics, algorithmic power, data ethics and justice and so on and they're trying to make the analysis much more kind of contextual in relation to kind of history, culture and domain knowledge and so this is kind of a it's kind of a a different kind of version if you like of urban science it's one that's much more kind of uh drawing on ideas from kind of critical social theory to kind of rethink how how research is conducted and how we make sense of the findings and how we employ them employ the knowledge and it's much more kind of aware around the kind of how the city is contested how it's messy and contingent and that the and the solutions to the problems facing urban societies needs to be more than technical solutionism you know so so urban science might be part of the answer but it's not the answer it's part of an answer with a with a set of other related kind of knowledges and expertise and so on so I'm going to I'm going to end at that point if you if you're interested in any of this kind of stuff there's a there's a few papers on the slides I can I can make the slides available and you can you can kind of follow up on on some of those kinds of analysis and I'll end there um thank you very much Dr. Kitchen for your talk I would like now to open up the session for questions as a reminder again to ask questions participants on zoom are encouraged to use the raise your hand button and I'll call you to my unmute and you can ask your question directly or you may also type your questions in the chat box and I can read them out and for audience in every one 14 you can raise your hand directly and I'll call you so you can ask your question okay we have a question in sorry um thank you so much Dr. Kitchen I really enjoyed going to say I have a particular interest in your discussion on data accessibility and literacy you named the couple really interesting methods in trying to improve literacy in these dashboards I was just wondering in your experience what has proven most effective by any standard of your metrics yeah I mean it's a shame that we haven't been able to do more more user testing afterwards I mean COVID just basically killed you know really limited our ability to be able to be able to do this and our initial feedback is really around making the data and its interpretation accessible to the right level so um so for members of the public the data stories approach is really about it's almost like data journalism it's about having the analytics and the story that goes with it and and you're kind of telling a story say we have an example around housing where we kind of tell the story of housing in the city illustrated with these analytics that you can then you can query so all of our all of our graphs and so on you can kind of you know you hover over them and bits of information pop up and and so on but the text kind of tells you how to interpret it now for the policy makers that's that's useful but they they normally want to make their own interpretation and so on and they want to be able to kind of query and ask questions of the data we're not really in the data stories we're not really trying to enable people to ask questions we're just trying to tell them a story on the relation to policymakers we're trying to allow them to kind of answer ask the questions that they that they want to that they want to pose and to get some of the some of the information back whereas our kind of power users our our kind of people who have a lot of background in data science or analytics they work in GIS or they work in you know a kind of a analytics unit within government or a company or something but they want the utility to be able to kind of do their own analytics and to and to you know kind of change how they query query the data and so on so it's really about trying to tailor what you're doing to different types of audiences and my impression of most city dashboards is is that they don't do user requirements they just simply put the dashboard together and then present it to you know to the to the to a public which is seemingly universal and uniform as opposed to having these kind of diverse kind of learning styles and literacies and so on so I you know I think it's just a you know if we're really going to do kind of public facing kind of geographic data science urban science then we have to work out a way where we can communicate to at different levels and we're really not there at the minute. Thank you and I think we have a question in the chat box from Wednesday so thanks Rob for the presentation I wonder if you could give some examples of what you think are situated contextual studies or projects in urban science and I wonder if there is a way of formalizing methods towards this direction. Oh god example oh no of course I'm not going to think of an example and I'm trying to think of one. I mean some of the work that there's a there's a paper by kind of Jeremy Crampton and co called Beyond the Geotag and in that paper they they they're kind of doing their kind of Twitter analytics and they're doing analytics on all the data and so on but they're then trying to place the data within historical kind of context and policy kind of context and they they make a case that that data has got to be complemented with kind of more localized studies more contextually placed studies that you know you can't you can't have a kind of a one-size-fits-all kind of analytics for every city but you have to kind of be cognizant of so I think in our city it might be about Louisville or something and they're kind of saying look if you don't if you don't take account of the racial politics of that city you're really not going to understand what's going on there if you just take the service level data you know you really have to place it into just longer history of deindustrialization very strong structural spatial divides within city persistent racial inequality and so on and you know and if you don't if you don't do that you're really just going to give this kind of surface level kinds of analysis and that's really the problem with some of these dashboards is that they are surface level they're just giving you kind of a kind of a general impression about what's kind of going on you know and it doesn't really matter what the domain is whether it's education or welfare or transport or health or whatever it might be if you if you really want to kind of understand you know if you really want to make sense of what's going going on you have to link you know like the idea that you're going to be able to say something sensible about housing policy while ignoring the last 50 years of previous housing policy would make no sense you know like in terms of you know what work what didn't work why it failed what kind of politics were going on what were the competing this competing ideas and so on and that if you want to solve the problems then you're going to need a kind of a multi-pronged approach onto that you know that it's going to need you know it's going to need different types of investments and different types of solutions whether that's community legal fiscal policy politics whatever it whatever it might be um I guess my try my try answer on some of this stuff is around you you know you're not going to fix you're not going to fix homelessness with an app you know like it's a it there's no technical solutionist way of solving homelessness you know it's a problem of deep structural inequality mental illness domestic violence all kinds of various things the political economy of the city the housing policy the city how rental markets work how finance capitalism works you know how financialization and so on like that's you you need a multi-pronged way into that uh an app an app isn't going to do it on its own thank you we have another question from the audience on zoom uh this is from Renjani so as you pointed out there are both benefits and shortcomings of using urban science in policymaking what measures are cities or governing bodies taking measures to address these critiques of urban science and I think the main one is is that is that is that a lot of people working within cities are still cognizant that they need that they need this multi-pronged approach and that and that you know things like politicians are still reliant on on people to vote for them so there's always going to be politics going on there there's always going to be uh kind of competing interests going on different constituencies and so on I mean some of that's ad hoc and uh happenstance and so on so even within things like municipalities you've got different constituencies who have different uh aims and objectives and and and uh will block uh different kinds of things going on or or oppose them and so on so a lot of it is just the fact that cities are kind of complex social systems in and of themselves will will uh limit will limit kind of urban science being the only approach into understanding understanding cities in terms of like um some of the kind of instrumental ways in which they're working on cities you know so things like implementing uh predictive policing or various kinds of control rooms and and so on uh you know so the kind of smart city technology I mean there's various pushbacks going on against that out of civil society out of NGOs out of out of political parties you know all kinds of people are kind of pushing back on bits of those and um and you know interestingly some cities are actually uh decommissioning so there's a there's a number of cities that used to have predictive policing programs that no longer do they've rolled back uh from from that approach principally because they've they've decided it doesn't work um and it doesn't it doesn't do what they thought that it that it would do and it's you know it's very costly way of um of kind of reorganizing policing uh services and so on um so yeah I you know I do get worried a bit around things like the data quality I don't I don't find people taking enough notice around uh the actual data that they're using and so on they test it I'm always surprised at how uncritically people will take data and use it without really uh kind of looking into what it is that they've actually got um so yeah I don't I'm sure I really answered your question um but yeah I I'm not too I'm not too worried in the sense of like like urban science isn't taking over the whole thing it's this like one approach amongst many um and if people don't like the answers it gives then they'll push back against it in any case thank you um do we have any more questions from the audience yes thank you so much for your talk it's been really interesting and I was just wondering um I was really interested in the um when you're talking about uh the idea of like there's uh spaces of silence and there's things that data analytic like the data um in these dashboards is not actually um showing us and I was wondering if you experimented at all with ways to point to those silences or point to those blind spots um on the dashboards themselves if there's been any like particularly effective ways to communicate to users what is not being shown uh yeah it's a great question no we haven't we haven't done that actually uh and it's probably something that we should do in in a way it's illustrated purely by the fact that there are so few datasets on you know so within within any say say on some of our kpi indicators you might only have three for education four for health five for transport or whatever I mean it's quite obvious in some ways that they're that they're very limited selected choices um but we don't actually directly point out all the things that are not measured I mean we we did some work I'm not not sure if you're familiar with this world world city council data projects which is uh which is a kind of a global project around the ISO 37023 or whatever it is there's a there's a new ISO standard for city indicators um and each city can apply to become ISO compliance around its indicators now we we did the work for Dublin to look at whether Dublin could apply to become ISO compliance around its city indicators now to become a to be fully compliant you there's a hundred indicators and you need you need a certain percentage uh in Dublin we could we could comply with 11 you know that there were only 11 that would meet the ISO requirements and the principal reason for that is is that the data for Dublin is actually not at the city scale it's at the regional scale so it includes a lot of the hinterland that's outside of the city um so it's not basically the spatial scale is the wrong scale and that's kind of ever in our in our dashboard what we're trying to do is do data that's at a kind of neighborhood scale and so on and we just we have very little of it in fact most of the data we have at that scale is from the census which is every five years so we have a snapshot data as opposed to having continuous data and even things with like some of the real-time data that we've got they're really for small parts of the city so the bike share scheme is really just the center of the city the sound sensor network is just again just really one one municipality out of the four and so on so the kind of absences are kind of are kind of evident if you know the city and you're looking at the data but you're right it probably would be a good idea to actually to actually say it would be really great to have these kinds of data and these are kinds of data that are kind of missing I do get worried a little bit around the ways of which people try to find surrogates or proxies to fill in some of those gaps so like using social media data as a proxy for kind of value values opinions beliefs and so on there's there's huge issues with that you know and not least of the data just not representative you know like if you're scraping off a twitter you know four percent of people over the age of 65 have a twitter account like it's it's not a representative of the whole population the views that you're that you're getting and it's the same whichever platform that you're that you're using so it can't be a replacement for other kinds of civic engagement to find out people's views about the city it can only be a supplement to it. Okay thank you so I think we're at time so on behalf of GSAP and the urban planning program in particular I'd like to thank you again Dr Kitchen for your great presentation today we really appreciate you taking the time to share your work with us and also thanks to everyone who attended both in person on zoom please make sure to join us next week at the same time for our next lecture by professor brandy thompson summers whose talk will be on spatial temporalities the future path of black dispossession thank you