Abstract

This paper presents the results of an experiment in using a big and granular dataset from Facebook to invent a new way of measuring political diversity in Copenhagen. The experiment was a collaboration between TANTLab and GEHL architects, and it was sparked by a shared concern that our cities are becoming political filter bubbles. Its outcome was the publication of an interactive datascape that enables the user to explore which urban spaces in Copenhagen break such bubbles. The paper discusses how this datascape reconfigured existing problematizations of urban diversity by problematizing existing assumptions about what urban diversity is and how to design for it. It unfolds how we produced the datascape, how the reception of this alternative cartography sparked a new debate about this issue, and how our findings relate to other literatures on urban diversity. The production of the datascape is an example of how scholars involved in data-driven urbanism can productively engage themselves in the production of critical metrics and imaginative cartographies rather than just using new data sources and algorithms in a framework of prediction and control. This is an important task because urban metrics and cartographic techniques ultimately shape the way we envision our cities.

Keywords: Soft City Sensing, urban cartography, political diversity, digital methods, critical metrics

Introduction

Urban metrics and cartographic techniques matter to the way we envision our cities. The empirical tools at our disposal guide the way we frame questions about them and ultimately how they can be steered and controlled (see, e.g., Wood and Fels 1992; Scott 2020; Kurgan and Brawley 2019). This is also why GEHL architects—the partner in this study—work under the motto “measure what you care about.” GEHL is built around a concern for the human aspects of the city, and its mission is to design measures and visualizations that put these aspects on the agenda of urban decision makers. The underlying thought is that your best bet for making your concerns and interests matter to others is to make them tangible, to give it a form through which it can be taken into account, valued, and shared. Designing metrics and maps in one powerful way of achieving this (Latour 1986; Kornberger et al. 2015).

This paper presents an experiment conducted in collaboration between computational humanities scholars at TANTLab and the innovation team at GEHL architects.1 Sparked by a concern that our cities are becoming political filter bubbles—and an interest in identifying where urban spaces break such bubbles—we set ourselves the task of creating an interactive cartography of political diversity in Copenhagen. As public voting data proved too spatially aggregated to give us relevant insights, we decided to repurpose a granular data set from Facebook that enabled us to unearth formerly unseen dimensions of urban diversity. We named the project “Do You Live in a Bubble” (DYLIAB), and the images in Figure 1 are static snapshots of the datascape we published for the public to explore. The full interactive version can be found here: https://tantlab.gehlpeople.com/

This paper describes how this datascape reconfigured existing problematizations of urban diversity—for both local practitioners of urban design and the publics interested in this specific issue. Whereas cartographies of urban diversity usually depict the sociodemographic traits of residents in predefined neighborhoods (see, e.g., Tam Cho ot al. 2013; Gimpel and Hui 2015), the digital traces at the heart of our map failed to provide such insights. Facebook data carry no metadata about the residential status of the data subjects. However, because it is filled with relational and granular traces of people’s interaction with content and events, it made it possible for us to detect patterns in the way 300,000 politically interested Facebook users interacted with so-called “third places”(Oldenburg 2013) in Copenhagen. Throughout the paper, I unfold how we produced the datascape, how the reception of this alternative cartography of diversity sparked a new debate about this issue, and how our findings relate to other literatures on urban diversity. I end the paper by suggesting that DYLIAB is an example of how scholars involved in data-driven urbanism can productively engage themselves in the production of critical metrics and imaginative cartographies rather than just using new data sources and algorithms in a framework of prediction and control.

Critical metrics and imaginative mapping

I use the concept critical metrics as a reference to practices of measurement that prompt a need to revisit the assumptions underpinning established modes of measuring, practices that expand or shift the way experts and the public can pose empirical questions about specific urban problems. Following Dewey (1929), I do not think of quantification as providing more objective representations of the world than what is possible with qualitative labels. Rather, practices of quantification are interesting because they involve acts of formalization that stimulate concrete discussions about the fundamental assumptions behind measures and representations. Echoing Peirce (2020), I want to argue that collaborations around operationalizations is a productive way to explicate the normative commitments that often underpin knowledge production.

This take on quantification is not dissimilar to the way maps and cartographies are theorized by scholars within the field of critical cartography, where it has been suggested that the critical potential in new conventions of mapping is not their potential to objectively represent the world. Rather, it is their ability to surface hidden assumptions in previous mapping conventions: “A critique is not a project of finding fault but an examination of the assumptions of a field of knowledge. Its purpose is to understand and suggest alternatives to the categories of knowledge that we use. Critique does not seek to escape from categories but rather to show how they came to be, and what other possibilities there are” (Crampton and Krygier 2005, 13).

This definition of cartographic critique is coupled with the diagnosis that digitization has “undisciplined” the practice of mapmaking. The fact that the digital offers new types of data and mapping techniques open for a potential to practice critical cartography in new ways. By embracing new digital opportunities, critical cartographers can intervene in the production of topical maps and propose alternatives to existing cartographic practices. This is what Crampton and Krygier (2005) term “imaginative mapping,” and it is not a new phenomenon. For instance, Snow’s early epidemiological maps shifted public perception of the contagion of cholera from being about the flow of air to being about the infrastructure of water (Vaughan 2018). Similarly, Booth’s poverty maps broke the public imagination that London was divided in two by showing how pockets of poverty existed in both the east and the west (Vaughan 2018). Recently, we have seen the development of the smart city framework being intimately tied to the invention of new modes of measuring and modeling coupled to the Internet of Things (Kitchin 2014b). This framework has once again shifted the ontology of the urban and thereby also the premises from which we can ask questions about its design.

While it did not pretend to be even remotely comparable in impact, DYLIAB aimed to create critical metrics and imaginative maps that would shift the empirical ground from which the debate about urban diversity could unfold. By teaming up with GEHL architects, I deliberately positioned myself in a situation from which it was possible to revisit inherited imaginations about this problem in the professional as well as public arena. The goal was to reconfigure problem spaces (Lury 2020) rather than provide evidence to predefined problems.

The computational turn: A fertile ground for critical metrics and imaginative mapping

The empirical ground for the diversity project was a big repository of granular traces of Danish Facebook users engagement with politics and urban events. Later, I will explain the details of how we translated these traces into the maps. For now, I want to briefly emphasize that the project actively leveraged the fact that the digitization of our urban lives (Halegoua 2020) has created a situation where urban analysts have access to traces of people’s urban engagement with a previously unseen granularity. They have access to algorithmic tools for finding patterns in them. This is what Berry (2011) has termed “the computational turn.” Following the lead of Crampton and Krygier (2005), I want to propose two reasons why this turn has the potential to change the way we pose problems about urban phenomena.

First, the rise of the social web and handheld apps have enabled a much more diverse set of people producing data about urban life (Halegoua 2020). People rate the urban environment on Yelp, they depict it on Instagram, and they indicate their willingness to engage in it on Facebook events. Such data are authored “from below” but heavily shaped by the algorithms and logics of the commercial platforms from which they are sourced. The fact that such data enables a different gaze on the social was already discussed a decade ago when it challenged the authority of established data producers in sociology (Savage and Burrows 2009) and has more recently been described as a “redistribution” of research methods in the social sciences (Marres 2012).

Second, digital traces have different empirical potentials than the bureaucratic measures that have previously dominated data-driven urbanism. They are simultaneously quantitative and qualitative, they are accompanied by a plethora of metadata conserving the context of their production (Munk 2019), and they are more granular and relational than normal urban data sources (Kitchin 2014a). Whereas data on urban life have often been aggregated on standardized spatial grids, such as the street, the block, and the neighborhood (Congress for the New Urbanism 2000), digital traces have a spatiotemporal granularity that makes it possible to investigate previously unexplored patterns.

Recent development in urban studies illustrate this potential. One example is the way Shelton et al. (2015) uses tweets in Louisville to identify a specific street corner where people from the east side and the west side are indeed co-located on specific time slots. Here, digital traces are used to question the popular idea that Louisville is divided across 9th street. This methodological strategy is not dissimilar to the way Booth took advantage of granularity in his maps of London to challenge engrained mental maps. Other examples of how social media data have been used as a basis for alternative cartographies with the potential to shift our urban perspective are the sounds maps of Aiello et al. (2016) and the smellscapes by Quercia et al. (2015). In a recent paper, we discuss more examples of this form of digital cartography as part of a larger methodological paradigm that we label “soft city sensing” (Madsen et al. 2022).

DYLIAB: (Re)Operationalizing urban diversity

We focused DYLIAB on the problem of diversity because it was already on GEHL’s agenda. For instance, the company had already developed a “public life diversity toolkit,” which included recipes for how to measure the level of “social mixing” in urban space (GEHL Studio SF 2015). Two of the central methods introduced in this toolkit are surveys and observations. The main focus is on the potential of using these methods to study sociodemographic mixing. This frame is nicely illustrated by the charts (see Figure 2) that report findings concerning encounters between people of different age and gender in Union Square (NYC) and Patricia’s Green (San Francisco).

These charts illustrate how the empirical techniques of surveys and observations successfully translate the issue of social mixing into a question concerning sociodemographic traits. However, GEHL’s diversity toolkit also includes a few tentative experiments in using digital traces to map this form of social mixing. For instance, by collecting the last 10,000 Instagram photos from Patricia’s Green and Union Square—and looking at previous photos from each active Instagrammer in these areas—the toolkit proposes ways to paint a portrait of the economic diversity of these areas. This is done by inferring the home neighborhood of each Instagrammer and using census data to estimate their economic status. The result is shown in figure 3 where the map to the right shows the distribution of photos and the circle charts show that Patricia’s Green attracts people from neighborhoods that are a little wealthier and a have a slightly higher employment rate than the crowd at Union Square. Traces on Instagram here come to constitute a proxy for its author’s economic class. This proxy approach enables a more fine-grained mapping of sociodemographic diversity than the earlier charts.

While the toolkit concludes that digital traces can indeed be used to explore the social dimensions of urban spaces, it also emphasizes the lack of standardized ways to infer the residential status of the Instagrammer as a serious drawback. While this problem is indeed important, it is also magnified by the GEHLs states intention to use digital traces to bring the census down to the street level. The census is still the ideal empirical technique that digital traces are evaluated against. Since this method is born out of an interest in understanding the sociodemographic composition of residential areas (Scott 2020), it makes sense that the possibility of connecting digital traces to ZIP codes is seen as an important test for the empirical value of digital traces. While arguably interesting and valuable, this specific experiment in using digital traces from Instagram to map urban diversity is not critical and imaginative in the sense outlined previously. It is not used to critically re-examine how to phrase problems about urban diversity. In GEHL’s existing tools, the empirical material made available by the computational turn is used to provide further evidence to a predefined problem that centers around people’s residence and their sociodemographic traits. In contrast, DYLIAB deliberately tried to unsettle this frame in two ways.

First unsettlement: Turning the gaze away from demography

Social media data is vague on demographics. Whereas the metadata accompanying API-generated data can be very detailed in terms of, for example, geography and temporality, it is often silent on traits such as gender, age, and income (Rogers 2013). This is also true for the data in DYLIAB, which we obtained from a project in TANTLab entitled “Atlas of Danish Facebook Culture” (Munk and Olesen 2020). Before Facebook limited the conditions for API access in the aftermath of the Cambridge Analytica scandal, this project constructed a database containing almost every post, comment, event, and reaction on public Danish Facebook pages from 2012 to 2018. These data points are accompanied by anonymized IDs of the users leaving these traces. It is accordingly a rich source for asking all sorts of questions about patterns in people’s engagement with urban phenomena (such as concerts, demonstrations, football games, and dining). However, demographic metadata is absent. If we were to answer questions within the standard sociodemographic frame, we would have to engage in the kind of (problematic) inferential work discussed earlier.

In DYLIAB, we decided to inquire into the political diversity of Copenhagen as an alternative to the demographic frame. One reason for this was that digital traces from Facebook simply affords the former type of analysis rather than the latter. By working with a new data source, we were prompted to imagine new ways of mapping diversity. Another reason was that political segregation was a topic of interest for both the TANTLab researchers and GEHL employees. We were all interested in understanding whether the city functions as a filter bubble where people meet like-minded people or whether urban life has the quality of building bridges between political opposites. This is a question that has been the agenda of urban planning since the publication of The Big Sort (Bishop 2009) and that has recently surfaced in debates about urban social infrastructure (Klinenberg 2018; Brown and Enos 2021). In order to profile anonymized IDs based on their political leanings, we went through the operational steps shown in figure 5.

First, we located all Danish Facebook pages describing themselves as political. We found around 700 such pages, and we manually assigned each of them to one of three “blocks” in Danish parliamentary politics. The red block comprises Facebook pages from five socialist/social democratic/green liberal parties. The blue block is economically liberal and comprises pages from three political parties. The yellow block is nationalist and anti-immigration and comprises pages from three political parties. This way of dividing the Danish political spectrum is standard in political analyses of Danish elections. From these 700 political pages, we downloaded 300,000 posts that we interpreted as political expressions with a specific leaning. For instance, if a post was from a Facebook page in the red block, we assumed it would carry a political statement in favor of that block.

Second, we made a list of the anonymized IDs of the users who had left more than two positive interactions—likes or loves (hearts)—on these posts. We defined these users as politically interested, and we made this specific cut with reference to previous research arguing that it does not take more than two likes to predict the political leanings of a Facebook user (Kristensen et al. 2017). Furthermore, we focused only on “likes” and “loves” because a previous experiment in our lab showed that these are the only two reactions on Facebook that carry an unambiguous signal of sentiment (Munk, Jacomy, and Knudsen 2021). For instance, the choice to leave an angry emoji on a political post can actually be intended as a support for that post in the sense of sharing its indignation about something. In our data set, more than 600,000 unique IDs had this level of political engagement. We decided that if a user had more than 60% of their positive interactions in one block, they would be considered leaning toward that block. This resulted in a list of the IDs with a metadata on their political leanings. This list was stored in a GDPR secured environment at Aalborg University.

Second unsettlement: Beyond the neighborhood unit

Most studies of urban political polarization concern the voting composition of residents in predefined neighborhoods. Rather than challenging this premise, the academic debate has primarily been focused on which type of voting trace would be the most valid (Abrams and Fiorina 2012). The maps emerging from this frame represent a form of “residential cartography” that ultimately builds on the assumption that the political identity of a neighborhood is characterized by the preferences of the people who sleep and pay taxes there (Harding and Blokland 2014). In that sense, it continues a rich theoretical history dating back to Perry’s (1929) concern with face-to-face interactions between residents in “neighborhood units.”

Again, digital traces perform badly when evaluated on their capability to answer questions within this established frame. If we want to know where users of social media live, we need to engage in the kind of inferences discussed earlier. The data we obtained from the “Danish Facebook Atlas” is much stronger when used as signals of people’s attachment to so-called “third places” that are neither home or work (Oldenburg 2013). For instance, it contains 130,000 Facebook events held in Copenhagen between 2012 and 2018 as well as a list of the user IDs signaling their interest or plan to attend them. We cannot know whether these users actually went to the event, but we do know that the choice to press the “interest” button means that they signaled to a group of friends that this was an event that they considered in tune with their online identity. Driven by an interest in understanding the way such events contribute to the political dynamics of Copenhagen, we decided to finish our diversity map through the following operational steps.

First, for every 130,000 events, we made a list of the anonymous IDs of users attending or showing an interest. Subsequently, we compared this list to the list of political IDs. Of the 600,000 political user IDs, we found that half of them had engaged with events in Copenhagen. Second, we used this comparison to calculate the percentage of the event crowd that was interested in politics as well as a percentage-wise breakdown between the blocks of this politically interested crowd. For instance, if an event had 100 interested users and 30 could be found in our list of politically interested IDs, this event would have a political crowd of 30%. If these 30 users were divided evenly across the three blocks, the political breakdown would be 33.3% red, 33.3% blue, and 33.3% yellow. Finally, we converted this breakdown into a diversity index of 0–100 for each event. An event scores 0 if the whole political crowd supports the same political block. It is a political bubble. By contrast, it scores 100 if its political crowd is evenly divided between the red, blue, and yellow blocks. It is a bridge. It is this index that is the basis for colors in the interactive datascape shown in the introduction.

By operationalizing diversity in this way, we shift the premises of the debate from a residential ontology to a focus on third places where the same user can in principle have interactions all over town. This follows in line with a recent focus on “informal segregation” (Alexander and Tredoux 2010) in the literature on diversity. Rather than looking at the preferences or traits of residents in a given area, diversity scholars have during the last decade begun to turn their analytical gaze toward everyday, interpersonal interactions between people in informal settings (Tredoux and Dixon 2009). DYLIAB can utilize the existence of digital traces to bring this focus to the topic of political diversity as well.

A note on data quality and the issue of bias

Before discussing the extent to which the datascape succeeded shifting established frames around urban diversity, it is important to note that we never thought of DYLIAB as a progression from the established methods at GEHL. First, there are very good reasons that most studies of urban diversity have been couched in a sociodemographic frame. This is a frame that is understood by policymakers, and there are arguably many important issues of inequality and segregation that hinge upon people’s race, gender, age, and other demographic attributes. The extent to which our residential areas succeed in mixing people on such attributes is absolutely vital to the social qualities of our cities. Accordingly, we do not claim that political diversity is more important than demographic diversity, and we do not claim that the role of third places are more important than where people decide to live. Our claim is merely that we should take advantage of the fact that the computational turn enables us to map diversity in new ways. This is a resource we can use to critically assess whether the frame within which we have hitherto discussed this issue tends to be too narrow.

It is important to note that we do not consider the data obtained from Facebook to be of higher quality than, for instance, surveys or voting data obtained from public resources. However, it is a data source with different potentials and drawbacks than the data that is already in standard use in analyses of urban diversity. The potentials have already been discussed, so I briefly touch upon the most important drawbacks. First, as the data is repurposed from a commercial platform with no intentions of producing a good infrastructure for research, it is inevitable that proprietary algorithms influence the behavior of users and thereby also the empirical traces of this engagement (Gillespie 2014). Furthermore, the users of a commercial platform can never be as representative of the general population as a randomized sample (Lomborg and Bechmann 2014). Whereas it makes sense to more or less directly compare survey answers across countries, digital traces need to be interpreted with reference to the local context of their production. Third, we cannot know whether a user attending an event on Facebook has been there in person. We can only interpret it as a signal that this is an event that the user is interested in and wanted to share with a community of Facebook friends.

In DYLIAB, we decided to use Facebook as our empirical ground because this platform has an extremely high penetration in Denmark. More than half of the adult population is active on Facebook more than once a week (Slots-og Kulurstyrelsen 2018), and there are no significant differences in use when it comes to gender, education, and location2 (Bechman 2019). This is different when it comes to age, where we see a slight user drop in use after the age of 65. However, more than half of the Danish population aged 65–74 still uses the platform (Slots-og Kulurstyrelsen 2018). Whereas the Facebook data we collected cannot be considered as representative as a survey, we concluded that its potentials outweighed its drawbacks for our explorative purposes. Our ambition was to create a map that could unsettle the problem space in which this issue of diversity was discussed and stimulate new forms of investigations into this problem. We thought of DYLIAB as a conversation starter—not the final proof.

A reconfigured problem space?

One of the aims of the datascape was to enable users to explore the political diversity of Copenhagen all the way down to the level of venues. Our hope was that this would stimulate a debate about urban diversity that was different from the one framed by demographics and residential voting patterns. Again, this is not a new idea. When Booth made his granular poverty maps of London, he similarly put great effort into exhibiting them for the public (Vaughan 2018). That was one of his tricks for ensuring that his maps would indeed spur the kind of critical reconfiguration of the debate on poverty that we are also seeking in the context of diversity.

After launching the datascape, we were therefore curious about the responses. Would it attract attention? Would it help establish a new framework around the debate on urban diversity? Would people be convinced by it or challenge its assumptions and operationalizations? The answer to our questions was that the datascape did receive attention. After its launch, it was featured in several national newspaper articles and was the topic of a one-hour podcast. It also sparked discussions among architects (also internally at GEHL) and led to conference invitations within the field of urban planning. Drawing on these different forms of feedback, I want to elaborate on two areas where our datascape succeeded in reconfiguring the debate about urban diversity.

Juxtaposing urban diversities (and illustrating the impossibility of a “catchall” strategy)

We did not design the datascape for completely open-ended exploration. Whereas the user can indeed choose to just explore the map in an open fashion, the datascape also includes our own attempt to narrate a data-driven story concerning the connection between urban design and political diversity. This story compares two areas in Copenhagen—Folkets Park in Nørrebro and Sundbyøster Plads at Amager. One element in the story is the two illustrations in Figure 7 that show how Folkets Park—the area with the highest ethnic and sociodemographic diversity in Copenhagen—came out as the least politically diverse. Despite being a melting pot in the classic demographic and residential frame, it came out as a bubble by our new digital metrics. In the story published on the datascape, we hypothesized different reasons for this. For instance, we suggested that the different types of amenities available in the two areas might have an influence on their political crowd.

The fact that we found areas of high political diversity to score low on demographic diversity (and vice versa) makes it relevant to think about urban life as comprising distinct forms of diversity that may sometimes counteract each other. In the public discussions in the aftermath of the project, I therefore proposed to conceptualize a diverse city as a conglomerate of “partially diverse pockets” that each serves as a home for selected groups of people. Taken in isolation, these pockets (e.g., Folkets Park) are partially diverse because they create a home for specific groups but implicitly exclude others. However, the totality of pockets with different diversity characteristics can, in sum, ensure a diverse city. This way of framing the problem provides an alternative to the way diversity has, for instance, been discussed in recent literature on universal design. In this literature, the stated goal for every building and urban space is to support access for all and eliminate any form of exclusion (Khalil et al. 2021).

Following the logic of universal design, the way to evaluate the inclusiveness of the two squares in Figure 7 would therefore to be to investigate whether each of them—in isolation—excludes specific groups. If yes, this would be a problem to take care of. However, following the frame of “partially diverse pockets,” any public space will inevitably exclude some people and invite others. Folkets Park excludes the political right, whereas Sundbyøster Plads excludes certain demographic groups. As the computational turn enables us to measure diversity on multiple parameters, we will probably realize that there is no chance for a universal “catchall” diversity. Therefore, it makes sense to look at the city as a conglomerate of partially diverse spaces and accept that no space can invite all groups. In this framing, it is the city as a whole—rather than the individual square—that is the relevant unit for evaluating urban diversity. We need to accept different forms of segregation mechanisms at the micro-level (e.g., a square at Nørrebro making room for immigrants and queer people at the expense of alienating a specific political subgroup) in order to ensure diversity in the city as a whole.

The fact that the map made space for this frame in the public debate illustrates one way in which it contributed to reproblematizing the issue of urban diversity. After the datascape was featured in an article in the newspaper Berlingske Tidende (Johansen 2019), this specific frame was taken up as a discussion point by people already engaged in debates about diversity. One example of this is a conversation under a LinkedIn post sharing the newspaper article.3 One participant in this conversation was Kasper—the head of innovation at an urban planning company in Copenhagen. He was drawn to the discussion because he found that the DYLIAB project used the concept of “diversity” in a way that misses the real issue. In making this point, he stated that “[…] it is important to clarify that social mix [mangfoldighed] is not the same as political diversity [politisk diversitet].” To him, it is important to make this distinction, because he argues that “[…] if you live in a neighborhood characterized by social mixing, then you streamline your political views. [This is] precisely, because you want a socially mixed city.”

Kasper’s comment is interesting in relation to the discussion about the hopes of universal design. He precisely distinguished between “social mix” and “political diversity” in order to make the argument that sometimes you need to sacrifice the latter to achieve the former. He explicitly mentioned the area of Nørrebro as an example of this trade-off. His argument is that Nørrebro is an area of social mix precisely because it squeezes out people who discriminate. In this reading, Nørrebro only functions as a “safe space” where different cultures, age groups, and sexualities can coexist because it is a “red bubble.” Interestingly, this narrative about Nørrebro is very close to the one officially promoted by VisitCopenhagen—the local tourist destination manager. In a campaign, it described Nørrebro as “a multicultural pot of diversity and tolerance […] a place where cultural diversity is practiced and appreciated” (KnowYourBro, n.d.).

Contrary to the frame of “pockets of diversity,” Kasper’s arguments in the LinkedIn thread imply that he sees a hierarchy among urban diversities. Demographic diversity is more valuable than political diversity. A few of the people from the DYLIA team intervened in the debate and suggested that our aim with the map was precisely to question such inherited narratives about what diversity is and which types of diversity a city should design for. While not immediately convinced, Kasper agreed that this would be an interesting question to pursue. Interestingly, the last comment in the thread was authored by the city architect of Copenhagen, Camilla van Deurs, who stated that she found the project “super important.” In that sense, it seems fair to say that the datascape did stimulate a conservation about diversity among local professionals in urban planning.

If we turn to the comment thread under the post that Berlingske Tidende made about the article,4 we can see how a similar conversation unfolded among its readers. For instance, the article ignited a discussion between four men who used the datascape as a starting point for normative reflections on how citizens should navigate urban space in a democratic society. One of them—Jannick—recognized the bubble dynamics shown in the map. He also turned to Nørrebro as an example of an area where he experienced what he calls “social shaming” in relation to politics. As he rhetorically puts it, “If you are [part of the yellow block], then just try to go to Nørrebro on a Friday or Saturday night to discuss politics. Try to see if people want to have a beer and a chat.” The clear subtext to this rhetorical question is that the answer is “no.” Jannick mentions a concrete example of a female politician who was chased out of Copenhagen bars because of her political views. Again, he rhetorically asks: “Is that OK? Just because she is of a different political opinion?”

The continuation of the Facebook thread contains very different answers to Jannick’s questions. For instance, it stimulated an interchange between Aske and Michael, who each interpreted the issue in their own distinct ways. As a direct response to Jannick, Aske suggested mechanisms of political exclusion can be justified. Referring to ongoing Danish debates on immigration and citizenship, he argued that when some political parties actually want to exclude his friends from the country, “[...] it makes sense not to talk to them.” He added that this is his red line and that he “[…] would be OK sharing a beer with anybody else.” In a direct response to Aske, Michael stated that his comments perfectly illustrate what he believes to be the core problem exposed by the map. He argued that “[…] if we can no longer engage in a conversation with those we disagree with, then we have no chance of achieving anything.” Aske’s reply is that he agreed that in the overall scheme of things, people should have conversations with those they disagree with, but he also maintains the right to choose a fun night out “[…] without having to act as a social pedagogy for people on the right wing. Even if it would be the best way to get to understand each other.” To Aske, there are limits to the meaningful conversations one can be expected to engage in, and talking to people who support the yellow block on a night out is one of those limits.

These conversations between professionals on LinkedIn and the general public on Facebook illustrate that the choice to design a cartography from the perspective of political diversity proved unsettling for people who are used to translating questions about diversity into questions concerning the coexistence of people with different demographic traits. It is clear that some of the commentators (Kasper and Aske) operate with a hierarchy of diversities wherein political diversity resides below demographic diversity. Others use the datascape to challenge this hierarchy. One consequence of DYLIAB is to make such assumptions about diversity hierarchies explicit. In fact, this issue was also raised at internal meetings within GEHL architects. In a follow-up interview, Jeff linked this to the strength and importance of identity politics as a framing for urban problems: “You saw in our meeting how our staff has kept on going back to demographics [...] we’re so trained to understand people through these demographic traits, and so it’s really hard to look beyond that, and then also, I think, in a world of identity politics and social justice [...] those are all determined by demographics, so, and I don’t think we should necessarily throw all that away” (personal interview).

In these discussions, it becomes apparent how urban metrics and norms about the good democratic city are deeply intertwined. This is something that has been discussed for two decades in valuation studies (see Kornberger et al. 2015), where the consistent finding is that values are not discovered but rather practiced into being, often through measurement. New metrics make new things valuable. When the value of political bridges becomes visible in our map, the official narrative about the diversity of an area such as Nørrebro is questioned.

The color of gray (and the challenge to design as a diversity driver)

Since our map is based on people's engagement with third places, it affords quite different hypotheses about the drivers of diversity than analyses focused on residential areas. Maps of the latter type can illustrate how specific neighborhoods have high or low diversity, which would then naturally stimulate hypotheses concerning aggregated neighborhood traits. On such a frame, diversity is often understood through the use of official statistics, such as income or education levels, and the question becomes how to improve those numbers for specific neighborhoods. By contrast, when a map offers the possibility of seeing patterns all the way down to the level of venue, it opens the way for thinking about the individual traits of such venues—or their immediate surroundings—as drivers of diversity. This possibility sparked a conversation in the DYLIAB team about the fact that some of the most diverse venues in Copenhagen seemed really dull and uninteresting on almost any design criterion. One example can be found in Figure 8. Here we see the concert venue ‘Viften’ which is located across from a shopping mall in a quite generic public space with well-ordered trees and a Danish flagpole next to a small pond. While this space seems unremarkable on images it does perform very well in terms of hosting events with high political diversity. However, it is unlikely that it would be chosen as an illustration of Copenhagen’s diversity by either urban planners or public officials.

The fact that these types of venues stood out as especially diverse gave rise to a working hypothesis that we—in the project team—half-jokingly titled “the color of gray.” The intuition behind this tagline was that the absence of any interesting signals may actually be part of the reason why a broad community feels at home in these venues: the colorful mix of red, blue, and yellow users was somehow related to the dull appearance and the lack of identity. This idea motivated us to qualitatively map the semiotics of the politically homogeneous square at Nørrebro. If the semiotics of diverse spaces were dull and inactive, could we see a different pattern in Nørrebro—the most politically homogeneous space? As can be seen from Figure 9, the answer is “yes.” We found that Folkets Park in Nørrebro had very active signage that was often shaped by specific political agendas. For instance, there appeared to be a consensus among the regulars in this area to refrain from taking down political stickers in the public interior. You could say that the regulars seem to select the visiting crowd by the way they manage signage.

This finding feeds into the previously mentioned literature on informal cues as drivers of diversity. One of the arguments in this literature is that a focus on, for example, semiotic spaces can expose problems that remain hidden at the demographic macro-scale. Furthermore, this literature emphasizes that a focus on semiotics results in a more dynamic view on diversity than studies looking at demographic neighborhood traits and residency. The reason is that the actual signs and people’s interpretation of them change way more frequently than people’s residential preferences. Whereas people rarely move residence (and select their residence primarily based on economic factors), they are constantly confronted with choices concerning the informal places they visit and the way they interpret their signage and identity (Zisman and Wilson 1992). Understanding the issue of diversity through informal signage thereby unlocks the possibility for more ad hoc solutions. To put it very simply, if the presence of stickers and symbols make some people feel unwelcome in an area, it is a pretty simple move to remove them (while keeping in mind that those same stickers may be what make other people feel a sense of belonging!).

The finding that places such as the skating rink and Viften are among the most diverse in Copenhagen also challenges established local assumptions about how to design for diversity. One example concerns the narrative about another area in Nørrebro called “Superkilen,” a public space that has received international acclaim for its diversity design (see figure 10). In 2016, it won the Aga Khan Award for its achievement in bringing people together across ethnicity, religion, and culture. That quality was also noted in the mayor’s reception speech, where he emphasized that the designers of the area had succeeded in creating a space with “room for everybody” (Dania 2016) . One of the design choices that has been praised for ensuring this mix is the choice of populating the square with urban interiors from the different countries of the residents in the area. Examples of this are a Moroccan fountain and a Russian hotel sign. The square also promoted inclusive symbols, such as bike parking painted in rainbow colors.

The professional assumption behind Superkilen is obviously that the task of designing for diversity is solved by stuffing an area with so many symbols that everyone feels a sense of belonging. The link between this design choice and the normative goal of universal inclusion is explained by Rasmus Nielsen, one of the responsible artists, who stated that “we wanted to create a public space where people from diverse backgrounds feel at home” (The Aga Khan Award 2016, 2 min 25 sec). While this is probably true when it comes to the demographic backgrounds of the residents in this area, our map shows that it is not true if we evaluate it through the lens of political backgrounds. In fact, the area is one of the stronger bubbles in our map. The dull gray containers depicted carry many more colors when visualized on our digital metric.

Of course, the next question is why they do so. Is it because the norm in those places is not to talk politics? In other words, are they political bridges because politics is downplayed to such an extent that it ceases to matter? This was a question that was also posed by some of the commentators in the Facebook thread discussed earlier. To investigate this, we looked at the “political crowd” of the areas—the percentage of the people attending events at a given venue who also figure as politically active in our data set. Interestingly, Superkilen and the skating rink score equally on this measure. Approximately 30% of their event attendees engage with politics on Facebook. Of course, this does not mean that those political interests are exhibited in the same way in these two spaces. It seems likely that people express their political commitments more at Superkilen, which is often a preferred starting point for political demonstrations. Investigating whether this is the case would require qualitative studies. However, even if places such as the skating rink attract a politically diverse crowd because politics is downplayed, this would still be interesting for the purpose of understanding urban diversity. While a skating rink that downplays political expressions and arguments could not be interpreted as a Habermasian forum for political discussion, it would still serve the function of giving people a shared point of reference and perhaps even a shared concern if, for instance, the funding of the skating rink was in danger. It would be an implicit bridge between people who may not even know that they are political opposites.

Outro: Interventionist scholarship in an age of new empirical traces

Throughout this paper, I have tried to describe how DYLIAB can be seen as an intervention that prompt a critical re-examination of previously established standards for measuring and mapping diversity. I have even implied that this kind of interventionist work can be a specific role that urban researchers can take upon themselves in the context of the computational turn. In a situation where empirical traces of the urban multiply, we are faced with interesting possibilities to unpack the normative assumptions about existing ways of measuring urban space. Instead of thinking about data as a tool for representation, prediction, and control, we can think of it as a way to pose new questions and design new visibilities. This is in line with the imaginative mapping agenda set out by critical cartographers, and it is a strategy that has recently been discussed under the slogan “speeding up data to slow down reasoning” (Jensen et al. 2021).

However, the story told in this paper echoes other studies that show how such an interventionist strategy is not without its complications (Madsen and Munk 2019). In DYLIAB, it is clearly the case that our attempt at creating critical metrics and imaginative maps intervene in established agendas that we may even sympathize with. An obvious example is the agenda to design urban spaces that bring people together across demographic traits. As we have just concluded that urban measurements are deeply intertwined with norms, it is not surprising that our project has been met with critical questions about our own position in the debate. In a sense, this resurfaces a dilemma that has accompanied sociologists at least since the 1960s, where Becker’s (1966) paper “Whose Side Are We on” discussed how scholars engaged in the study of deviance clearly saw it as their duty to take an ethical stance and help deviant communities get a voice (Seeley 1966). Following this line of reasoning, one could argue that our map does a disservice to the deviant communities of Nørrebro by challenging the engrained narrative that their community is the most tolerant and inclusive in Copenhagen.

While this is arguably a relevant critique, I do want to suggest that this line of reasoning is also characterized by having already made the normative diagnosis beforehand. By contrast, the choice to intervene with critical metrics is driven by the idea that the existing normative diagnosis could be unpacked and refined in the course of experimentation. This is what Zuiderent-Jerak (2015) points to when he argues that the aim of interventions is to reconfigure problem spaces rather than providing evidence to predefined problems/diagnosis. The metrics and maps we developed made such reconfigurations possible because they were used to simultaneously detect patterns in political diversity and amplify this specific aspect of urban life in public discussions (Law 2004). As stated by Zuiderent-Jerak (2015), such experimental amplification puts the social researcher in a quite different normative role than the one envisioned by the scholars of deviance. This role is “one of dealing with the overflowing normativities encountered in a setting one is involved in.” These are overflows that one has a responsibility for.

However, the fact that the interventionist researcher is willing to make their normative diagnosis while intervening should not be mistaken for a normative deficit. As argued, the choice to invent new metrics is normative in the first instance. For instance, our project was driven by a belief that the possibility for a city to have pockets where people of different political orientations can meet is valuable. The realization that this form of diversity often occurs in areas that are low on demographic diversity creates a normative overflow. If we deal with different forms of diversities, it becomes hard to say that one supports a diverse city per se. If there is no “catchall” diversity, as proposed by adherents of universal design, the diversity agenda suddenly has to include debates about the dynamic between different types of diversity and even a prioritization concerning the kinds of diversity one wants to design for. Perhaps urban centers need some areas that cater to one demographic diversity and others that cater to political diversity in order to be truly diverse.

Critical metrics can help raise such questions, but they cannot answer them. Managing the normative overflow requires a democratic conversation concerning what we believe to be good democratic urban spaces. Facilitating this debate is part of the role of the interventionist researcher, and one important inspiration for how to conduct this role can be found in Williams (2020) on “data action.” In this book, she suggests that ethical use of urban data could be inspired by insights from collaborative planning theory that have demonstrated the importance of building trust among stakeholders. Working with data and new forms of quantification can be a way to build new communities of interest. The idea that data practices need to involve the voice of those affected by the measures is also the guiding thought between the framework of Participatory Data Design (Jensen et al. 2021). Drawing on the Scandinavian tradition of participatory design, this STS-inspired approach suggests addressing dilemmas of quantification by involving the relevant public early in the data work. In a world where urban metrics matter, I take the design of such inclusive data processes as one of the main tasks of the data-driven urban researcher.