940698

research-article20202020
SMSXXX10.1177/2056305120940698Social Media <span class="symbol" cstyle="Mathematical">+</span> SocietyBoy and Uitermark

SI: Studying Instagram Beyond Selfies

Social Media + Society

Lifestyle Enclaves in the Instagram City?

July-September 2020: 1­–10 
© The Author(s) 2020
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/2056305120940698
https://doi.org/10.1177/2056305120940698
journals.sagepub.com/home/sms

John D. Boy1 and Justus Uitermark2

Abstract
Commentators and scholars view both social media and cities as sites of fragmentation. Since both urban dwellers and social
media users tend to form assortative social ties, so the reasoning goes, identity-based divisions are fortified and polarization
is exacerbated in digital and urban spaces. Drawing on a dataset of 34.4 million interactions among Amsterdam Instagram
users over half a year, this article seeks to gauge the level of fragmentation that occurs at the interface of digital and urban
spaces. We find some evidence for fragmentation: users form clusters based on shared tastes and leisure activities, and
these clusters are embedded in four distinct lifestyle zones at the interface of social media and the city. However, we also
find connections that span divisions. Similarly, places that are tagged by Instagram users generally include a heterogeneity of
clusters. While there is evidence that Instagram users sort into groups, there is no evidence that these groups are isolated
from one another. In fact, our findings suggest that Instagram enables ties across different groups and mitigates against
particularity and idiosyncrasy. These findings have important implications for how we should understand and study social
media in the context of everyday life. Scholars should not only look for evidence of division through standard network
analytic techniques like community detection, but also allow for countervailing tendencies.

Keywords
Instagram, social media, segregation, urban space, integration

Although social media corporations and their representatives suburbs light up. The city of Amsterdam, our case study area,
continue to argue that their platforms incubate a global com- also exhibits plain contrasts between overwhelmingly left-
munity, commentators and scholars nowadays stress that leaning inner-city areas and more right-leaning outer bor-
social media are at risk of undermining cohesion and democ- oughs. These differences in political preferences are tightly
racy. The behemoth Facebook is a case in point: CEO Mark coupled with differences in lifestyles and identities, resulting
Zuckerberg published a lengthy manifesto about his compa- in “lifestyle enclaves” (Bellah et al., 1985; DellaPosta et al.,
ny’s role in fostering global community at the start of 2017, 2015). It seems plausible, perhaps even inescapable, that
just a little over a year before he was called to testify before processes of self-segregation online and offline work together
Congress about the company’s role in the American political to generate increasingly fragmented social landscapes.
crisis. Danah boyd’s (2017) verdict was merciless. She called This article addresses these concerns by studying social
Zuckerberg “naive as hell” for believing in the dream that relations of Instagram users in Amsterdam, examining how
“he could build the tools that would connect people at they form groups, segregate, and claim different places
unprecedented scale” (n.p.). A large and growing literature within the city. While we find some evidence of “lifestyle
documents how social media’s penchant for reinforcing enclaves” among Amsterdam’s Instagram users, we also
assortative ties results in polarization, balkanization, echo highlight connections between groups and processes of inte-
chambers, and filter bubbles (Del Vicario et al., 2016; Pariser, gration. In short, we try to answer the question to what degree
2011). Such concerns over social media-induced fragmenta- and in what ways processes of fragmentation and integration
tion dovetail with anxieties about geographic segregation. shape the relations of Amsterdam-based Instagram users. In
While titles like American Apartheid (Massey & Denton,
1998) signal long-standing concerns about racial segrega- 1
Leiden University, The Netherlands
tion, more recently commentators have expressed worries 2
University of Amsterdam, The Netherlands
about lifestyle segregation. In the United States, progressive
Corresponding Author:
coastal states are pitted against the conservative Midwest John D. Boy, Institute of Anthropology and Sociology, Leiden University,
and South. Looking at a lower level of granularity, stark divi- Wassenaarseweg 52, P.O. Box 9555, Leiden 2333 AK, The Netherlands.
sions between progressive inner-cities and conservative Email: j.d.boy@fsw.leidenuniv.nl

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further permission
provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/open-access-at-sage).
2 Social Media + Society

the following, we start by presenting our perspective on frag- village community suppresses differences, the city rein-
mentation and integration. We specifically focus on the inter- forces and amplifies them (Fischer, 1982; Wirth, 1938). This
face of Instagram and the city, examining whether Instagram process makes for extraordinarily vibrant environments of
users indeed self-segregate online and in the city. Using con- highly diverse subcultures, but also results in fragmentation
cepts and methods that are widely used in contemporary and its associated evils of anomie, collective paralysis, and
debates on polarization and social media, we then empiri- failures of empathy.
cally show that Instagram users in Amsterdam do, in fact, It is no coincidence that contemporary anxiety over social
sort into groups with specific appearances and lifestyles. We media echoes historical concerns over cities. The move to
further demonstrate that these groups construct zones at the cities and the development of modern communication tech-
online–offline interface, that is, symbolic and material nologies are essentially two sides of the same coin: both
domains that serve as stages for the enactment of identity and developments emancipate people from the communities they
the performance of status. While our analyses, therefore, were born into and allow them to associate with people of
confirm that tendencies toward fragmentation are present, their own choosing. It is plausible that, when the city and
we go on to complicate this conclusion by dissecting its con- social media become intertwined, differences are multiplied,
ceptual and methodological premises. To put it bluntly, we reinforcing mechanisms of fragmentation (Bastos et al.,
find fragmentation, but only if we neglect any countervailing 2018; De Waal, 2014; Graham, 2005; Kitchin & Dodge,
processes. The second part of the empirical analysis, there- 2011; Wang et al., 2018). In this article, we pursue this line
fore, takes a different angle in analyzing our dataset and of thought by examining how subcultures emerge at the
demonstrates that there are indeed formidable processes of interface of cities and social media (Boy & Uitermark, 2016,
integration at play. The Instagram city, we argue, may be a 2017). We show how different social groups claim their
much more integrated, and much more boring, place than the space and mark their territory. Places figure into this story as
tidings of fragmentation and conflict would suggest. stages for expressing individual status and group belonging.
Social media users generally do not picture quotidian activi-
ties like visiting the supermarket, but rather share experi-
Understanding Fragmentation and ences of places for aspirational consumption (see also Boy &
Integration in the Instagram City Uitermark, 2016, 2017; cf. Currid-Halkett, 2017). Social
As social media allow us to associate with like-minded peo- media, in this line of thinking, are hyper-segregated: by
ple, so communis opinio holds, we are inadvertently yet selectively displaying where social media users are, they
ineluctably drawn into echo chambers or filter bubbles. To reflect and reinforce segregation on the ground.
make things worse, algorithms reinforce our propensity to However, the evidence of a connection between social
associate with those like us by suggesting we befriend our media use and fragmentation is moderate and mixed, even in
friends’ friends or read more from the blogs we just visited. the United States context, on which most of the work in this
Social media, then, feed on our differences and reinforce field is focused (Boxell et al., 2017; Garrett, 2009). We
them, resulting in fragmentation (Pariser, 2011; Sunstein, therefore need to develop a perspective that allows not just
2001). While we are increasingly connected to others who the possibility of fragmentation but also of its opposite, that
are just like us, the distance to others grows. American com- is, integration. In addition, concerns over fragmentation are
munication scholar danah boyd (2017, n.p.) sums up this typically voiced in relation to Facebook and especially
pessimistic diagnosis: Twitter (Tufekci, 2014), much less in relation to other plat-
forms, including Instagram. Since Instagram is much more
Ironically, in a world in which we have countless tools to visual than either Facebook or Twitter, we may expect that it
connect, we are also watching fragmentation, polarization, and is less likely to induce acrimonious debate. The budding
de-diversification happen en masse. The American public is research literature on Instagram emphasizes that the platform
self-segregating, and this is tearing at the social fabric of the is not a staging ground for symbolic resistance (Manovich,
country. 2016, p. 23) but gives ample space to corporate-sponsored
influencers to shape tastes and desires (Abidin, 2016; Boy
These concerns about fragmentation sound all too famil- et al., 2018) while compelling users to enact idealized selves
iar to students of the city. The scholars of the Chicago and hide stress and strains (Duffy & Hund, 2015). Whereas
School of sociology argued in their classic works that the many studies on Twitter highlight polarization, this literature
disintegrating forces of modernity—as observed by first- on Instagram conjures up the image of users connecting in an
generation sociologists like Ferdinand Tönnies and Émile environment where beauty, wealth, and success are cele-
Durkheim—reached their apex in cities. Robert Park (1915) brated and estheticized (cf. Boy & Uitermark, 2017;
wrote of cities that “every social group tends to create its Marwick, 2015).
own milieu . . . The processes of segregation establish moral When we speak of “integration,” we therefore do not nec-
distances which make the city a mosaic of little worlds essarily mean a benign process where people with different
which touch but do not interpenetrate” (p. 608). While the preferences and interests come together in harmony. We
Boy and Uitermark 3

view social media as stages for the expression of status that ethnographic case studies thus suggest that social media do
are characterized by mutual monitoring, the collapse of pri- not solely amplify difference but could also facilitate integra-
vate–public distinction, and, most importantly, stratified sys- tion and promote conformity. We therefore examine both
tems of rank (Boy & Uitermark, 2019). What is essential to fragmentation and integration at the interface of Instagram
social media is not that people share images of posts, but that and the city. Before we do so, we discuss our methods and
these contributions are appraised by proximate and distant data.
others. “Integration” means that people come to depend on
and surveil one another. Although this process may be har-
Methods and Data
monious, it can also involve competitive individualism and
breed anxiety as social media users are compelled to antici- Our analysis in this article is based on a corpus of 709,348
pate appraisals by others. geotagged Instagram posts gathered over half a year, between
We should thus take seriously the possibility that the 1 December 2015 and 31 May 2016. On Instagram, users can
social media promote integration and breed conformity opt into geotagging (attaching a location to posts) on a post-
instead of amplifying difference. There are prima facie rea- by-post basis. Since our main interest is in how city dwellers
sons to consider the scenario. A number of commentators use locative social media in their everyday lives, our corpus
have observed how, in the age of social media, radical and only includes posts by users with at least two geotagged
deviant subcultures have withered away. For instance, Jessa posts at least 4 weeks apart to eliminate likely tourists. The
Lingel (2017) has written about several groups that used to total number of users in our corpus is 78,207, equivalent to
make a home on the open web during the 1990s, only to be about one-tenth of Amsterdam’s population. On 1 June 2016,
gradually displaced by the ascendant social media platforms Instagram severely restricted the data that could be accessed
when the 2000s came along. One of her studies focuses on through its application programming interface (API), which
extreme body modification, a subculture that for many years is why we focus on the period up to 31 May 2016.
had a virtual meeting place at Body Modification Ezine With our corpus, we can investigate how social media is
(BME). The founders of BME positioned the platform as an implicated in the creation of subcultures and social divisions.
“online haven for outsiders” (Lingel, 2017) where members Our main data are the 34.4 million “likes” and comments
could share their experiences with face tattoos, scarification, among the users in our dataset. While we acknowledge that
subincisions, stretched ears, piercings, flesh pulls, split the meaning of likes and comments varies across contexts
tongues, and the like (p. 37). The platform flourished in the and situations, we pragmatically consider a reciprocated tie
late 1990s but faltered as Facebook rose to prominence. (I comment on, or like, your post and you comment on, or
Although the story of BME’s decline is complicated, it is like, my post) as a proxy of affinity between users. Out of the
clear social media feature prominently. The promise of a interactions in our data, 130,665 are reciprocated, and we use
wider audience pulled members away from BME and onto these mutual ties to identify groups and the relations between
social media like Facebook where body modifiers’ sense of them. Since we use different methods to study fragmentation
alterity and community dwindled. and integration, respectively, we provide further details on
Writing on Rotterdam’s gay scene, Ferrie Weeda (2018) these methods in the following, empirical sections.
also relates the ascendancy of social media, and specifically
the dating app Grindr, to the disappearance of a subcultural
Fragmentation
milieu. After the dating app Grindr allowed its users to seek
hook-ups and partners online, the number of gay meeting There is plenty of evidence of fragmentation if we look for it.
places dropped precipitously. One after another, gay bars and The network of reciprocated likes or comments has a modu-
clubs have closed down. While LGBTQ folks may benefit larity score of circa 0.6, a relatively high score (Newman,
from the efficiency of the dating app to arrange a tête-à-tête, 2006), suggesting that interactions tend to occur within rela-
more radical and collective expressions of difference lose tively more densely knit subgroups of users and providing
their place within the city. Grindr allows its users to search prima facie evidence of the fragmentation often associated
for specific “tribes” (otter, bear, geek, twink, trans, etc.), but with city life and social media. We can identify subgroups
its effect is less to reinforce difference than to have LGBTQ thanks to a procedure called community detection. By apply-
people retreat into privacy or blend into the public. ing the Louvain method of modularity optimization to the
These examples suggest social media may spur the inte- undirected network of reciprocated ties (Blondel et al.,
gration of deviant groups into the mainstream and contribute 2008), we find a total of 31 clusters with at least 200 users.
to the dissolution of subcultural milieus. Difference does not We subsequently engage in qualitative coding. We identify
so much disappear but comes to be expressed through ever the 10 most central users in each cluster, look for common-
more subtle and individual strategies of distinction—a pro- alities among them, and characterize the clusters accord-
cess Elias (1994) captures with the phrase “diminishing con- ingly. For instance, when we spot pictures of people flexing
trasts, increasing varieties” (pp. 382–386). The ambiguous their muscles in gyms and references to “personal body
evidence of social media-induced polarization as well as plans” in bios, we label the cluster as “fitness enthusiasts.”
4 Social Media + Society

Table 1.  The nine largest clusters of Instagram users. When users within the same clusters use places as staging
grounds for their status displays, we assume a link between
# Label Users Posts Median number
of followers those places; the more links between places, the more likely
they are part of the same zone. We construct a proximity
1 City Consumers 4,642 86,925 684 matrix of places based on how frequently they are tagged by
2 Hedonist Lifestyles 4,412 57,879 785 people in the same cluster. We then apply the same method of
3 City Imageers 3,977 81,482 562 community detection as discussed above, first turning the
4 Rich Kids 3,196 20,225 552 proximity matrix into a co-occurrence network. This yields
5 Gay Performers 2,843 31,904 550 four clusters of places, three of which are of roughly equal
6 Refined Lifestyles 2,097 33,682 857
size (between 500 and 600 places), and one that is smaller.
7 Clubbing 2,073 22,234 707
These zones are not contiguous areas, as in Parks’ “natural
8 Beliebers 1,952 11,740 498
areas” or Burgess’ concentric zones model (Park & Burgess,
9 Fitness 1,773 23,301 520
1925), but sets of places in the city that figure into status
displays on Instagram. See Table 2 for an overview of the
four zones, which also presents some additional data on the
Through such qualitative coding, we are able to character- places we sourced from Yelp, the popular social reviews
ize these clusters in terms of shared interests and lifestyles. site.2 Before considering in greater detail how these zones
In addition to the nine largest clusters shown in Table 1, we and the places that make them up are bound up with status
identify several smaller ones. They include, in order of displays in the city, we first describe their features.
decreasing size, international students (1,120 users), The nightlife zone consists, at its core, of places associ-
Amsterdammers of Turkish descent (821), coffee aficiona- ated with the city’s clubbing scenes. According to Yelp, the
dos (769), Russian-speaking expats (519), evangelicals review site that we mined for additional data on places tagged
(395), CrossFit adherents (350), and electronic dance music by Instagram users, a typical closing time for places in the
enthusiasts (304). Table 1 shows that users in the cluster nightlife zone is 6:00 a.m. Concert venues like Paradiso and
organized around “refined lifestyles” have somewhat more Melkweg and clubs like Jimmy Woo, Bitterzoet, and Club
followers than users in the other clusters. However, the great- Air are at the center. Generally, we find most of the city’s
est inequality is found within clusters rather than among dance clubs and a high number of bars and cafes within this
them: all the clusters have very uneven distributions of both zone. Footwear and sportswear stores carrying local brands
likes and comments. We thus find that Instagram users self- also rank highly, suggesting that some sartorial and con-
organize into clusters of different sizes but all with a median sumer choices predominate in the city’s clubbing scenes and
number of followers between 403 (the CrossFit adherents) serve as a source of subcultural capital (Thornton, 1996).
and 857 (the cognoscenti of refined lifestyles). While there Images tend to show performers and groups swept up in the
are differences among the clusters in terms of follower count action. More than in any other zone, the images taken here
or activity, what stands out are qualitative differences in show moments of collective enjoyment.
terms of interests and lifestyles. If the nightlife zone is about dancing, the lifestyle zone
Since we are especially interested in the interface of online appears to be about eating. The yuppie’s favorite meal,
and urban spaces, we subsequently examine which places brunch, is an important occasion to frequent places in this
these different groups tag. A place tag is a predefined location zone, in which hotels, cafes, and restaurants that serve brunch
name that can be attached to a post. This is a form of metadata staples like poached eggs and pancakes predominate. At the
that enables geographic exploration on Instagram, but also center of this zone, we find a number of upmarket hotels—
enriches individual posts with additional information.1 the Conservatorium, The Hoxton, and W Hotel—where
In line with the perspective we outlined, we do not inter- patrons like to picture beautifully plated French toast and
pret place tags as “trace data” that can be used to track users’ bespoke cocktails. Boutique coffee places, such as Coffee &
trajectories through the city, but as features of status displays. Coconuts in De Pijp, as well as a slew of restaurants serving
Users typically do not post about their daily shopping at the various cuisines also form part of this zone. Several locations
supermarket or their ride to work, but selectively and strate- on the city’s luxury shopping street, the P.C. Hooftstraat, are
gically use Instagram as a platform to live out their identities among the lifestyle locations, as are other places associated
and showcase their social contacts, sense of style, achieve- with fashion and design, such as the Dutch headquarters of
ments, or new purchases (cf. Boy & Uitermark, 2017; Hearst, publishers of Elle, Harper’s Bazaar, Esquire, and
Hochman & Manovich, 2013; Zasina, 2018). Different kinds Cosmopolitan; a showroom of Dutch design; and a fashion
of places offer resources (props and audiences) for different retailer specializing in “the good things in life.” Looking
kinds of displays that garner esteem and prestige in different again at Yelp reviews, places in this zone have comparatively
social scenes. Places are not just physical settings, but also high ratings and high prices.3 Images foreground moments of
social situations that encourage contextually appropriate consumption, often conspicuous, or at least indicative of
expressions of conformity and distinction. sophistication (Currid-Halkett, 2017; Veblen, 1899/1934).
Boy and Uitermark 5

Table 2.  Zones at the interface of the city and Instagram.

Nightlife Lifestyle Culture Fitness

Tagged places 604 588 533 91
Average latest closing 6:00 a.m. 6:00 p.m. 5:00 p.m. 6:30 p.m.
time
Most common hashtags #music #party #paradiso #food #love #coffee #rijksmuseum #netherlands #fitness #workout #gym
#iamsterdam
Posts 71,312 49,431 42,151 10,976
Users 19,912 13,085 15,648 3,945
Focal areas Leidseplein, Herengracht, De Pijp Museumplein, Amsterdam Zuid, Noord
Rembrandtplein Arena
Standout places Paradiso, Jimmy Woo The Hoxton, Rijksmuseum, Eye Film Changing Life Hub,
Conservatorium Hotel Museum Vondelgym
Typical images Bands, performers, Food, clothes, group Art, architecture, outside Groups, action shots,
dancing shots, selfies views selfies, outfits

The cultural zone revolves around the city’s museums, clear, albeit less pronounced, tendencies. Nightlife locations
with the iconic Rijksmuseum at the helm. Alongside it are are tagged by users in the Hedonist Lifestyles cluster and the
other well-known landmarks and cultural institutions—the Clubbing cluster, while locations in the culture zone are
zoo, botanical garden, opera house as well as the public tagged by the cluster of City Imageers as well as internation-
library’s central branch—which are frequented by the city’s ally oriented clusters of expats and foreign-exchange stu-
cultural connoisseurs and pictured for distant audiences who dents. Lifestyle locations are tagged by various clusters of
appreciate images of Amsterdam cityscapes and sights. Users apparel and fashion enthusiasts and the CrossFit cluster
tagging these locations frequently invoke the city brand (which branches out from the fitness zone into other territo-
#iamsterdam, which in its sculpture form is an inescapable ries). Unsurprisingly, we also find the cognoscenti of refined
sight on Instagram, serving as a metonymy for the city as a lifestyles represented here.
whole (which lacks instantly recognizable landmarks on par This aspirational dimension of Instagram use comes out
with the Eiffel Tower or Big Ben). They are tagged by expats not only in the places that are tagged but also in what is por-
and people with a more international audience, suggesting trayed in these places and how. We could start with the most
they signal well to these international audiences as markers notorious genre of social media post, the selfie. Generally,
of being in Amsterdam. commentary on the selfie is out of proportion to its actual
Finally, the smallest of the four is the fitness zone, which prominence on most social media, and in our data, too, self-
unsurprisingly revolves around sports clubs and gyms. ies account for only a small proportion of the total volume
Amsterdam’s CrossFit gyms and yoga studios can be found (cf. Boy & Uitermark, 2017; Manovich et al., 2014). In the
in this zone along with Yoghurt Barn franchises. This zone fitness zone, however, posts bearing a #selfie hashtag (or
comprises not only places appealing to the health-conscious ironic variations like #shamelessgymselfie) can be found
but also establishments that cater to other practices involving much more frequently, which speaks to the centrality of the
the body, such as tattoo parlors and a cryotherapy center physical self to status displays staged in this zone.4 In the
(where customers can subject their bodies to temperatures of culture zone, architectural details and outside views predom-
110°C below zero for 3 min). More than in other zones, the inate because here recognition hinges on one’s identification
ideal of expressive individualism shines through in the dis- with the branded image of Amsterdam. In the lifestyle zone,
plays from this zone (Turner, 2011). bands, performers, and party people literally take center
The uneven presence of clusters in different territories stage—here, esteem is rewarded on evidence of hedonistic
supports the assumption we made at the outset of this inves- pursuits. Unlike in the selfie-saturated fitness zone, portraits
tigation: that places are used strategically for displays that are more likely to show groups than individuals. Finally, in
play to different social scenes, garnering rewards in the form the lifestyle zone, still-lifes of desirable items—especially
of esteem or recognition. The fitness zone, in particular, food, fashion, and furniture—speak to the ways in which
stands out as the preferred domain of several clusters that conspicuous consumption continues to be an avenue toward
revolve around fitness and tattooing, suggesting that esteem prestige. In short, by combining community detection with
in these groups is bound up with particular places and their geographic analysis, we can show how subcultural groups
affordances—in this case, getting and maintaining an attrac- use digital technologies to mark their territory within the city,
tive, fit, and healthy body. The three larger zones which demonstrating that fragmentation occurs not only through
revolve around nightlife, culture, and lifestyle also show residential segregation but also through more complex
6 Social Media + Society

Figure 1.  Heat maps of Instagram activity for different groups centered on the same coordinates. Hotter colors indicate more place
tags.
The place tag and the location from which a message is posted do not necessarily coincide—a user may append a place tag for “Club Vividio” while
posting from their homes.
Boy and Uitermark 7

spatial sorting on the interface of social media and the city gravity in the center of the city. There are a couple of clusters
(Graham, 2005; Wang et al., 2018). that also show a lot of activity in Amsterdam South East
because of the concert venues in that neighborhood but this
Integration is hardly distinctive. If there is one cluster that stands out, it
is a cluster with Amsterdammers of Turkish descent (16) that
There is a large literature that directs our attention to divi- shows a lot of activity in the Western part of the city.
sions online and in the city. Part of the impetus for this litera- However, this cluster, too, gravitates to the center of the city.
ture is the concern that society will fall apart into different While both classic writings on the city and contemporary
groups that at best live past one another, and at worst will writing on social media would lead us to expect stark differ-
clash. The smartphone, as a territory device (De Waal, 2014), ences, we do not at all find that groups sort into internally
is believed to further buttress these divisions by algorithmi- homogeneous “natural areas.”
cally reinforcing and fortifying urban dwellers’ propensity to One might counter that the maps in Figure 1 are not at the
find kindred spirits. We have a range of tools and measures at right scale. Perhaps different clusters all tend to post from the
our disposal to capture such tendencies toward fragmenta- city center but from different places within it. The mixing of
tion. For instance, like many other researchers, we use com- different groups in the city center would, in this scenario,
munity detection to identify different groups. However, we reveal profound processes of segregation operating at a lower
need to appreciate that a community detection algorithm will scale where people might keep out others by constructing of
identify communities even in random networks. Identifying a parochial realm (Lofland, 1998). Although members of dif-
groups and their places within the city, as we have done ferent clusters traverse the same spaces in the city’s center,
above, risks overemphasizing the differences and fragmenta- they might ultimately self-segregate into different places—a
tion. We need to look more closely, and we may also need to pattern referred to in the literature as “social tectonics”
look differently, if we are to understand the relational struc- (Robson & Butler, 2001; see also Jackson & Butler, 2015).
tures that social media users spawn. However, when analyzing at the level of places, we do not
There are good reasons to not only look at fragmentation find strong support for this scenario. If we look at the 100
but also integration. While we preferentially connect with places that are tagged the most (in at least 309 posts), we find
like-minded people on social media, strangers are never far not a single place where posts originate from one cluster
away and our audiences are always multiple, at least poten- only. While some places are more parochial than others, as a
tially. As we explained above, social media insert users into general rule, members of different clusters rub shoulders in
systems of standardized rank and into each others’ purview bars, squares, restaurants, parks, clubs, or boutiques.
(Boy & Uitermark, 2019). Social media users know this all While it is now clear that members of different clusters
too well and tend to adjust their posts accordingly. Through traverse the same spaces and rub shoulders in places, perhaps
their exposure to multiple audiences, social media users have segregation operates in still more subtle and insidious ways.
to cope with or internalize different kinds of expectations The literature suggests that urbanites who move around in
and pressures. Acknowledging that social media bring differ- the same neighborhood and even frequent the same places
ent people together, we can write a different story of the still may have little to no contact. They may “live together
same network, using different measures and data points. apart” as they use digital devices to carve out their parochial
For instance, while we followed convention by character- domain (De Waal, 2014). If this would be so, there should be
izing clusters according to their most central nodes, we could little online interaction between members of different clus-
also look at randomly selected rank and file users. When we ters. This is, again, contrary to what we actually find.
know which cluster they belong to, it is usually not so diffi- Although the network of interactions has a relatively high
cult to see why that would be so. For example, users in the modularity score of 0.6 (as reported above), remarkably a
cluster of “gay performers” might present themselves as gay whopping two out of three interactions are between, not
or performers. However, gay people and performers are also within, clusters. The clusters observed through communities
to be found in other clusters, which means it is very difficult may be distinct, but they are also perforated and intercon-
to guess which clusters randomly designed users belong to, nected. As we can see in Figure 2, even though nodes within
suggesting that the communities we find through community each community cluster together, nodes of different colors
detection are not status groups in Weber’s sense or tribes in are also interspersed, indicating that there are numerous ties
Mafessoli’s sense (Maffesoli, 1988/1996; Weber, 1921/2010). between communities. Similarly, while we can identify
What is true for community detection also holds for our zones that serve as the domain of specific groups, we should
strategy of identifying zones: it is a method designed to high- also tell a different story. The place network—where differ-
light difference by filtering out similarity. When we look at ent places are connected when they are tagged by people in
the spatial footprint of the different clusters, as we do in the same cluster—has a modularity of only 0.15, signaling
Figure 1, we get a different sense of how Instagram users are that there are many connections between places.
positioned within the city. What is remarkable is that the heat Communities are neither bubbles nor bounded fields, and
maps are so much alike: all clusters have their center of interactions frequently span across clusters. Amsterdam
8 Social Media + Society

we can discern distinct groups around specific foci like hob-

bies, professions, or lifestyles. We further identify distinct
zones: sets of interconnected places that serve as the domain
for particular kinds of groups. And yet, that is not the whole
story. Our findings do not conform to the dystopian image of
deep and algorithmically fortified divisions. Even when
users socialize in a community of CrossFit fanatics, they are
never far from users with other interests, such as Beliebers or
coffee aficionados. Users coalesce into groups, so much is
true, but the boundaries of such groups are fuzzy. This casts
social media in a different light; perhaps they are best seen as
vehicles of integration rather fragmentation.
We come to our conclusions based on computational anal-
ysis of a slice of data produced by a specific population in a
particular place using a platform designed to facilitate visual
communication. In grounding our study in Amsterdam, we
have chosen a location that at least historically has resisted
tendencies toward disintegration and growing inequality
(Uitermark, 2009). Today Amsterdam is known as a liberal
city, and perhaps our results would have looked differently if
we had focused on a more divided city. Considering these
specificities, we concede that our case is likely not represen-
tative of other populations or platforms. But the same is true
for other research based on data sourced from, say, Twitter or
Facebook, that informs tidings of fragmentation. One way to
account for the differences between our observations of
Figure 2.  A graph representation of Instagram users in our Instagram and others’ observations of Twitter and Facebook
dataset.
Nodes are colored according to the clusters they belong to (see Table 1
is to trace them back to the affordances of different platforms
for a description) and scaled according to their eigenvector centrality. (cf. Van Dijck, 2013; Wellman et al., 2003). On this reading,
the patterns of interconnection and pressures toward confor-
mity we observe are peculiar to Instagram and the specific
Instagrammers organize into clusters according to their life- functions the application offers its users. Although we read-
styles and backgrounds but neither on Instagram nor in the ily agree that platforms have different affordances, we never-
city are they far removed from others. People may have a theless feel this kind of argument is limited by its privileging
primary reference group that is most consequential for how of the technological underpinnings of social relations.
they understand and comport themselves, but this primary Theorizing of affordances originated from the need to move
reference group is not apart from rest of the social world. beyond technological determinism and explicitly acknowl-
While they associate with people with similar interests and edge that the same technological set-up allows for different
lifestyles, they generally do not form enclaves. The entan- kinds of social relations to emerge. And yet, technology
gled networks documented in this article are the structural remains the starting and end point of analysis—whatever
backdrop of the cross-pressures that users experience as they happens, happens because technology affords it, leading
consider posting to the platform. The aggregate result of researchers to scrutinize design decisions in minute detail.
these cross-pressures is that Instagram breeds conformity: While we do not take issue with this interpretation, we do
the platform is used for a range of purposes by different want to consider another. There certainly are important varia-
groups but it nevertheless has aesthetic and social norms that tions between platforms, but it is nevertheless possible to dis-
all users have to reckon with (Manovich, 2016). cern trends. While the internet initially functioned and felt like
an alternate reality, it is now increasingly woven into everyday
life. Social media accounts make the internet more personal-
Conclusion
ized, intimate, and visual, while also making interdependen-
Both social media and the city are widely seen as spaces of cies more extensive, differentiated, and dense. The relational
fragmentation. In these spaces, commentators expect and patterns we identify here emerge within this structural context:
fear, people will flock to each other, forming enclaves or as we construct our personae and connections through social
bubbles, losing touch with the wider society. Departing from media, we are compelled to take into account the views of
this perspective, this article traces the formation of groups at proximate and distant others. The processes and mechanisms
the interface of the city and Instagram. We indeed find that we identify on Instagram may be less salient on other
Boy and Uitermark 9

platforms, but we surmise they are present there, too. So what Notes
is the use of our case? We would suggest that Instagram pro- 1. Place tags are often rather generic. For instance, users could
vides an alternative starting point for theorizing. Where tag their picture with “Amsterdam” or “Amsterdam West.”
researchers of political communication on Twitter or Facebook In the analysis that follows, we restrict our analysis to clearly
use their specific cases to theorize about fragmentation, we defined places on the assumption that they convey a status
can use our study of Instagram to highlight mechanisms of signal when they are tagged. We eliminate places with a lot
integration. Where researchers of political communication of variation in the rooftop coordinates associated with them.
view social media posts as expressions of opinion, our case This includes place tags for large parks, long streets, or entire
pushes us to consider them as status displays. Our theoretical neighborhoods. We manually verified the remaining places,
keeping 51.2% of tagged locations. There are 1,750 places in
perspective applied to a specific set of data enables us to iden-
Amsterdam that people across clusters tag. Using this list, we
tify processes and dimensions that may not have caught the look at the co-occurrence of places within these clusters.
attention of researchers working from a different theoretical 2. We were able to gather Yelp data for just over half of the places
perspective and studying different platforms. in our database.
Yes, there are radical or outlandish views even on 3. The average Yelp rating of 4.15 is noticeably higher than the
Instagram,5 but there are also powerful pressures toward overall average (4.0), indicating that places in this zone are
conformity that render countercultures precarious. Users viewed favorably not just by the Instagrammers who tag them,
pursue distinction, but in a conformist way—they know what but by Yelp reviewers as well. Second, Yelp indicates how
the norms are and they abide by them. In this article, we costly establishments are through the use of repeated dollar
bring into view the wider set of relations through which signs ($, $$, $$$, and $$$$). Again, the lifestyle establishments
norms are maintained: the fine-grained and cross-cutting score highest, with an average of 2.4 dollar signs, compared to
an overall average of just 2.28. In both cases, the differences
linkages within and between communities. The sorts of com-
are slight but significant and further support the impression
munities discussed at the beginning of this article require a that the lifestyle zone comprises high-status establishments.
degree of closure to shield its participants from the dominant We performed a t-test of statistical significance (p < .01).
gaze and have low chances of survival within this constella- 4. In the fitness territory, 3.4% of post captions contain “selfie,”
tion of the fine-grained and cross-cutting linkages. Instagram as opposed to 0.9% overall.
users in Amsterdam form an integrated, albeit differentiated, 5. At the time of writing, several high-profile members of far-
social world. Social media are the interface through which right groups continued to be present on Instagram after having
we negotiate what is acceptable, exceptional, or beautiful. been banned from other platforms; see Sommer (2018).
The aggregate outcome of these processes of negotiation is
not an online space partitioned into a wide range of commu- References
nities that each have their own ideas or norms, but an expand- Abidin, C. (2016). “Aren’t these just young, rich women doing vain
ing web of relations that bring people into dependence and things online?” Influencer selfies as subversive frivolity. Social
implores them to take others’ views into account. Media + Society, 2(2), 1–17.
Bastos, M., Mercea, D., & Baronchelli, A. (2018). The geographic
Acknowledgements embedding of online echo chambers: Evidence from the Brexit
campaign. PLOS ONE, 13(11), Article e0206841.
We are grateful to Petter Törnberg and Willem R. Boterman for Bellah, R. N., Madsen, R., Sullivan, W. M., Swidler, A., & Tipton,
taking the time to read this article and provide us with helpful feed- S. M. (1985). Habits of the heart: Individualism and commit-
back and suggestions. We also thank Special Issue editor Alessandro ment in American life. University of California Press.
Caliandro for his gracious and incisive comments, and the anony- Blondel, V. D., Guillaume, J. L., Lambiotte, R., & Lefebvre, E.
mous reviewers for numerous suggestions for improving the article. (2008). Fast unfolding of communities in large networks. Journal
of Statistical Mechanics: Theory and Experiment, 2008(10),
Declaration of Conflicting Interests 10008. https://doi.org/10.1088/1742-5468/2008/10/P10008
Boxell, L., Gentzkow, M., & Shapiro, J. M. (2017). Greater internet
The author(s) declared no potential conflicts of interest with respect
use is not associated with faster growth in political polarization
to the research, authorship, and/or publication of this article.
among us demographic groups. Proceedings of the National
Academy of Sciences of the United States of America, 114(40),
Funding 10612–10617.
The author(s) disclosed receipt of the following financial support Boy, J. D., & Uitermark, J. (2016). How to study the city on
for the research, authorship, and/or publication of this article: We Instagram. PLOS ONE, 11(6), Article e0158161. https://doi.
gratefully acknowledge funding from the ODYCCEUS Project as org/10.1371/journal.pone.0158161
financed through the European Union’s Horizon 2020 program Boy, J. D., & Uitermark, J. (2017). Reassembling the city through
(grant agreement no. 732942). Instagram. Transactions of the Institute of British Geographers,
42(2), 612–624.
Boy, J. D., & Uitermark, J. (2019). Theorizing social media with
ORCID iD Elias: Status displays, mutual monitoring, and the genesis of
John D. Boy https://orcid.org/0000-0003-2118-4702 new sensibilities. https://doi.org/10.31235/osf.io/phm5x
10 Social Media + Society

Boy, J. D., Uitermark, J., & Wiersma, L. (2018). Trending #hijab- Park, R. E., & Burgess, E. W. (1925). The city. University of
fashion: Using big data to study religion and the online–urban Chicago Press.
interface. Nordic Journal of Religion and Society, 31(1), 22–40. Robson, G., & Butler, T. (2001). Coming to terms with London:
boyd, d. (2017). Why America is self-segregating. https://points.data- Middle class communities in a global city. International
society.net/why-america-is-self-segregating-d881a39273ab Journal of Urban and Regional Research, 25(1), 70–86.
Currid-Halkett, E. (2017). The sum of small things: A theory of the Sommer, W. (2018). Instagram is the alt-right’s new favorite haven.
aspirational class. Princeton University Press. The Daily Beast. https://www.thedailybeast.com/instagram-is-
DellaPosta, D., Shi, Y., & Macy, M. (2015). Why do liberals drink the-alt-rights-new-favorite-haven
lattes? American Journal of Sociology, 120(5), 1473–1511. Sunstein, C. R. (2001). Republic.com. Princeton University Press.
Del Vicario, M., Vivaldo, G., Bessi, A., Zollo, F., Scala, A., Thornton, S. (1996). Club cultures: Music, media, and subcultural
Caldarelli, G., & Quattrociocchi, W. (2016). Echo chambers: capital. Wesleyan University Press.
Emotional contagion and group polarization on Facebook. Tufekci, Z. (2014). Big questions for social media big data:
Scientific Reports, 6, 37825. https://doi.org/10.1038/srep37825 Representativeness, validity and other methodological pitfalls.
De Waal, M. (2014). The city as interface: How digital media are In Proceedings of the eighth international Association for the
changing the city. Nai010. Advancement of Artificial Intelligence conference on weblogs
Duffy, B. E., & Hund, E. (2015). “Having it all” on social media: and social media (ICWSM ’14). https://arxiv.org/ftp/arxiv/
Entrepreneurial femininity and self-branding among fashion papers/1403/1403.7400.pdf
bloggers. Social Media + Society, 1(2). Turner, B. S. (2011). Religion and modern society: Citizenship,
Elias, N. (1994). The civilizing process. Blackwell. secularisation and the state. Cambridge University Press.
Fischer, C. (1982). To dwell among friends: Personal networks in Uitermark, J. (2009). An in memoriam for the just city of
town and country. University of Chicago Press. Amsterdam. City, 13(2–3), 347–361.
Garrett, R. K. (2009). Echo chambers online? Politically motivated Van Dijck, J. (2013). The culture of connectivity: A critical history
selective exposure among internet news users. Journal of of social media. Oxford University Press.
Computer-Mediated Communication, 14(2), 265–285. Veblen, T. (1934). The theory of the leisure class: An economic
Graham, S. (2005). Software-sorted geographies. Progress in study of institutions. The Modern Library. (Original work pub-
Human Geography, 29(5), 562–580. lished 1899)
Hochman, N., & Manovich, L. (2013). Zooming into an Instagram Wang, Q., Phillips, N. E., Small, M. L., & Sampson, R. J. (2018).
city: Reading the local through social media. First Monday, Urban mobility and neighborhood isolation in America’s
18(7). https://doi.org/10.5210/fm.v18i7.4711 50 largest cities. Proceedings of the National Academy
Jackson, E., & Butler, T. (2015). Revisiting “social tectonics”: The of Sciences of the United States of America, 115(30),
middle classes and social mix in gentrifying neighbourhoods. 7735–7740.
Urban Studies, 52(13), 2349–2365. Weber, M. (2010). The distribution of power within the commu-
Kitchin, R., & Dodge, M. (2011). Code/space: Software and every- nity: Classes, stände, parties. Journal of Classical Sociology,
day life. MIT Press. 10(2), 137–152. (Original work published 1921)
Lingel, J. (2017). Digital countercultures and the struggle for com- Weeda, F. (2018). Hoe ik Rotterdam heb zien veranderen via datin-
munity. MIT Press. gapps als Grindr. Vers Beton. https://versbeton.nl/2018/09/hoe-
Lofland, L. (1998). The public realm: Exploring the city’s quintes- ik-rotterdam-heb-zien-veranderen-via-datingapps-als-grindr/
sential social territory. Aldine De Gruyter. Wellman, B., Quan-Haase, A., Boase, J., Chen, W., Hampton,
Maffesoli, M. (1996). The time of tribes: The decline of individual- K., Díaz, I., & Miyata, K. (2003). The social affordances
ism in mass society (D. Smith, Trans.). SAGE. (Original work of the internet for networked individualism. Journal of
published 1988) Computer-Mediated Communication, 8(3). https://doi.
Manovich, L. (2016). Designing and living Instagram photogra- org/10.1111/j.1083-6101.2003.tb00216.x
phy. http://manovich.net/index.php/projects/instagram-and- Wirth, L. (1938). Urbanism as a way of life. American Journal of
contemporary-image Sociology, 44(1), 1–24.
Manovich, L., Stefaner, M., Yazdani, M., Baur, D., Goddemeyer, Zasina, J. (2018). The Instagram image of the city: Insights from
D., Tifentale, A., & Chow, J. (2014). Selfiecity. http://selfiec- Lodz, Poland. Bulletin of Geography: Socio-Economic Series,
ity.net/ 42(42), 213–225. https://doi.org/10.2478/bog-2018-0040
Marwick, A. (2015). Instafame: Luxury selfies in the atten-
tion economy. Public Culture, 27(1), 137–160. https://doi.
org/10.1215/08992363-2798379 Author biographies
Massey, D. S., & Denton, N. A. (1998). American apartheid:
John D. Boy (PhD, City University of New York) is an assistant
Segregation and the making of the underclass. Harvard
professor of sociology in the Institute of Cultural Anthropology and
University Press.
Development Sociology at Leiden University. His research inter-
Newman, M. E. J. (2006). Modularity and community structure in
ests include urban studies, digital sociology, the sociology of reli-
networks. Proceedings of the National Academy of Sciences,
gion, and social and cultural theory.
103(23), 8577–8582.
Pariser, E. (2011). The filter bubble: What the internet is hiding Justus Uitermark (PhD, University of Amsterdam) is a professor
from you. Penguin Books. of Urban Geography at the Faculty of Social and Behavioural
Park, R. E. (1915). The city: Suggestions for the investigation of Sciences of the University of Amsterdam (UvA). His research
human behavior in the city environment. American Journal of interests include digital platforms, rapid urbanization, and urban
Sociology, 20(5), 577–612. policy.