Article

Environment and Planning A

Speculative futures: Cities, 2016, Vol. 48(9) 1691–1708
! The Author(s) 2016
data, and governance Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav

beyond smart urbanism DOI: 10.1177/0308518X16651445

epn.sagepub.com

Agnieszka Leszczynski
School of Environment, University of Auckland, Auckland, New Zealand

Abstract
In this paper, I examine the convergence of big data and urban governance beyond the discursive
and material contexts of the smart city. I argue that in addition to understanding the intensifying
relationship between data, cities, and governance in terms of regimes of automated management
and coordination in ‘actually existing’ smart cities, we should further engage with urban
algorithmic governance and governmentality as material-discursive projects of future-ing, i.e., of
anticipating particular kinds of cities-to-come. As urban big data looks to the future, it does so
through the lens of an anticipatory security calculus fixated on identifying and diverting risks of
urban anarchy and personal harm against which life in cities must be securitized. I suggest that
such modes of algorithmic speculation are discernible at two scales of urban big data praxis: the
scale of the body, and that of the city itself. At the level of the urbanite body, I use the selective
example of mobile neighborhood safety apps to demonstrate how algorithmic governmentality
enacts digital mediations of individual mobilities by routing individuals around ‘unsafe’ parts of the
city in the interests of technologically ameliorating the risks of urban encounter. At the scale of
the city, amongst other empirical examples, sentiment analytics approaches prefigure ephemeral
spatialities of civic strife by aggregating and mapping individual emotions distilled from
unstructured real-time content flows (such as Tweets). In both of these instances, the urban
futures anticipated by the urban ‘big data security assemblage’ are highly uneven, as data and
algorithms cannot divest themselves of urban inequalities and the persistence of their
geographies.

Keywords
Algorithmic governance, big data, future, security, smart cities

Introduction
In this paper, I examine the convergence of big data and urban governance beyond the
material and discursive context of the smart city. I argue that in addition to
understanding the intensifying relationship between data and governance in terms of
regimes of automated management and coordination in ‘actually-existing’ smart cities

Corresponding author:
Agnieszka Leszczynski, School of Environment, University of Auckland, Private Bag 92019, Auckland 1142, New Zealand.
Email: [email protected]
1692 Environment and Planning A 48(9)

(Shelton et al., 2015), we should further engage with modes of urban algorithmic governance
and governmentality as material-discursive projects of future-ing, i.e., of anticipating
particular kinds of cities-to-come. Indeed, David Lyon (2014) has argued that one of the
defining characteristics of big data is that it ‘‘focus[es] on the future more than on the present
and the past’’ (6). For him and others such as Aradau (2015), the future orientation of big
data is latent in a speculative ethos which underwrites a complex of analytics practices
designed to anticipate and shape the unfolding of possibilities, particularly those around
social deviance, risk, and unrest. The notion of ‘the future’ is often discursively articulated as
something necessarily different from the present and the past (Anderson, 2010; Kinsley,
2012; Massumi, 2007). While this promise of divergence may provoke affects of hope
bound up in the prospect of change for the better (Anderson, 2006, 2014), the radical
openness of the future as an endless set of potentialities simultaneously carries with it
threats of disaster, emergency and anarchy against which contemporary liberal-democratic
life must be securitized (Amoore, 2013; Anderson, 2010; Aradau and Blanke, 2015; de Goede
and Randalls, 2009).
Per Amoore (2011) such securitization may be realized through the operationalization of
a risk calculus which abstracts individual subjects into fragmented encodings she terms ‘data
derivatives.’ Here, the security calculus is an algorithmic process by which data derivatives
are cast onto a horizon of possibilities such that potential scenarios may themselves be
ascribed degrees of ‘‘risk, suspicion and . . . value’’ (Amoore, 2014, 2011; Crampton, in
preparation: 11). Decoupled from identifiable objects/subjects and stripped of context, the
data derivative is theorized as being agnostic to current (or ‘actual’) conditions (Amoore,
2014). Data derivatives feed algorithmic determinations of potentiality which establish
‘‘what can be imagined and inferred about who [subjects] might be[come]’’ (Amoore,
2011: 24), rendering these speculative subjectivities actionable in the here-and-now in ways
that are increasingly computationally enacted (i.e., data-driven and automated; Introna,
2016; Kinsley, 2012; Kitchin and Dodge, 2011).
Contemporary data-driven security as a safeguard against social as well as natural
disaster is inherently a project of future-ing (Anderson, 2010), i.e., of making secure,
resilient futures by circumscribing the horizon of possibilities to exclude potential
scenarios deemed undesirable or deleterious. Futuring is organized around anticipatory
regimes of governance which hold axiomatic that the ‘‘unforeseen can be made
foreseeable, can be somehow folded into present decision’’ (Amoore, 2007: 226). The
security–risk calculus which underwrites these modes of anticipatory governance seeks to
identify, manage, minimize and insulate against exposure to adverse possibilities through
shaping particular kinds of subjects, such as the ‘‘proto-suspect,’’ who expressly contain or
redirect future threats to equilibrium, stability, and order by actively participating in their
own algorithmic self-regulation (Amoore, 2007: 226; Anderson, 2007; Crampton, in
preparation). Accordingly Introna (2016) clarifies that ‘algorithmic governance’ is often
intended to actually refer to techniques and practices of algorithmic governmentality:
modes of governing whereby subjects are affected to govern themselves, as well as the
ways in which data and algorithms have become a fixture of the entire ensemble

formed by the institutions, procedures, analyses and reflections, . . . calculations and tactics that
allow the exercise of this very specific albeit complex form of power, which has as its target
population, as its principal form of knowledge political economy, and as its essential technical
means apparatuses of security. (Foucault, 1991: 102)

While anticipatory algorithmic governmentality has often been narrowly engaged in terms of
shaping particular kinds of subjects (see for example Amoore, 2011; Amoore and Hall, 2009;
Leszczynski 1693

Cheney-Lippold, 2011, Introna, 2016; Sadowski and Pasquale, 2015), Crampton (in
preparation) is mindful to point out that these active productions of subjectivities have
geographical outcomes. Algorithmic governance (as governmentality) is productive of
subjects and spaces. And as cities have come to be constituted as much by their digital
shadows—content generated about city spaces from city spaces—as by bricks and mortar
(Graham, 2013), much like subjects, they are subject to being dis/assembled across
information flows. Inasmuch as it is possible to theorize a data derivative as an
abstraction of individuals into discrete variables for a security–risk calculus (Amoore,
2011, 2014), it is possible to similarly posit an urban derivative. (Big) data about/from
cities likewise feed a speculative security calculus that projects urban derivatives onto ‘‘an
array of uncertain futures’’ in the interests of securitizing against that very uncertainty by
rendering it actionable in the present through various kinds of preemptive urban
interventions (Amoore, 2011: 24).
This is intimately bound up with the ways in which data have become ‘‘the modus
operandi and raison d’etre’’ of contemporary modes of urban governance (Shelton et al.,
2015: 16), particularly those immediately synonymous with ‘smart city’ initiatives which give
precedence to data capture and analytics in support of a range of urban processes and
practices including city planning, infrastructure operation, preemptive policing, and the
management of urban mobilities and flows (see also Kitchin, 2014b; Roche, 2014,
2016; Sadowski and Pasquale, 2015). Smart cities are driven by and organized around
efficiency gains on the time horizon of the present. Most real-time urban big data—such
as that from road traffic sensors, pollution monitors, flood sensors—are appropriated within
a very immediate time-frame to manage the here-and-now, whether interfaced via urban
dashboards or centralized within command-and-control facilities such as the Operations
Centre of the City of Rio (Kitchin, 2014b; Kitchin et al., 2015; Mattern, 2014, 2015).
But beyond the temporalities of the ‘real-time city’ (Kitchin, 2014b) and the concomitant
preoccupations with maximizing efficiencies in the present, (big) data as they are enrolled
within modes of urban governance simultaneously actualize ‘‘particular vision[s] of the
future city’’ (Shelton et al., 2015: 13). Accordingly, it serves us to ask, what kind of a city
is envisioned and affected by big data as it looks to the future? Or, alternatively stated, what
is the array of potential urban futures enacted by urban derivatives as they are projected
onto the horizon of possibilities via the data-security calculus? Questioning what kind of a
city big data anticipates involves analytically and discursively moving beyond the emphases
on efficiency and temporal immediacy implicit in narratives of smart urbanism. We must
further account for the material workings and effects of longer-term temporalities which
inflect modes of speculative algorithmic governance and governmentality in the service of
anticipate particular kinds of cities. These anticipatory algorithmic governance/
governmentality regimes are themselves characteristically oriented around a telos of
security rather than efficiency, as is consonant with an increasing intimacy between big
data and security (Amoore, 2016; Aradau, 2015; Aradau and Blanke, 2015; Lyon, 2014).
I argue that as the city is subsumed within the data-security assemblage as an object/subject
of algorithmic governance (as governmentality), urban big data do not actually envision a
radically different city as would be consistent with an understanding of the future as
organically open, as something that always comes as a surprise (Anderson, 2010).
Algorithmic governmentality cannot divest itself of actual urban realities of socio-spatial
stratification to which the derivative is theoretically indifferent. Extant inequalities are
abstracted into data flows, informing and propagating through the calculation of
algorithmically anticipated urban futures. Even where only speculative, this can be
1694 Environment and Planning A 48(9)

evidenced at two scales of digital urban praxis: the scale of the urbanite body, and that of the
city itself.
There are clearly many forms of urban big data that map at (and onto) the body and city
scales. At the scale of the embodied urban subject, for example, this can include practices of
logging and sharing personal spatial trajectories through practices of the quantified self; or,
the continuous monitoring, tracking, and identification of individual bodies via distributed
CCTV networks coupled to automated facial recognition technology seamlessly embedded
across numerous cityscapes of the world. Here, however, I present and work through two
selective examples at each of the scales of both the body and the city to illustrate how the
dividuation of the city into urban derivates feeds security calculi which are speculative in
their logic, operationalizing an anticipation of cities-to-come in the interests of their
securitization. The first example involves digital mediations of the individual urban
experience through locative media, which affect subjects to opt-in to the algorithmic
management of their mobilities as a practice of self-securitization against risks of
particular kinds of urban encounter. I discuss these with reference to neighborhood safety
applications. The second involves the ordering and regulation of city spaces achieved by
tapping into spatialized real-time urban digital content productions and misappropriating
them as proxies for collective dis/affection. These distillations of affect inform practices of
prefiguring ephemeral spatialities of urban deviance and risk through sentiment analytics.
I take each of these up in turn below, but first attend to the relationality between big data
and cities.

Urban, big, data

‘Big data’ has emerged as a nebulous placeholder for varied aspects of the contemporary
pervasiveness of data—digital abstractions of ‘‘observations, computations, experiments,
[and records]’’ of phenomena—in the spaces and practices of everyday life (Kitchin,
2014a: 2). This pervasiveness has conventionally been characterized in terms of the sheer
volume, variety, and velocity (the ‘three V’s’) of continuous, real-time flows of information
commensurate with the rise of content, the cloud, mobile computing, transactional capta,
distributed sensor networks, and the digitization of records (see for example Mayer-
Schonberger and Cukier, 2013). More recently, however, social scientists in particular
have moved past instrumentalist definitions emphasizing the size, speed, and diversity of
data generation by engaging ‘big data’ as socio-technical productions which form and
function as data assemblages (Aradau and Blanke, 2015; Kitchin and Lauriault, 2014).
‘Assemblage’ here designates the institutional, political, economic, technological, affective
and social arrangements and practices which enable and prioritize enrollments and
deployments of data within technologically enacted processes (such as data analytics)
leveraged towards the realization of particular kinds of social ends (boyd and Crawford,
2012; Crampton, in preparation; Crawford et al., 2014; Kitchin, 2014a; Kitchin and
Lauriault, 2014). Theorizing big data as an assemblage deemphasizes debates about
thresholds at which data productions become sufficiently sizeable to be considered ‘big,’
and foregrounds instead the unprecedented extent to which data have become entirely
ordinary and expected presences on ‘the street’ (Taylor et al., 2014).
The city itself has also been theorized as an assemblage, as comprising a uniquely
‘‘processual, relational, mobile and unequal’’ spatiality (McCann, 2011; McCann and
Ward, 2011; McFarlane, 2011a, 2011b: 649; McGuirk, 2011). Geographers know that
practices, actors, objects, institutions and processes come together in and occur (i.e.,
assemble) in space. In this way, it is possible to think about the city-assemblage in two
Leszczynski 1695

senses of the term: as a spatial assemblage (a unique spatiality that may be engaged as an
assemblage), and the city as actively spatializing (as giving spatial form to various
assemblages). As an example of the latter, Haggerty and Ericson’s (2000) ‘surveillant
assemblage,’ which captures the emergent, rhizomatic nature of contemporary
dataveillance, is theorized by the authors as inherently urban, enabled by the density of
surveillance apparatuses in cities. Kitchin (2015) builds on this in his recent theorization of
continuous geosurveillance as a phenomenon of the sensor-seeded, extensively monitored
landscapes of the ‘smart city’. These distributed sensor networks capture urban processes
and abstract mobilities as data to be enrolled as functional inputs towards a series of
(semi)automated operations and optimizations. Simultaneously, the city is also constituted
by volumes of data generated about city spaces from city spaces through quotidian
enrollments of mobile, digital information and communication technologies (ICTs) and
the platforms and services that run on these devices. To quote Graham (2013), ‘‘[c]ities
ooze data[;] they cast innumerable data shadows’’ (117). This is not only the effect of the
urban density of denizens and devices, but also of the ways in which content production is a
demonstrably urban phenomenon. In their analysis of volunteered geographic information
(VGI) contributions across three platforms—Twitter, Flickr, and Foursquare—Hecht and
Stephens (2014) found a statistically significant urban bias in the per-capita user-generation
of geospatial content. ‘Urban big data’ accordingly designates these multiple articulations of
cities and data: the urban as site, subject, and seat of data capture and production practices.
The ways in which code, data, and algorithms assemble the city are geographically
uneven. This is by no means a novel claim, but rather an enduring reality of the ways in
which the ‘softwirization’1 of the city has underwritten the intensification of various practices
of socio-spatial stratification that are rendered transparent by virtue of becoming
progressively digitally mediated and automated (Dodge et al., 2009; Graham, 2005;
Kitchin and Dodge, 2011; Monahan, 2008). Software, in other words, sorts the city
(Graham, 2005): bodies are positioned as belonging in or excluded from certain spaces,
and city spaces are characterized and geographically demarcated on the basis of the
presences of the kinds of bodies deemed expected or anomalous therein.
As cities are increasingly constituted not only by software and code but also by data
(Graham, 2013), the digital shadows they cast are not uniform but rather (re)enact urban
inequalities. For example, a Google Maps search for ‘restaurant’ in Tel Aviv conducted from
the same location in Arabic and Hebrew, respectively, returns radically divergent results,
with different establishments appearing at the top of the results chain for a query conducted
in either language (Graham and Zook, 2013). The result is a material reification
of imaginaries of which kinds of bodies belong in which kinds of spaces (e.g., Arab
speakers being directed to Arab sectors of the city; Hebrew speakers directed to Jewish
quarters of the city), and a shaping of commercial patronage behaviors and urban
mobilities of individuals and social groups. As further expounded upon by Graham
(2013), the results of such Balkanizations of spatial content (in this case along linguistic
lines) are the concomitant material productions of socio-economically stratified cities.
Graham and Zook’s (2013) analysis of the spatial contours of geolocated content is
illustrative of how content and algorithms come together in the material (re)production of
highly uneven urban geographies in the here and now, in real-time and on-demand. Yet as
urban content, algorithms, and governance become ever more intimately bound in a ‘(big)
data-security assemblage’ that privileges and values the operationalization of data within
analytics practices in the service of various forms of spatial securitization (of bodies, cities,
nations; Aradau and Blanke, 2015), the security calculus does not only (re)produce extant
urban inequalities in the present, but it also actively projects them forward in space and time.
1696 Environment and Planning A 48(9)

The material spatial effects and affects bound up in anticipations of unequal urban futures
may be observed across what may initially appear to be distantiated and unrelated moments
of contributing and tapping into spatial big data flows to support activities across multiple
scales of digital practice. At one end of a spectrum of urban big data practices are highly
individualized, ad-hoc enrollments of location-aware mobile devices and applications (apps)
that mediate the personal urban experience by proposing to optimize individual bodily
mobilities. At the opposite pole of the spectrum are city-scale practices of appropriating
flows of real-time geocoded social media content to optimize regulatory spatial interventions
by prefiguring and preempting spatial pre-futures of urban disorder.

Containing risks of urban encounter

Locationally aware mobile devices are ubiquitous presences on city streets around the world.
The urban experience is now thoroughly mediated by content, interfaces, and algorithms (de
Souza e Silva and Frith, 2012; Sutko and de Souza e Silva, 2010). Location-enabled
hardware/software objects do more than enhance and streamline quotidian activities such
as navigating through traffic or making public transportation connections. Increasingly these
mediations are real-time; mobile applications and services continuously interface circuits of
uninterrupted content flows by exploiting the locational awareness of positionally enabled
personal devices, often through the passive operation of locational utilities in the
background of the mobile operating system. The utilities of these applications and
services are varied, but a substantial proportion is predicated on anticipating individual
motilities by seeking to manage, rather than outright curtail (i.e., discipline), the
movements of bodies through cities.
The socio-spatial stratification effect bound up in the management of individual urban
mobilities may be evidenced in a class of mobile applications that may loosely be referred to
as neighborhood safety applications (or apps). These come in essentially two varieties. The
first group consists of navigation-centric applications, such as Microsoft’s patent for a
Pedestrian Route Production utility, which dynamically route individuals around
‘dangerous’ areas of cities, re-calculating possible lines for travel on the basis of a user’s
current geographic position, spatial history, social interactions, and behavior, all of which
are subject to capture as data. Apps of the second variety, such as the now-defunct
SketchFactor and Ghetto Tracker (renamed Good Part of Town prior to its quick demise)
released in major US cities, solicit user-contributed personal reviews of urban and suburban
neighborhoods, at times combining these contributions with public data such as crime
statistics, socio-economic and demographic data from the census. These services
crowdsource and subsequently present this content in the form of aggregate ratings of the
relative ‘sketchiness’ and ‘ghettoness’ of city spaces.
In the case of SketchFactor, these aggregate scores were presented as Chernoff face-like
placemarks meant to inform non-locals about neighborhoods they should avoid visiting and/
or travelling through (McEnry, 2014). Contributors could indicate other factors (such as
whether the area was poorly lit, or whether they observed instances/evidence of ‘crime’), and
could leave descriptive comments about what made the neighborhood ‘sketchy’ in their
opinion. As described by Biddle (2014), the very branding of the app carries with it racist,
classist connotations, as ‘sketchy’ is a term used almost exclusively by ‘‘young white
people . . . to describe places where they don’t feel safe because they watched all five
seasons of [the HBO television series] The Wire’’ (a show set in inner-city Baltimore
which explored, amongst others, themes of urban decay, social inequality and racial
segregation). The racist framing of an app named Ghetto Tracker is more explicit,
Leszczynski 1697

with ‘ghetto’ in the North American context constituting a derogatory designator for poor
inner-city neighborhoods inhabited by racial and ethnic minorities often residing in housing
projects (government housing). Even once renamed Good Part of Town, this branding carries
with it reminders of the enduring spatial legacies of residential redlining in the US, which
stripped racial and ethnic minorities of their rights to equitable homeownership (and
accordingly equal opportunities in education, the accrual of wealth and home equity, and
social mobility). Like SketchFactor, Ghetto Tracker/Good Part of Town similarly solicited
user feedback about the relative safety of urban and suburban areas, making summative
ratings and user comments accessible via map-based pop-ups under the auspices of
providing ‘travel advice’ (Narula, 2013).
Both these navigation-centric and review-based types of neighborhood safety apps enact
forms of algorithmic governmentality that prefigure and shape self-regulating subjects who
willing position themselves within, and actively contribute to, an urban security calculus as a
means of minimizing their own personal exposure to urban risks presented by risky bodies in
risky spaces. Microsoft’s Pedestrian Route Production patent for a dynamic foot-traffic
travel application to run on the Windows phone is emblematic of the ways in which such
modes of algorithmic governmentality are predicated on an anticipatory logic that
anticipates unequal urban futures. Popularly dubbed the ‘Avoid Ghetto GPS’ but never
actually developed for market, the Pedestrian Route Production patent envisioned a map-
based service that would dynamically (re)calculate a set of walking directions for users
taking into account ‘‘user history, weather information, crime statistics, demographic
information, etc.’’ and subjecting this data to the routing algorithm, which would
optimize travel across the city ‘‘according to at least one criterion, such as keeping a user
safe,’’ by ‘‘taking the user through neighborhoods with violent crime statistics below a
certain threshold’’ (Tashev et al., 2012: n.p.). What was to constitute ‘demographic
information’ or ‘crime statistics,’ and/or where such data was to be sourced, is nowhere
specified in the patent application. This similarly holds for the vague reference to ‘crime
statistics.’ The repeated use of language around the safety utility of the service to be realized
through its routing of pedestrians around ‘‘unsafe neighborhood[s]’’ designates soft-speak
for racialized and ethnic-minoritized poor inner-city American urban imaginaries of what
those spaces look like and the kinds of bodies that are thought to populate them (Tashev
et al., 2012: n.p.).
Accordingly Thatcher (2013) identifies the Pedestrian Route Production patent as more
than simply a prospectus for a navigational utility that optimizes pedestrian movement
through the city. In its very envisioning, the service anticipates futures that are always-
already raced and classed. The promise of the technology to ensure ‘‘efficiency, safety,
and new forms of coordination’’ is inseparable from, and is indeed predicated upon, the
‘‘[opening up of] a future wherein encounters on the city street are sorted by race; an unseen
algorithm enabling users to only ever encounter those already sorted as demographically
similar’’ (Thatcher, 2013: 967, 974). The Microsoft filing may in this way be thought of as a
patent for a highly individualized risk calculus. It looks to affect subjects to assume
responsibility for their own security by opting-in to an app that algorithmically triggers
behaviors that effectively avert and manage the risks of encountering risky (racialized,
classed) bodies in risky (ethnic-minoritized, impoverished) spaces by simply not projecting
these onto the array of potential futures (as possible trajectories for travel). The
operationalization of this calculus depends on the abstraction of both subjects and cities
into derivatives—geographic position of the user, crime statistics, weather—to be
reassembled via the mobile interface, which recalculates the horizon of possibilities on the
fly to reflect changes in real-time conditions.
1698 Environment and Planning A 48(9)

For the purposes of this discussion, what is important about these apps is how they
position themselves as material-discursive devices for enacting data-driven urban futures.
They are highly speculative in orientation, anticipating cities that are risky by virtue of being
socio-spatially stratified—around every city street loom threats to personal harm and safety
bound up in the possibility of undesirable urban encounter. Informed by this anticipatory
logic, these apps appropriate city spaces and denizens as data derivatives that may
subsequently be reassembled by a speculative calculus of risk, such as that which would
be operationalized by algorithmically routing pedestrians around ‘unsafe’ neighborhoods.
As proposed, the continuous (re)calculation of routes for travel described in Microsoft’s
Route Production patent application is an enactment of future-ing, of addressing normative
neoliberal subjects to self-securitize by actively participating in the algorithmic
governmentality of their personal mobilities so as to offset the hypothetical risks
associated with ‘unsafe’ urban encounters that loom on the horizon of possibilities. In the
second instance, as is characteristic of the crowdsourced nature of both SketchFactor and
Ghetto Tracker/Good Part of Town, these services assume hyper-responsibilized subjects
who not only willingly position themselves within the remit of a spatial algorithmic calculus
(of navigation, for example), but also actively contribute to it by generating highly affective,
emotional content productions in the form of ratings and commentaries about city
neighborhoods. This (re)enacts an urban geography in which risks such as ‘poor lighting,’
observations of ‘crime,’ and even the presence of bodies deemed unsavory (homeless persons,
residents of particular ethnic and racial backgrounds) can effectively be contained. Yet by
spatially containing purported threats of urban encounter, these neighborhood safety apps
would fail to actually enact any radically different city. In abstracting extant socio-spatial
disparities as derivatives to be projected onto the horizon of possible urban futures, the
algorithmic security calculus cannot divest itself of actual conditions on the ground,
anticipating a socio-spatially stratified city.

Securitizing the city

While the city is pervasively interfaced via mobile devices which digitally mediate individual
experiences of the urban, cities themselves also function as interfaces for data capture,
generation, circulation, and aggregation (Barreneche, 2012; de Waal, 2014). The anticipatory
ethos which underwrites speculative proposals for the containment of risks of urban encounter
similarly informs city-scale forms of anticipatory governmentality oriented towards securitizing
the city against the risks of city-wide social disorder latent in unstructured, non-curated data
flows generated by largely unaffiliated2 individuals participating on social media. When
subjected to practices of big data handling known as ‘sentiment analytics,’ these piecemeal
derivatives of urban emotion are algorithmically assembled to (supposedly) betray collective
affects of alienation, incitement to violence, and lack of confidence in law enforcement. Such
affects carry with them the threat of potentially materializing as contagious behaviors of urban
disruption including protest, rioting, and looting. Sentiment analytics operationalizes a synthetic
distillation of the emotional pulse of a city to inform a speculative anticipation of pre-futures of
urban anarchy immanent in codified expressions of civic strife that can be mapped onto an
urban geography. Neighborhoods or boroughs determined to be teetering on the verge of
descent into disorder are effectively rendered ripe for various kinds of interventions—such as
heightened surveillance and increased police presence—designed to preempt the materialization
of undesirable urban futures.
Sentiment analytics are part of a group of approaches to operationalizing big data
referred to as ‘predictive analytics,’ which describes a ‘‘variety of statistical, modeling,
Leszczynski 1699

data mining, and machine learning techniques to study recent and historical data, thereby
allowing analysts to make predictions about the future’’ (Bertolucci, 2013: n.p.). These
anticipated futures may then be shaped and/or preempted through combinations of
formative (suggesting book titles to Amazon customers) and preventative (stocking up on
flu vaccine) interventions that enable or forestall the unfolding of unfavorable/favorable
scenarios (additional purchases, seasonal disease epidemics). ‘Sentiment analytics’
specifically are predictive-analytics based systems that interface social media platforms
through their public streaming APIs and capture content as it is being generated by
individuals in real-time in order to model a ‘‘population’s behavior (opinions and
sentiments) so as to produce geographical risk alerts’’ of threats to public order
(Barreneche, 2012: 215; The Economist, 2012). While several such analytics suites have
been developed (see Barreneche, 2012; The Economist, 2012), a particularly illustrative
example is the British sentiment analytics engine EMOTIVE (Extracting the Meaning
of Terse Information in a Geo-Visualization of Emotion). EMOTIVE was designed to
monitor and map shared atmospheres of affect as expressed through individualized
contributions to the social media platform Twitter (EMOTIVE, 2015; Press Association,
2013; Sykora et al., 2013).
EMOTIVE processes up to 2,000 Tweets per minute through a lexicon-based natural
language processing (NLP) algorithm that matches emoticons and strings of words
extracted from sparse, unstructured texts to a limited set of possible emotions—including
anger, disgust, fear, happiness, sadness, loneliness, surprise, shame, and
inadequateness—mapped in a formal ontology (EMOTIVE, 2015; Marsden, 2013; Sykora
et al., 2013). Once filtered through the NLP pipeline, EMOTIVE subsequently processes
Tweets through a geolocation module that scrapes them for spatial referents, including
coordinates (for natively geocoded Tweets) and other geographical metadata (EMOTIVE,
2015; Marsden, 2013). This allows emotions distilled from Tweets to be spatially aggregated
in dynamic ‘mood maps’ of UK cities that can be used by law enforcement to inform
preemptive measures for suppressing the organization and social contagion of public
expressions of strife and alienation (for example, blocking off access routes into/out of
neighborhoods as an anti-kettling tactic) (EMOTIVE, 2015; Marsden, 2013: n.p.; Press
Association, 2013).
EMOTIVE is a proof-of-concept. But as with neighborhood safety apps introduced in the
previous section, what matters for the purposes of this discussion is not so much how many
law enforcement agencies have incorporated this particular sentiment analytics engine within
predictive policing practices. What is significant is the way in which EMOTIVE is a proof-of-
concept for operationalizing a preemptive algorithmic calculus for securitizing the city that is
axiomatic in its underlying logic and anticipatory in scope. The instrumentalist utility of the
ability to tap into social media streams in real-time is that it extends an ability to intervene
not only in the future (by preempting it in the present), but also somehow in a data-
determined pre-future, before the contagion of data-latent disaffection may be projected
forward in time and space. Data-latent contagion here refers to the propensity of digital
phenomena to go ‘viral’—i.e., to be suddenly and perhaps without precedent be widely
shared by users across digital platforms to the extent that it reaches the status of cultural
reference, even if only ephemerally. This preoccupation with contagion and the risks of its
materialization as disorganized/organized behaviors branded ‘anti-social’ informs not only
EMOTIVE, but also other sentiment analytics platforms, such as Condor, an earlier
example designed specifically for modeling protest dynamics (The Economist, 2012).
Condor factors in the reputation, or clout, of Twitter users in determining the likelihood
and viability of public demonstrations. If negative sentiments and/or calls for collective
1700 Environment and Planning A 48(9)

action are shared by users with a large number of followers, they are likely to be re-tweeted
(spread) with greater frequency and rapidity, having increased influence in incentivizing
demonstration, civic disobedience, etc. (The Economist, 2012).
EMOTIVE seeks to securitize the city against risks of the contagion of negative social
affects such as anger and disgust which carry with them the threat of social disintegration,
even where such risks exist only in speculative form as codified urban derivatives abstracted
from performative content productions reduced to a formal ontology of emotion. Yet
although the EMOTIVE security calculus is anticipatory in its orientation, it cannot
divorce itself from the historical and enduring socio-spatial inequalities that are
anticipated by its algorithmic assembly of cities-to-come. This is because the very
inception of the platform is informed by uneven social geographies that are always being
negotiated and contested in any public expressions and demonstrations of discontent, such
as organized protests. The researchers behind the design of the EMOTIVE platform detail
having developed it in the wake of the 2011 protests and subsequent riots which spread
through socioeconomically deprived and racial/ethnic minoritized boroughs of London and
outwards to other UK cities (EMOTIVE, 2015). They identified the ‘‘urban disturbances’’ of
the ‘‘fast-moving events of Summer 2011,’’ and the ‘‘key role played by social networks’’ in
spurring the spread of socially disruptive behaviors as an event horizon precipitating the
development of a sentiment analytics suite to help both city police departments and national
security agencies ‘‘to predict and monitor selected events’’ across UK cities (EMOTIVE,
2015: n.p.). Significantly, the ‘urban disturbances’ referred to in the framing of the
EMOTIVE analytics suite originated in response to the police shooting of an unarmed
Black man at point-blank range in what marked the climax of an escalation of simmering
tensions between the police and ethnic-minoritized communities (Lewis, 2011; Newburn
et al., 2011). At the time, law enforcement and the press identified mobile social
media—in particular Twitter, BlackBerry Messenger, and Facebook—as having enabled
the coordination of rioting and subsequent looting and vandalism (see for example
Adams, 2011; Press Association, 2013).
Insofar as EMOTIVE ‘looks to the future,’ it is predicated on an imaginary of cities as
tending towards disorder, where disorder itself is a function of socio-spatially stratified
urban geographies. Riots, for example, are axiomatically assumed as likely to occur in
certain kinds of (socioeconomically deprived) areas, and to be incentivized and performed
by certain kinds of (Black, Asian, student, White working class) bodies that are seen to
normatively inhabit those spaces. As a mode of algorithmic governmentality oriented
towards the securitization of city spaces, EMOTIVE presumes that the urban derivatives
(geo-located Tweets distilled to an emotional ontology) projected onto the array of possible
futures already carry with them the latency of descent into anarchy, and that this tendency to
disorder can be isolated and spatially aggregated by its security calculus (the NLP pipeline
and geo-location module). But as with the neighborhood safety apps, the urban derivative
cannot be divorced from stratified urban geographies actually existing on the ground.

Conclusion
Cities are constituted by their data shadows (Graham, 2013). These data do not merely
emanate into the ether, but are actively enrolled within contemporary modes of urban
governance which privilege data as a fixture and fixation of ordering and organizing
persons, places, objects, events, and flows within and across city spaces (Shelton et al.,
2015). To date, much of the discussion of the binding of data and urban governance in
the literature has been framed in terms of maximizing efficiencies in the ‘real-time’
Leszczynski 1701

smart city.3 In this paper, I suggest that in addition to understanding big data and urban
governance in terms of automated management in ‘actually existing’4 smart cities, we may
further make sense of the appropriation of urban (big) data within modes of algorithmic
governance/governmentality as a project of future-ing. This provides a means of making
sense of the intensifying relationship between cities, governance, and data beyond the sole
discursive rubric of smart urbanism by accounting for the longer-term temporalities and
security preoccupations of data-driven algorithmic anticipations of cities-to-come.
The actually existing city and the anticipated future city are of course co-constitutive. The
extensive networks of sensors seamlessly distributed across cityscapes which allow for real-
time monitoring and synchronicity simultaneously serve as a highly effective urban
surveillance grid; efficiency and security are not antithetical but rather co-implicated.
In this paper, I foreground security over efficiency to emphasize the ways in which modes
of urban algorithmic governance (as governmentality) are operationalized as a means of
controlling for the unfolding of undesirable urban futures by affecting particular kinds
of subjects and spaces that redirect, deflect, and spatially contain imminent risks of
contemporary urban life latent in urban derivatives. As I have introduced it in this paper,
the ‘urban derivative’ parallels Amoore’s (2011, 2014) theorization of data derivatives as the
abstraction of individuals into decontextualized encodings which serve as functional inputs
to speculative calculi that anticipate particular kinds of subjects who may themselves be
positioned (or position themselves) within the remit of said calculi. Similarly, the ‘urban
derivative’ designates the disassembly of cities into discrete codifications of places, denizens,
flows, and events reassembled across data flows via an algorithmic calculus that speculates
on the imminence of particular kinds of city-assemblages that loom on the horizon of
possibility.
I have suggested that such speculative modes of urban algorithmic securitization may be
most immediately evidenced at two ends of a continuum of spatial scales of digital practice:
that of the individual urbanite body, and that of the city itself. Drawing on the selective
example of neighborhood safety apps such as that described in Microsoft’s patent filing for a
pedestrian navigation app, algorithmic governmentality of the individual body is
operationalized via a routing algorithm that effects the spatial containment of risks of
urban encounter. The risks of encountering the racialized other in socioeconomically
deprived neighborhoods is technologically ameliorated by an algorithm that
circumnavigates these areas. The speculative design of Microsoft’s Pedestrian Route
Production service axiomatically presumes normative, risk-averse neoliberal subjects who
will self-securitize—i.e., participate in their own algorithmic self-governmentality—by
intentionally adopting and utilizing the service as a means of mitigating any threats to
their safety associated with walking through ‘unsafe’ neighborhoods. Other instances
participating in the ‘neighborhood safety apps’ class, such as SketchFactor and Ghetto
Tracker/Good Part of Town, assume hyper-responsibilized neoliberal subjects who further
contribute to the security calculus by contributing ratings of urban neighborhoods to these
platforms. These contributions (reviews and commentaries) are the definition of the urban
derivative—decontextualized by virtue of being contributed by individuals whose (race,
gender, socioeconomic) positionality is likely obfuscated, and aggregated in the form of
seemingly neutral summaries of the relative safety of different urban areas.
Elsewhere, practices of algorithmic governmentality as a mode of securitizing the city
against various risks of urban disorder do not necessarily depend on active crowdsourcing
but may instead repurpose user-generated content by tapping into continuous flows of
piecemeal, non-curated data from social media (Facebook likes, Tweets, etc.). This
method is evident in urban sentiment analytics approaches to preemptive urban
1702 Environment and Planning A 48(9)

securitization, which in this paper I have discussed with reference to EMOTIVE, a proof-of-
concept platform for generating real-time ‘mood maps’ of UK cities. The platform consists
of a NLP pipeline through which Tweets, extracted through Twitter’s public streaming API,
are processed in real-time at a rate of 2,000 per minute. The NLP algorithmic reduces
unstructured strings of texts and emoticons in the content of a Tweet to an ontology of
discrete emotions which are then geolocated on the basis of geographical referents and Tweet
metadata. The speculative utility of EMOTIVE is latent in its purported ability to distill the
emotional pulse of a city, which can inform various kinds of material interventions—such as
increased surveillance or stepped-up police presence—enacted to preempt and spatially
contain not only risks of urban anarchy bound up in collective expressions of strife and
disaffection, but the very contagion of sentiments deemed negative.
EMOTIVE is furthermore interesting insofar as it is demonstrative of the ways in which
corporate and state forms of governance (and governmentality) are increasingly difficult to
separate as data and security have become progressively imbricated in an ever-intensifying
‘(big) data-security assemblage’ which privileges and enables data-driven forms of
securitization (Aradau and Blanke, 2015). EMOTIVE may have been designed for state
actors (urban law enforcement) as end users, but its operationalization is entirely
dependent on the presence of public entities (social media companies) whose products
generate flows of content made available to third parties through public use of their APIs.
This is a defining feature of emergent modes of state-enacted signals intelligence activities
crystallizing around big data, of which the public has been made aware of through the
release of the Snowden documents and subsequent revelations (Amoore and Piotukh,
2015; Aradau, 2015; Crampton, 2015; Leszczynski, 2015; Lyon, 2014). These include the
interception of deep-sea fiber-optic cables owned and operated by commercial interests, the
mass collection and retention of communications metadata from telecom firms, and more
recent (though to date unsuccessful) efforts of some securities services at breaking mobile
operating system encryption.
These intricacies notwithstanding, following Aradau (2015), the ‘signature’ of big data
security is that it is anticipatory in scope—securitization is operationalized by speculating on
data-driven futures assembled across content flows and rendered actionable in the present.
As modes of algorithmic governmentality deployed within and across city spaces assemble
the city by appropriating data derivatives within a calculus of risk, what kind of a city do
they anticipate? In order to answer this question, I have examined how the anticipatory ethos
of big data security informs enactments of algorithmic governmentality that address,
respectively, both individual bodies in cities and city spaces themselves. In the instance of
Microsoft’s patent for a speculative pedestrian navigation app, the design of the service is
predicated on an urban imaginary of city futures that are always-already raced and classed,
and spatially segregated along those lines. These racially and socioeconomically stratified
urban geographies represent imminent risks of encounter that may be redirected for
privileged subjects by routing them around ‘unsafe’ neighborhoods.
Moving from the scale of the individual body to that of the city, the very design of the
sentiment analytics platform EMOTIVE is a response to uneven urban geographies. By
‘response,’ I do not mean that it represents an effort to ameliorate forms of urban
segregation. Rather, it is a technologically solutionist reaction that betrays a commitment
to an imaginary of the city as a socio-spatial assemblage always teetering on the brink of
descent into disorder and disarray. The propensity towards urban anarchy is latent within
data derivatives which encode urban disaffections abstracted from highly individualized data
productions distilled to a formal ontology of emotion. Securitization is achieved by
rendering urban spaces algorithmically isolated as hotspots of civic strife ripe for
Leszczynski 1703

preemptive material interventions. Yet the calculus which informs this determination
understands disaffection as something fomented in particular kinds of urban (working
class, ethnic-minoritized) area by certain kinds of (student, classed, racialized) subjects.
Insofar as (urban) big data look to the future, it cannot but anticipate socio-spatially
stratified cities. This is because the security calculus is dependent on the urban derivative
which, contra Amoore’s (2011, 2014) theorization of the data derivative, is not indifferent to
actual conditions on the ground. The urban derivative abstracts extant inequalities into
algorithmic projections of possible urban futures, such that enactments of algorithmic
governmentality reassemble the city in its own geographically uneven image. As a project
of future-ing, big data in the service of security seeks to securitize the city against the future
by contending with the uncertainty posed by the future as a radically open set of possibilities.
The material realities of the ways in which uneven urban geographies propagate through and
are projected onto the array of possible futures via enactments of algorithmic
governmentality operationalized in and across city spaces destabilizes utopian narratives
of urban big data solutionism.
In the vein of such utopianism, Nigel Thrift (2014) has recently suggested that big data
will reduce urban poverty by making it visible (prominent as a trend in the data) and thereby
less amenable to obfuscation. There is evidence to support his conjecture about the visibility
of inequality. A recent study using anonymized mobile phone records for Rwanda found
that an individual’s digital footprints derived from mobile phone transactions—locations
where individuals are placing and receiving mobile calls—was a strong predictor of their
socioeconomic status, accurately reproducing maps of extant geographies of relative
socioeconomic deprivation (Blumenstock et al., 2015). At the same time, however, such
optimistic pronouncements ignore the deep-seated inequalities that are echoed in urban
big data productions themselves. While mobile devices and now smartphones are
ubiquitous presences on city streets worldwide, the adoption of and participation on
social platforms are led by youth and young adults (see Leszczynski, 2015 for a summary
of relevant reports). As such, elderly urban residents living below the poverty line are
unlikely to own a smartphone and participate in generating content about their everyday
mobilities, such that they ‘‘do not [necessarily] register as digital signals’’ in content flows
(Crawford, 2013; Crawford et al., 2014: 1667). This problem of (under)representation
plagues any attempts at algorithmically distilling real-time knowledge of the emotional
pulse of a city by scraping sentiment from social media data. Content generated across
social platforms is not only highly performative in nature (Wilson, 2015), but indeed has
been contributed by a highly select and self-selecting segment of any urban population. The
result is that determinations of real-time urban affects cannot be generalized to the scale of
the city as EMOTIVE purports to do because the samples from which any algorithmic
aggregations are made is simply not representative. This does not even begin to address
the polemics of the axiomatic assumption that the algorithmic distillation of an individual’s
emotional state from highly unstructured, decontextualized strings of text is valid, desirable,
or in any way meaningful or indeed ethical.
In other words, it is not only the outcomes of analytics that are questionable. The
analytics themselves are also fraught. Urban informatics may indeed allow ‘‘phenomena
like [urban] poverty [to] become real to more people’’ (Thrift, 2014: 1264). But as Bowker
(2014) points out, this is likely a best-case scenario. Socioeconomic disparities may certainly
be rendered more immediately material for those who embody privileged subjectivities and
inhabit privileged spaces of the city, but only when these trends in the data actively inform
policies that intervene in the production of urban inequality. This however is contingent on
the ability of policymakers to make sense of data. Data are easy enough to appropriate,
1704 Environment and Planning A 48(9)

but valid analysis and meaningful interpretation of results are another matter. The coder
behind some of the US National Security Agency’s (NSA) analytics suites has for example
publicly stated that the agency’s fixation on intercepting, amassing, and collecting vast
swathes of data is actually hindering its surveillance efforts; the agency is simply awash in
too much data (Angwin, 2013).
It is for this reason that Bowker (2014) suggest that in the worst case scenario, ‘‘and most
commonly,’’ big data may deny that urban inequalities exist in refusing ‘‘that there are
indeed broad social forces’’ by rendering what are outcomes of intense, continuous
negotiations, and contestations of urban geographies to be the neutral outcomes of big
data analytics in which data have spoken for themselves (1797). Bowker is not
articulating a deterministic claim about the inherent deleterious social consequences of big
data. It is not big data per se but the logics and intentions which inform how data are
appropriated and operationalized, and to what societal ends, that is important. These
may either effect the reproduction of socio-spatial inequalities or disrupt them. Examples
of the latter do exist, such as the nonprofit Invisible Institute’s5 recent release of a
comprehensive dataset of 56,000 reports of police misconduct involving 8,500 Chicago
police officers over a five-year period (Arthur, 2015). Assembled as part the Citizens
Police Data Project, this dataset inverts the power imbalance inherent in practices of
predictive policing by providing Chicago residents with a rich data resource for predicting
incidences of bad policing (racial profiling, over-policing, physical aggression and violence)
in their city (Arthur, 2015).
The arguments made in this paper are not that urban big data are inherently pernicious
nor that widening divides between the urban rich and poor would be absent if not for
enactments of algorithmic governmentality latent in a patent application for a speculative
pedestrian routing service or a proof-of-concept for a sentiment analytics platform. Rather,
the argument is about the very material nature of urban inequalities and the persistence of
their geographies, both actual and anticipated, from which the urban derivative cannot be
divested. The effect is that as urban big data looks to the future through the lens of an
algorithmic security calculus, it seeks to securitize contemporary urban life against the risks
of uncertainty latent in the radical openness of possibilities, which include amongst them
scenarios of personal harm, anarchy, and disorder. In so doing, it projects the certainly of
what is—the characteristic unevenness of the contemporary social geography of
cities—forward in time and space, anticipating equally uneven cities-to-come.

Acknowledgements
I am most grateful to Rob Kitchin for extending an invitation to present this work at the September
2014 Code & The City workshop in Maynooth, Ireland, as well as to the other participants for their
generous feedback. I further appreciate the thoughtful comments and suggestions of the Reviewers
who took the time to engage with the ideas expressed herein.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or
publication of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Leszczynski 1705

Notes
1. Burrows and Ellison (2004).
2. Individuals not linked by any social ties or connections, as expressed through their social graphs.
3. Kitchin (2014b).
4. Shelton et al. (2015).
5. http://invisible.institute/police-data/

References
Adams WL (2011) Were Twitter or BlackBerrys used to fan flames of London’s riots? Time. Online), 8
August. Available at: http://content.time.com/time/world/article/0,8599,2087337,00.html (accessed
3 December 2015).
Amoore L (2007) Vigilant visualities: The watchful politics of the war on terror. Security Dialogue 38:
215–232.
Amoore L (2011) Data derivatives: On the emergence of a security calculus for our times. Theory,
Culture & Society 28: 24–43.
Amoore L (2013) The Politics of Possibility: Risk and Security Beyond Probability. Durham, NC: Duke
University Press.
Amoore L (2014) Security and the claim to privacy. International Political Sociology 8: 108–112.
Amoore L (2016) History, algorithm, ethics. Paper presented at the Association of American
Geographers Annual Meeting, 29 March–02 April, San Francisco, CA.
Amoore L and Hall A (2009) Taking people apart: Digitised dissection and the body at the border.
Environment and Planning D: Society and Space 27: 444–464.
Amoore L and Piotukh V (2015) Life beyond big data: Governing with little analytics. Economy and
Society 44: 431–366.
Anderson B (2006) Becoming and being hopeful: Towards a theory of affect. Environment and Planning
D: Society and Space 24: 733–752.
Anderson B (2007) Hope for nanotechnology: Anticipatory knowledge and the governance of affect.
Area 39: 156–165.
Anderson B (2010) Preemption, precaution, preparedness: Anticipatory action and future geographies.
Progress in Human Geography 34: 777–798.
Anderson B (2014) Encountering Affect: Capacities, Apparatuses, Conditions. Farnham, Burlington: Ashgate.
Angwin J (2013) NSA struggles to make sense of flood of surveillance data. The Wall Street Journal. .
25 December. Available at: http://www.wsj.com/articles/SB10001424052702304202204579252022
823658850 (accessed 20 December 2015).
Aradau C (2015) The signature of security: Big data, anticipation, surveillance. Radical Philosophy 191:
21–28.
Aradau C and Blanke T (2015) The (big) data-security assemblage: Knowledge and critique. Big Data
& Society. July–December: 1–12. DOI: 10.1177/2053951715609066.
Arthur R (2015) How to predict bad cops in Chicago. Five Thirty Eight Politics. 15 December.
Available at: http://fivethirtyeight.com/features/how-to-predict-which-chicago-cops-will-commit-
misconduct/ (accessed 19 December 2015).
Barreneche C (2012) Localizing the media, locating ourselves: A critical comparative analysis of socio-
spatial sorting in locative media platforms (Google and Flickr 2009–2011). Doctoral Thesis,
University of Westminster.
Bertolucci J (2013) Big data analytics: Descriptive vs. predictive vs. descriptive. Information Week. 31
December. Available at: http://www.informationweek.com/big-data/big-data-analytics/big-data-an
alytics-descriptive-vs-predictive-vs-prescriptive/d/d-id/1113279 (accessed 16 December 2015).
Biddle S (2014) Smiling young white people make app for avoiding black neighborhoods. Valleywag. 7
August. Available at: http://valleywag.gawker.com/smiling-young-white-people-make-app-for-avo
iding-black-1617775138 (accessed 14 December 2015).
Blumenstock J, Cadamuro G and On R (2015) Predicting povety and wealth from mobile phone data.
Science 50: 1073–1076.
1706 Environment and Planning A 48(9)

Bowker GC (2014) The theory/data thing. International Journal of Communication 8: 1795–1799.

boyd d and Crawford K (2012) Critical questions for big data. Information, Communication & Society
15: 662–679.
Burrows R and Ellison N (2004) Sorting places out? Towards a social politics of neighborhood
informatization. Information, Communication & Society 7: 321–336.
Cheney-Lippold J (2011) A new algorithmic identity: Soft biopolitics and the modulation of control.
Theory, Culture & Society 28: 164–181.
Crampton JW (in preparation) Assemblage of the vertical: Commercial drones and algorithmic life.
Geographica Helvetica. Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id¼2693625
(accessed 30 November 2015).
Crampton JW (2015) Collect it all: National security, big data and governance. GeoJournal 80: 519–531.
Crawford K (2013) The hidden biases in big data. HBR Blog Network. 1 April. Available at: http://
blogs.hbr.org/2013/04/the-hidden-biases-in-big-data/ (accessed 6 October 2014).
Crawford K, Miltner K and Gray ML (2014) Critiquing big data: Politics, ethics, epistemology.
International Journal of Communication 8: 1663–1672.
de Goede M and Randalls S (2009) Precaution, preemption: Arts and technologies of the actionable
future. Environment and Planning D: Society and Space 27: 859–878.
de Souza e Silva A and Frith J (2012) Mobile Interfaces in Public Spaces: Locational Privacy, Control,
and Urban Sociability. New York, NY: Routledge.
de Waal M (2014) The City as Interface: How New Media Are Changing the City. Rotterdam: nai010
publishers.
Dodge M, Kitchin R and Zook M (2009) Guest editorial: How does software make space? Exploring
some geographical dimensions of pervasive computing and software studies. Environment and
Planning A 41: 1283–1293.
EMOTIVE (2015) Emotive. Available at: http://emotive.lboro.ac.uk/ (accessed 20 December 2015).
Foucault M (1991) Governmentality. In: Birchell G, Gordon C and Miller P (eds) The Foucault Effect:
Studies in Governmentality. Chicago: University of Chicago Press.
Graham M (2013) The virtual dimension. In: Acuto M and Steele W (eds) Global City Challenges:
Debating A Concept, Improving The Practice. London: Palgrave.
Graham M and Zook M (2013) Augmented realities and uneven geogrpahies: Exploring the geo-
linguistic contours of the web. Environment and Planning A 45: 77–99.
Graham SDN (2005) Software-sorted geographies. Progress in Human Geography 29: 562–580.
Haggerty KD and Ericson RV (2000) The surveillant assemblage. British Journal of Sociology 51:
605–622.
Hecht B and Stephens M (2014) A tale of cities: Urban biases in volunteered geographic information.
In: Adar E and Resnick P (eds) Proceedings of the eighth international AAAI conference on weblogs
and social media. Palo Alto: AAAI Press.
Introna LD (2016) Algorithms, governance, and governmentality: On governing academic writing.
Science, Technology, & Human Values 41: 17–49.
Kinsley S (2012) Futures in the making: Practices to anticipate ‘ubiquitous computing’. Environment
and Planning A 44: 1554–1569.
Kitchin R (2014a) The Data Revolution: Big Data, Open Data, Data Infrastructures and Their
Consequences. London: SAGE.
Kitchin R (2014b) The real-time city? Big data and smart urbanism. GeoJournal 79: 1–14.
Kitchin R (2015) Continous geosurveillance in the ‘‘Smart City’’. DIS Magazine. February. Available at:
http://dismagazine.com/issues/73066/rob-kitchin-spatial-big-data-and-geosurveillance/ (accessed 21
February 2015).
Kitchin R and Dodge M (2011) Code/Space: Software and Everyday Life. Cambridge, MA: The MIT Press.
Kitchin R and Lauriault TP (2014) Towards critical data studies: Charting and unpacking data
assemblages and their work. In: Eckert J, Shears A and Thathcher J (eds) The Geoweb and Big
Data. Omaha: University of Nebraska Press.
Kitchin R, Lauriault TP and McGardle G (2015) Knowing and governing cities through urban
indicators, city benchmarking and real-time dashboards. Regional Studies, Regional Science 2: 6–28.
Leszczynski 1707

Leszczynski A (2015) Spatial big data and anxieties of control. Environment and Planning D: Society
and Space 33: 965–984.
Lewis P (2011) ‘‘Tottenham riots: A peaceful protest, then suddenly all hell broke loose’’. The
Guardian. 7 April. Available at: http://www.theguardian.com/uk/2011/aug/07/tottenham-riots-pea
ceful-protest (accessed 6 July 2014).
Lyon D (2014) Surveillance, snowden, and big data: Capabities, consequences, critique. Big Data &
Society. July–December: 1–13. DOI: 10.1177/2053951714541861.
Marsden R (2013) Emotive: Mapping the mood of a nation. The Business of Emotions. 9 October.
Available at: http://businessofemotions.org/2013/10/09/emotive-mapping-the-mood-of-a-nation/
(accessed 28 July 2014).
Massumi B (2007) Potential politics and the primacy of preemption. Theory & Event. 10. Available at:
https://muse.jhu.edu/journals/theory_and_event/v010/10.2massumi.html (accessed 22 November
2015).
Mattern S (2014) Interfacing urban intelligence. Places. April. Available at: https://placesjournal.org/
article/interfacing-urban-intelligence/ (accessed 28 April 2016).
Mattern S (2015) Mission control: A history of the urban dashboard. Places. March. Available at:
https://placesjournal.org/article/mission-control-a-history-of-the-urban-dashboard/ (accessed 29
April 2016).
Mayer-Schonberger V and Cukier K (2013) Big Data: A Revolution that will Transform How We Live,
Work and Think. London: John Murray.
McCann E (2011) Veritable inventions: Cities, policies and assemblage. Area 43: 143–147.
McCann E and Ward K (2011) Cities assembled: Space, neoliberalization, (re)territorialisation, and
comparison. In: McCann E and Ward K (eds) Mobile Urbanism: Cities and Policymaking in the
Global Age. Minneapolis: University of Minnesota Press.
McEnry T (2014) New app will keep you away from ‘sketchy’ areas. Crain’s. 07 August. Available at:
http://www.crainsnewyork.com/article/20140807/TECHNOLOGY/140809904/new-app-will-keep-
you-away-from-sketchy-areas (accessed 8 August 2014).
McFarlane C (2011a) Assemblage and critical urbanism. City 15: 204–224.
McFarlane C (2011b) The city as assemblage: Dwelling and urban space. Environment and Planning D:
Society and Space 29: 649–671.
McGuirk P (2011) Geographies of urban politics: Pathways, intersections, inventions. Geographical
Research 50: 256–268.
Monahan T (2008) Editorial: Surveillance and inequality. Surveillance & Society 5: 217–226.
Narula SK (2013) The real problem with a service called ‘ghetto tracker’. The Atlantic. 6 September.
Available at: http://www.theatlantic.com/technology/archive/2013/09/the-real-problem-with-a-ser
vice-called-ghetto-tracker/279403/ (accessed 7 September 2013).
Newburn T, Lewis P and Metcalf J (2011) A new kind of riot? From Brixton 1981 to Tottenham 2011.
The Guardian. 9 December. Available at: http://www.theguardian.com/uk/2011/dec/09/riots-1981-
2011-differences (accessed 6 July 2014).
Press Association (2013) New computer program analyses Twitter to map public sentiment. The
Guardian. 6 September. Available at: http://www.theguardian.com/technology/2013/sep/06/comp
uter-program-twitter-public-mood-emotive (accessed 6 July 2014).
Roche S (2014) Geographic information science I: Why does a smart city need to be spatially enabled?
Progress in Human Geography 38: 703–711.
Roche S (2016) Geographic information science II: Less space, more places in smart cities. Progress in
Human Geography 40: 565–573.
Sadowski J and Pasquale F (2015) The spectrum of control: A social theory of the smart city. First
Monday 20. Available at: http://firstmonday.org/ojs/index.php/fm/article/view/5903/4660 (accessed
29 November 2015).
Shelton T, Zook M and Wiig A (2015) The ‘actually existing smart city’. Cambridge Journal of Regions,
Economy and Society 8: 13–25.
Sutko DM and de Souza e Silva A (2010) Location-aware mobile media and urban sociability.
New Media & Society 13: 807–823.
1708 Environment and Planning A 48(9)

Sykora M, Jackson TW, O’Brien A, et al. (2013) EMOTIVE ontology: Extracting fine-grained
emotions from terse, informal messages. IADIS International Journal on Computer Science and
Information Systems 8: 106–118.
Tashev IJ, Couckuyt JD, Black NW, et al. (2012) Pedestrian route production. Microsoft Corporation.
Available at: https://www.google.com/patents/US8090532 (accessed 6 December 2015).
Taylor AS, Lindley S, Regan T, et al. (2014) Data and life on the street. Big Data & Society.
July–December: 1–7. DOI: 10.1177/2053951714539278.
Thatcher J (2013) Avoiding the Ghetto through hope and fear: An analysis of immanent technology
using ideal types. GeoJournal 78: 967–980.
The Economist (2012) The science of civil war: What makes heroic strife. The Economist. 21 April.
Available at: http://www.economist.com/node/21553006 (accessed 11 July 2014).
Thrift N (2014) The promise of urban informatics: Some speculations. Environment and Planning A 46:
1263–1266.
Wilson MW (2015) Morgan Freeman is dead and other big data stories. Cultural Geographies 22:
345–349.