1. Software-Sorted Cities
The New Architectures of Inequality
Stephen Graham
Newcastle University

2. Recap: *Infrastructure
“forgotten, the
background, the frozen in
place” (Leigh-Star)
* Splintering Urbanism
and the
Architectures of
Infrastructure

3. I Introduction:
• ”The modern city exists as a haze of
software instructions (Amin Thrift)
• Computer software mediates,
saturates and sustains contemporary
capitalist societies.
• Enrolled into complex technoscientific
systems, stretched across time-space
• Vast universe of code provides the
hidden “calculative background” to
the functioning and ordering of such
societies
• This “background” has become
the ‘ordinary’ socio-technical world

4. • Ubiquitous, pervasive, interlinked spaces, systems, and
equipment
• Link databases to sensor,
tracking, logistics and access
control systems
• Continuously classify,
standardise, and demarcate
rights, privileges, inclusions,
exclusions, mobilities and
normative social judgements
across multiple scales
• Geographies of exclusion are
now performed through the
agency of software algorithms
Software-Sorting:
The Agency of
Algorithms

5. • Socio-technical architectures of
software-sorting are tending to
blend into the wider urban
environment
• “Surveillant assemblages”
embedded into infrastructures and
urban spaces: Transponders, RFID
Chips, ‘smart’ cards, biometrics,
mobile handsets etc
• Urban everyday life becomes series
of obligatory passage points policed
by software-sorting systems (with
proliferating codes, identifiers, scans,
passwords)
‘The Most Profound
Technologies
Are Those
That Disappear’

6. Software-Sorting:
A New Technological Politics?
• Software challenges us to
understand new forms of
technologies politics and new
practices of political invention,
legibility and intervention that
we are only beginning to
comprehend as political at
all (Thrift and French, 2002)

7. • Helping to facilitate neoliberal
infrastructural transformations
• Undermine barriers to
recommodification, mass-customisation,
Software-
and individualisation
Sorting and
• From standardised service regimes based
Neoliberal
on generalised tariffs or universal service
Transformations
obligations and supply monopolies
• To complex infrastructural marketplaces
where each user is surveilled, tracked and
treated differently based on normative
judgments of their fitness, worth or
profitability

8. III Code Space: Software-Sorted Mobilities
• (a) Biometric IDs and Airport
Immigration Filtering
• Opt-in biometric bypass for elite
travellers combined with
tightened controls for those
identified by risk profiling
• the control of international
mobilities that cross through
airports and border zones are
effectively managed, filtered and
screened within these sites (Peter
Aday, 2004),
• Politics of differential speed and
anticipatory profiling

9. • To Peter Aday, this “facilitates the ease
of speed for trusted, ‘good’ and
economically sound business travellers
and yet impede the flow of ‘bad guys’
or secondary processing – where
officers ‘really don’t care how long it
takes’ to process their entry
• Aday concludes, the airport is now a
surveillance machine— an assemblage
where webs of technology and
information combine. Movement, and,
increasingly, the body, identity, and
objects are made legible, momentarily
fusing with technology and virtual
realism (2004).

10. (b) Road Pricing:
Unbundling Public Roadspace Monopolies:
• Most road-pricing (e.g.
London) still based on
standard tariffs at all times
• Rely on transponders,
automatic number plate
recognition, call centres,
cameras and databases
• Do allow differential
pricing
• London proposing to
charge big extra levy for
4x4s by 2009

11. The Quest for ‘Real-Time’ Road Pricing
• Software-sorting being introduced to
display variable pricing in real time -for example I-15 highway in San Diego
• This is based on algorithms which
estimate exactly the level of price per
journey that is likely to deter enough
drivers to guarantee free-flowing traffic
-- no matter how bad the congestion is
on the surrounding public highway
system.
• May also be built into proposal for EU
roadspace pricing using GPS
technology

12. (c ) The Internet: From ‘Best Effort’
to ‘Squelching the Scavenger Class’
• Internet originally developed to
accord all the ‘packets’ of
information that flowed within it
equal status. This was the so-called
‘best effort’ model of packet
switching
• Now, under pressure of congestion
and pressures to introduce
neoliberal service regimes, the entire
Internet is being reengineered into a
corporately controlled system of
systems dominated by a wide range
of commercial services

13. • Software-sorted Internet
systems allow a guaranteed
quality of service to premium
users and prioritised services,
even at times of major internet
congestion
• But those packets deemed
unprofitable will slowed down
or actually ‘dropped’
• This is likely to lead to a
dramatic deterioration in the
electronic mobilities of
marginalised users or nonprioritised services

14. • World’s largest provider of Internet Routers,
Cisco (2002), describe how premium internet
services can now be offered to what they call the
transactional/interactive data class of users, whilst,
at the same time, what they term the scavenger
class will be actively impeded based on softwaresorting of every single Internet packet.
• The Scavenger class [categorisation] is intended
to provide differential services, or ‘less-than-BestEffort’ services, to certain applications, the
document suggests. Applications assigned to this
class have little or no contribution to the
organizational objectives of the enterprize.
Assigning a minimal bandwidth queue to Scavenger
traffic forces it to be squelched to virtually nothing
during periods of congestion .

15. (d) Call Centres: The Politics of
Speed-Up and Slow-Down
• Call centres can detect the
telephone numbers of
incoming calls, and instantly
check these against
customer and
geodemographic databases
• Use software-sorting
techniques to queue ‘good
customers’ for shorter
times than ‘bad’
customers.

16. Initially, Mediated by Call Centre Operator
• A marketing brochure from the
Avaya Corporation (2000):
One of your best customers dials
the national customer service
number for your company. The ANI
[Automatic Number Identification]
database reveals the customer to be
among the top 5% of your
customers. [Our system] routes the
customer at high priority. When the
agent picks up the call, he hears a
whispered announcement that this
caller is ‘Top 5’

17. But Shifting to Automated Prioritisation
• Ian Davis, a customer relations manager at the
IT company ATG: “It’s all about finding out who
the customer is, and putting then in the correct
bucket. The unprofitable customers never hear
about the discounts and promotions”.
• Can be used to allocate scarce operator time.
The phone company Orange, for example,
allows immediate access to a human being only
to those users who sign up for a premium
‘Panther’ service. The Virgin call centre,
Thetrainline, deters first time callers with lengthy
interactive voice response menus whilst
prioritising regular, business, passengers for
tailored, human, support.

18. IV Code Place: Software-Sorted Cities
• Geosurveillance and
geotracking boom
(e.g. Radio Frequency
Identification Tags)
• Databases link to
Geographic
Information Systems
(GISs) to offer
different services to
different
neighbourhoods

19. Supports Geographical Unbundling of
Previously Standard Services/Prices
• E.g. Amazon has offered
different DVD prices to
different customers in USA
• UK train travellers now
access a labyrinth of different
tariffs and prices based, using
software-sorted web sites
and call centres, on when
they book their tickets, who
they are, and even where
they live.

20. RFIDs: The Politics of
Ubiquitous Tracking?
• Once RFID based ‘tracking’ becomes routine, software-sorting
techniques will move beyond crude, generalised, and ‘lumpy’
simulations of places, to personalised advertising and real-time
tracking
• In Japan, owners of malls or privatised public spaces are
experimenting with using RFID cards to identify each individual who
enters their realm covertly and automatically. May allow trackings of
their tastes, wealth, habits, associations and potential profitability.
• Could allow extra services and benefits to be offered to those
deemed most desirable whilst supporting attempts to remove or
discourage those deemed to be problematic
• So may work to “chill [speech, actions or assembly deemed]
irregular, deviant or unpopular” in such places (Kang and Cuff, 2005)

21. V Code Face: Software-Sorted Streets?
• Building on the massive and rapid
diffusion of analogue CCTV, which
relies on the (expensive) ‘MK1
eyeball’ of human operators, scanning
monitors and recording footage using
banks of domestic-style video
recorders, a major effort is now being
made to install much cheaper,
automated, facial recognition, or
‘event-driven,’ CCTV in the place of
such systems
• Here algorithms scan for ‘abnormal or
‘target’ events or people based on
software which monitors and learns
the putative ‘normal’ background

22. Face-Recognition CCTV
• Many technological problems inhibit FR CCTV on open city streets
(as opposed to airport passage points)
• But major R and D to address these
• ”Unlike other biometrics [facial recognition CCTV] can operate
anonymously in the background (Koskela, 2003)
• The code within the facial recognition system becomes a key
political site because its operation automatically stipulates the
subjects, locations or behaviours that are deemed by the
operators to be ‘abnormal’, ‘threatening’ and worthy of further
scrutiny or tracking.
• There are very real risks that the multiple ‘islands’ of private and
public CCTV systems, each monitored by its own human
operators, could quickly merge, or link, into much more massive
and geographically-stretched facial recognition CCTV systems.

23. The Politics of Facial Tracking
• Mitchell Gray (2003): As the technology advances, the clear
risk is that the software will effortlessly track individuals
moving through urban space, public and private. Any
appearance of a person deemed threatening can be set to
trigger an alarm, assuming that that person’s face has been
recorded in a linked database.
• Inherent biases. On one trail identification rates for males
were 6% to 9% points higher than for female. Recognition
rates for older people were higher than for younger
people (Introna, 2003).
• Also, the trial report states that Asians are easier [to
recognise] than whites, African-Americans are easier than
whites, [and] other race members are easier than
whites (FRVT, 2002)

24. VI Conclusions: The Agency of the Algorithm
• This paper has sought to underline the centrality of
software-sorting in structuring contemporary social and
geographical inequalities.
• It has also attempted to illustrate the need to maintain a
broad, multisectoral perspective which can capture how
different software-sorting techniques are encroaching
across different dimensions of contemporary societies.
• Clearly, though, much more detailed analyses on
software sorted assemblages in practice is needed

25. (i)
Software-Sorting: New ‘Digital Divides’?
(ii) Politics of Speed and (Relative)
Immobilisation
(iii) Challenges of Transparency
and Visibility