In 2011, Harvard’s John F. Kennedy School of Government and the National
Institute of Justice published a paper titled “Police Science: Toward a New
Paradigm,” the ideas of which were developed at the Executive Session on
Policing and Public Safety hosted at Harvard University. The paper calls for a
“radical reformation of the role of science in policing” that prioritizes
evidence-based policies and emphasizes the need for closer collaboration
between universities and police departments. In the opening paragraph, the
authors, David Weisburd and Peter Neyroud, assert that “the advancement of
science in policing is essential if police are to retain public support and
legitimacy.”
Given that critics of the police associate law enforcement with the arbitrary
use of force, racial domination, and the discretionary power to make decisions
about who will live and who will die, the rebranding of policing in a way that
foregrounds statistical impersonality and symbolically removes the agency of
individual officers is a clever way to cast police activity as neutral,
unbiased, and rational. This glosses over the fact that using crime data
gathered by the police to determine where officers should go simply sends
police to patrol the poor neighborhoods they have historically patrolled when
they were guided by their intuitions and biases.
This “new paradigm” is not merely a reworking of the models and practices used
by law enforcement, but a revision of the police’s public image through the
deployment of science’s claims to objectivity. As Zach Friend, the man behind
the media strategy of the start-up company PredPol (short for “predictive
policing”), noted in an interview, “it kind of sounds like fiction, but it’s
more like science fact.” By appealing to “fact” and recasting policing as a
neutral science, algorithmic policing attempts to solve the police’s crisis of
legitimacy.
The Crisis of Uncertainty
Whereas repression has, within cybernetic capitalism, the role of warding off
events, prediction is its corollary, insofar as it aims to eliminate all
uncertainty connected to all possible futures. That’s the gamble of statistics
technologies. Whereas the technologies of the Providential State were focused
on the forecasting of risks, whether probabilized or not, the technologies of
cybernetic capitalism aim to multiply the domains of responsibility/authority.
—Tiqqun, The Cybernetic Hypothesis [footnote Tiqqun, “The Cybernetic Hypothesis
(L’Hypothèse cybernétique),” Tiqqun 2 (2001): 21.]
Uncertainty is at once a problem of information and an existential problem that
shapes how we inhabit the world. If we concede that we exist in a world that is
fundamentally inscrutable for individual humans, then we also admit to being
vulnerable to any number of risks that are outside our control. The less “in
control” we feel, the more we may desire order. This desire for law and
order—which is heightened when we are made aware of our corporeal vulnerability
to potential threats that are unknowable to us—can be strategically manipulated
by companies that use algorithmic policing practices to prevent crime and
terrorism at home and abroad. Catastrophes, war, and crime epidemics may
further deepen our collective desire for security.
In the age of “big data,” uncertainty is presented as an information problem
that can be overcome with comprehensive data collection, statistical analysis
that can identify patterns and relationships, and algorithms that can determine
future outcomes by analyzing past outcomes. Predictive policing promises to
remove the existential terror of not knowing what is going to happen by using
data to deliver accurate knowledge about where and when crime will occur. Data
installs itself as a solution to the problem of uncertainty by claiming to
achieve total awareness and overcome human analytical limitations. As Mark
Andrejevic writes in Infoglut, “The promise of automated data processing is to
unearth the patterns that are far too complex for any human analyst to detect
and to run the simulations that generate emergent patterns that would otherwise
defy our predictive power.”
The anonymous French ultraleftist collective Tiqqun links the rise of the
crisis of uncertainty to the rise of cybernetics. Tiqqun describes
cybernetics—a discipline founded by Norbert Wiener and others in the 1940s—as
an ideology of management, self-organization, rationalization, control,
automation, and technical certitude. According to Tiqqun, this ideology took
root following World War II. It seeks to resolve “the metaphysical problem of
creating order out of disorder” to overcome crisis, instability, and
disequilibrium, which Tiqqun asserts is an inherent by-product of capitalist
growth. However, the “metaphysical” problem of uncertainty that is created by
crisis enables cybernetic ideology to take root. Drawing on Giorgio Agamben’s
State of Exception, Tiqqun writes, “The state of emergency, which is proper to
all crises, is what allows self-regulation to be relaunched.” Even though, by
nearly every metric, “Americans now live in one of the least violent times in
the nation’s history,” Americans believe that crime rates are going up.
Empirically, there is no basis for the belief that there is an unprecedented
crime boom that threatens to unravel society, but affective investments in this
worldview expand the domain of surveillance and policing and authorizes what
Manuel Abreu calls “algorithmic necropower.” The security state’s calculation
of risk through data-mining techniques sanctions the targeting of “threats” for
death or disappearance. Though the goal of algorithmic policing is, ostensibly,
to reduce crime, if there were no social threats to manage, these companies
would be out of business.
Whether or not we accept Tiqqun’s account of how capitalist growth generates a
metaphysical crisis that enables the installation of cybernetic governance, it
is clear that PredPol appeals to our desire for certitude and knowledge about
the future. UCLA anthropology professor Jeffrey Brantingham emphasizes, in his
promotion of PredPol, that “humans are not nearly as random as we think.”
Drawing on evolutionary notions of human behavior, Brantingham describes
criminals as modern-day urban foragers whose desires and behavioral patterns
can be predicted. By reducing human actors to their innate instincts and
applying complex mathematical models to track the behavior of these urban
“hunter-gathers,” Brantingham’s predictive policing model attempts to create
“order” out of the seeming disorder of human behavior.
Paranoia
But what does PredPol actually do? How does it actually work? PredPol is a
software program that uses proprietary algorithms (modeled after equations used
to determine earthquake aftershocks) to determine where and when crimes will
occur based on data sets of past crimes. In Santa Cruz, California, one of the
pilot cities to first use PredPol, the company used eleven years of local crime
data to make predictions. In police departments that use PredPol, officers are
given printouts of jurisdiction maps that are covered with red square boxes
that indicate where crime is supposed to occur throughout the day. Officers are
supposed to periodically patrol the boxes marked on the map in the hopes of
either catching criminals or deterring potential criminals from committing
crimes. The box is a kind of temporary crime zone: a geospatial area generated
by mathematical models that are unknown to average police officers who are not
privy to the algorithms, though they may have access to the data that is used
to make the predictions.
What is the attitude or mentality of the officers who are patrolling one of the
boxes? When they enter one of the boxes, do they expect to stumble upon a crime
taking place? How might the expectation of finding crime influence what the
officers actually find? Will people who pass through these temporary crime
zones while they are being patrolled by officers automatically be perceived as
suspicious? Could merely passing through one of the red boxes constitute
probable cause? Some of these questions have already been asked by critics of
PredPol. As Nick O’Malley notes in an article on PredPol, “Civil rights groups
are taking [this] concern seriously because designating an area a crime hot
spot can be used as a factor in formulating ‘reasonable suspicion’ for stopping
a suspect.”
When the Cleveland police officer Timothy Loehmann arrived on the scene on
November 22, 2014, it took him less than two seconds to fatally shoot Tamir
Rice, a twelve-year-old black boy who was playing with a toy gun. This raises
the question—if law enforcement officers are already too trigger-happy, will
the little red boxes that mark temporary crime zones reduce the reaction time
of officers while they’re in the designated boxes? How does labeling a space as
an area where crime will occur affect how police interact with those spaces?
Although PredPol conceptualizes the terrain that is being policed as a field
where natural events occur, the way that data is interpreted and visualized is
not a neat reflection of empirical reality; rather, data visualization actively
constructs our reality.
Furthermore, how might civilians experience passing through one of the boxes?
If I were to one day find myself in an invisible red box with an officer, I
might have an extra cause for fear, or at least I would be conscious of the
fact that I might be perceived as suspicious. But given that I am excluded from
knowledge of where and when the red boxes will emerge, I cannot know when I
might find myself in one of these temporary crime zones. Using methods that are
inscrutable to citizens who do not have access to law enforcement knowledge and
infrastructure, PredPol is remaking and rearranging the space through which we
move. That is the nature of algorithmic policing; the phenomenological
experience of policing is qualitatively different from “repressive” policing,
which takes place on a terrain that is visible and uses methods that can be
scrutinized and contested. Predictive policing may induce a sense of being
watched at all times by an eye we cannot see. If Jeremy Bentham’s
eighteenth-century design of the “panopticon” is the architectural embodiment
of Michel Foucault’s conception of disciplinary power, then algorithmic
policing represents the inscription of disciplinary power across the entire
terrain that is being policed.
False Positives
Given the difficulty of measuring the efficacy of predictive policing methods,
there is a risk of falsely associating “positive” law enforcement outcomes with
the use of predictive policing software such as PredPol. The literature on
PredPol is also fuzzy on the question of how to measure its success. When
police officers are dispatched to the five-hundred-by-five-hundred feet square
boxes marked in red on city maps, are they expected to catch criminals in the
act of committing crimes, or are they supposed to deter crime with their
presence? The former implies that an increase in arrests in designated areas
would be a benchmark of success, while the latter implies that a decrease in
crime is proof of the software’s efficacy. However, both outcomes have been
used to validate the success of PredPol. A news clip from its official YouTube
account narrates the story of how the Norcross Police Department (Georgia)
caught two burglars in the act of breaking into a house. Similarly, an article
about PredPol published on Officer.com opens with the following anecdote:
“Recently a Santa Cruz, Calif. police officer noticed a suspicious subject
lurking around parked cars. When the officer attempted to make contact, the
subject ran. The officer gave chase; when he caught the subject he learned he
was a wanted parolee. Because there was an outstanding warrant for his arrest,
the subject was taken to jail.”
Much of the literature PredPol uses for marketing offers similarly mystical
accounts of the software’s clairvoyant capacity to predict crime, and these are
substantiated by anecdotes about officers stumbling upon criminals in the act
of committing these crimes. However, PredPol consistently claims that its
efficacy can be measured by a decrease in crime. Yet across the country, crime
rates have been plummeting since the mid-1990s. In some cases, the company
tries to take credit for crime reduction by implying there is a causal
relationship between the use of PredPol and a decrease in crime rates,
sometimes without explicitly making the claim. In an article linked on
PredPol’s website, the author notes, “When Santa Cruz implemented the
predictive policing software in 2011, the city of nearly 60,000 was on pace to
hit a record number of burglaries. But by July burglaries were down 27 percent
when compared with July 2010.”
Yet crime rates fluctuate from year to year, and it is impossible to parse
which factors can be credited with reducing crime. Though the article does not
explicitly attribute the crime reduction to PredPol, it implicitly links the
use of PredPol to the 27 percent burglary reduction by juxtaposing the two
separate occurrences—the adoption of PredPol and the decrease in burglaries—so
as to construct a presumed causal relation. The article goes on to use
explanations made by Zach Friend (about why and how PredPol works) to validate
its efficacy. Friend is described as “a crime analyst with the Santa Cruz PD”;
however, Friend actually left the Santa Cruz Police Department to become one of
the main lobbyists for PredPol soon after the company was founded.
By scrutinizing the PR circuits that link researchers like UCLA’s Brantingham
to the police, and link Silicon Valley investors to the media, one realizes
that essentially all claims about the efficacy of PredPol loop back to the
company itself. Though PredPol’s website advertises “scientifically proven
field results,” no disinterested third party has ever substantiated the
company’s claims. What’s even more troubling is that PredPol offered 50 percent
discounts on the software to police departments that agreed to participate as
“showcase cities” in PredPol’s pilot program. The program required
collaboration with the company for three years and required police departments
to provide testimonials that could be used to market the software. For
instance, SF Weekly notes that the city of Alhambra, just northeast of Los
Angeles, purchased PredPol’s software in 2012 for $27,500. The contract between
Alhambra and PredPol includes numerous obligations requiring Alhambra to carry
out marketing and promotion on PredPol’s behalf. Alhambra’s police and public
officials must “provide testimonials, as requested by PredPol,” and “provide
referrals and facilitate introductions to other agencies who can utilize the
PredPol tool.”
In “The Difference Prevention Makes: Regulating Preventive Justice,” David Cole
describes five major risks that come with the adoption of the “paradigm of
prevention” in law enforcement. He notes that “it is not just that we cannot
know the efficacy of prevention; our assessments are likely to be
systematically skewed.” Others have raised similar concerns with PredPol.
According to O’Malley, “The American Criminal Law Review has raised concerns
the program could warp crime statistics, either by increasing the arrest rate
in the boxes through extra policing or falsely reducing it through diffusion.”
The Politics of Crime Data
Crime has never been a neutral category. What counts as crime, who gets labeled
criminal, and which areas are policed have historically been racialized.
Brantingham, the anthropologist who helped create PredPol, noted, “The focus on
time and location data—rather than the personal demographics of
criminals—potentially reduces any biases officers might have with regard to
suspects’ race or socioeconomic status.” Though it is true that PredPol is a
spatialized form of predictive policing that does not target individuals or
generate heat lists, spatial algorithmic policing, even when it does not use
race to make predictions, can facilitate racial profiling by calculating
proxies for race, such as neighborhood and location. Furthermore, predictive
models are only as good as the data sets they use to make predictions, so it is
important to interrogate who collects data and how it is collected. Although
data has been conceptualized as neutral bits of information about our world and
our behaviors, in the domain of criminal justice, it is a reflection of who has
been targeted for surveillance and policing. If someone commits a crime in an
area that is not heavily policed—such as on Wall Street or in the white
suburbs—it will fail to generate any data. PredPol’s reliance on the dirty data
collected by the police may create a feedback loop that leads to the
ossification of racialized police practices. Furthermore, when applied to
predictive policing, the idea that “more data is better,” in that it would
improve accuracy and efficiency, justifies dragnet surveillance and the
expansion of policing and carceral operations that generate data.
Though PredPol presents itself as race-neutral, its treatment of crime as an
objective force that operates according to laws that govern natural phenomena,
such as earthquake aftershocks—and not as a socially constructed category that
has meaning only in a specific social context—ignores the a priori
racialization of crime, and specifically the association of crime with
blackness. Historian Khalil Gibran Muhammad’s The Condemnation of Blackness:
Race, Crime and the Making of Modern America traces how “at the dawn of the
twentieth century, in a rapidly industrializing, urbanizing, and
demographically shifting America, blackness was refashioned through crime
statistics. It became a more stabilizing racial category in opposition to
whiteness through racial criminalization.”
Muhammad describes how data was used primarily by social scientists in the
North to make the conflation of blackness and criminality appear objective and
empirically sound, thus justifying a number of antiblack social practices such
as segregation, racial violence, and penal confinement. The consolidation of
this “scientific” notion of black criminality also enabled formerly
criminalized immigrant populations—such as the Polish, Irish, and Italians—to
be assimilated into the category of whiteness. As black Americans were
pathologized by statistical discourse, the public became increasingly
sympathetic to the problems of European ethnic groups, and white ethnic
participation in criminal activities was attributed to structural inequalities
and poverty, as opposed to personal shortcomings or innate inferiority.
According to Mohammad, the 1890 census laid much of the groundwork for this
ideology. He describes how statistics about higher rates of imprisonment among
black Americans, particularly in northern penitentiaries, were “analyzed and
interpreted as definitive proof of blacks’ true criminal nature.”Thus,
biological and cultural racism was eventually supplanted by statistical racism.
While the methods developed by PredPol themselves are not explicitly
racialized, they are implicitly racialized insofar as geography is a proxy for
race. Furthermore, given that crime has historically been racialized, taking
crime for granted as a neutral—or rather, natural—category around which to
organize predictive policing practices is likely to reproduce racist patterns
of policing. As PredPol relies on data about where previous crimes have
occurred, and as police are more likely to police neighborhoods that are
primarily populated by people of color (as well as target people of color for
searches and arrests), then the data itself that PredPol relies on is
systematically skewed. By presenting its methods as objective and racially
neutral, PredPol veils how the data and the categories it relies on are already
shaped by structural racism.
Conclusion
The story of policing in the twenty-first century cannot be reduced to the
stereotypical image of bellicose, meathead officers looking for opportunities
to catch bad guys and to flaunt their institutional power. As Donnie Fowler,
the PredPol director of business development, was quoted saying in the Silicon
Valley Business Journal, twenty-first-century policing could more accurately be
described as “a story about nerds and cops.”
However, more than a story of an unlikely marriage between data-crunching
professors and crime-fighting officers, the story of algorithmic policing, and
PredPol in particular, is also a story of intimate collaboration between
domestic law enforcement, the university, Silicon Valley, and the media. It is
a story of a form of techno-governance that operates at the intersection
between knowledge and power. Yet the numerical and data-driven approach
embodied by PredPol has been taken up in a number of domains. In both finance
and policing, there has been a turn toward technical solutions to the problem
of uncertainty, solutions that attempt to manage risk using complex and opaque
mathematical models. Yet, although the language of risk has replaced the
language of race, both algorithmic policing and risk-adjusted finance merely
code racial inequality as risk. It is important that we pay attention to this
paradigm shift, as once the “digital carceral infrastructure” is built up, it
will be nearly impossible to undo, and the automated carceral surveillance
state will spread out across the terrain, making greater and greater intrusions
into our everyday lives. Not only will the “smart” state have more granular
knowledge of our movements and activities, but as the carceral state becomes
more automated, it will increase its capacity to process ever-greater numbers
of people, even when budgets remain stagnant or are cut.
Though it is necessary to acknowledge the invisible, algorithmic (or
“cybernetic”) underside of policing, it is important to recognize that
algorithmic policing has not supplanted repressive policing, but is its
corollary. “Soft control” has not replaced hard forms of control. Police have
become more militarized than ever as a result of the $34 billion in federal
grants that have been given to domestic police departments by the Department of
Homeland Security in the wake of 9/11. While repressive policing attempts to
respond to events that have already occurred, algorithmic policing attempts to
maintain law and order by actively preventing crime. Yet is it possible that
the latter actually creates a situation that leads to the multiplication of
threats rather than the achievement of safety? As predictive policing practices
are taken up by local police departments across the country, perhaps we might
consider the extent to which, as Tiqqun writes, “the control society is a
paranoid society.”
×
https://www.e-flux.com/journal/87/169043/this-is-a-story-about-nerds-and-cops-predpol-and-algorithmic-policing/
This text is an excerpt from Carceral Capitalism by Jackie Wang, forthcoming
from Semiotext(e) in February 2018.
Jackie Wang is a student of the dream state, black studies scholar, prison
abolitionist, poet, performer, library rat, trauma monster, and PhD candidate
at Harvard University. She is the author of a number of punk zines including On
Being Hard Femme, as well as a collection of dream poems titled Tiny Spelunker
of the Oneiro-Womb (Capricious). She tweets at @loneberrywang and blogs at
loneberry.tumblr.com.
# distributed via <nettime>: no commercial use without permission
# <nettime> is a moderated mailing list for net criticism,
# collaborative text filtering and cultural politics of the nets
# more info: http://mx.kein.org/mailman/listinfo/nettime-l
# archive: http://www.nettime.org contact: [email protected]
# @nettime_bot tweets mail w/ sender unless #ANON is in Subject:
