IEEE Technology and Society Magazine - June 2019 - 46

sensitivity to its various implications and specific implementations, pursuing such intuitions blindly can quickly
lead to problems. Unfortunately, the strength of this
intuition and its simple logic make it an easy policy
argument to make in many institutional and bureaucratic settings. One might even argue that this is the
"default" policy argument in the age of data, and thus
Models of Threat is the default approach to predictive
policing. And it is safe to assume that without critical
reflection and active awareness on the part of systems
designers, something similar will be the likely default
goal of most AI systems. To better understand how the
design of systems can mitigate or exacerbate the problems inherent in data-driven management, we now turn
to two examples of predictive policing.

One City, Two Cases of Predictive Policing
The City of Chicago, IL, has seen a spike in gun violence
in recent years. The city has led the United States in the
number of shootings and gun homicides, peaking with
758 total homicides and more than 4300 shootings in
2016, and down slightly in 2017 [36]. This has led to a
serious effort by the Chicago Police Department (CPD)
to address this spike by focusing on neighborhoods and
individuals most likely to become involved in gun violence. A number of studies, experiments, and policies
have been tested and implemented in recent years. By
comparing different applications of data-driven interventions occurring in the same city at the same time period, we can develop insights into the implications of data
for shaping policing practices.
Two such experiments, in particular, offer a good
insight into the ways in which data can be applied to
address gun violence, and also into the ways that the
implementation and utilization of those insights can
have radically different social and ethical implications.
One has been the subject of critical scrutiny by journalists and researchers, called the Strategic Subjects List.
More often called the "heat list" by police officers, it was
first used by CPD in 2012, and its use continues, though
under a revised set of guidelines following criticism of
the early uses described here. The other started in the
summer of 2011 as a pilot research program implemented by the City of Chicago, and was studied the following year by University of Chicago researchers. Called
One Summer, it has since been adopted as an annual
program by the City of Chicago. While both started out
as academic research projects, both were analyzed by
outside researchers in 2012, and both utilized data to
assess and identify youth who are at-risk of being
involved in gun violence, in most other ways the two
programs are very different.
The two projects can best be characterized as illustrative case studies, embodying two different philosophies

of predictive policing, and perhaps two extremes thereof. They accordingly have very different ways of thinking
about what being an "at-risk" youth means, and consequently pursue very different approaches to intervening
so as to reduce that risk. More importantly, they also had
very different outcomes in terms of their effectiveness in
reducing gun violence and in influencing the life outcomes for those identified as "at-risk" in each program.
In short, the Strategic Subjects List can be described as
taking a "Models of Threat" approach to at-risk youth.
That is, at-risk youth in that project are primarily viewed
as threats to the community because they are at-risk,
and interventions are targeted at increased police scrutiny and enforcement against those individuals. Whereas
the One Summer program takes an "Ethics of Care"
approach to at-risk youth, in which at-risk youth are given
access to social services and resources aimed at reducing their risks of becoming involved in violence.12 Like
their philosophies, their outcomes were also dramatically different, despite resting on similar data-driven assessments of being "at-risk."

The Heat List
The Strategic Subject List (SSL) algorithm was developed as an experiment by a researcher at the Illinois
Institute of Technology, and was utilized by CPD starting
in 2012 and continuing until today. In its early iterations
and implementations, it took data about individuals
from CPD arrest records, taking into account some 48
factors, including number of arrests, convictions, drug
arrests, gang affiliations, and being the victim of crimes
or violence [38]. The SSL then went further, taking into
account these factors for the individual's social network
as determined by who was arrested together with an
individual [39]. These factors were weighted and compiled into an overall SSL score from 1-500. The initial
implementation contained over 398 000 individuals
drawn from police arrest records, and identified 1400
as being at "high-risk" of being involved in violence.
While some 258 received the top score of 500 points,
only 48% of these had previously been arrested for a
gun crime, and many people on the list had never themselves been arrested, but rather were victims or were in
the social networks of victims or perpetrators [39]. Many
police officers reported that they were not fully
informed of how the list was compiled. They assumed,
or were led to believe, that everyone on the list was a
perpetrator of violence and was likely to commit more
violence, whereas the SSL scores combined those atrisk of being victims with those at-risk of being perpetrators in a single metric of "being involved in violence."
12

The slogan of the One Summer program is "Nothing Stops a Bullet Like a
Job" [37].

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

∕

JUNE 2019

IEEE Technology and Society Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - June 2019