CASE STUDY
BDR
ExclusionaryRule.pdf
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
561
The Exclusionary Rule in the Age of
Blue Data
Andrew Guthrie Ferguson*
In Herring v. United States, Chief Justice John Roberts
reframed the Supreme Court’s understanding of the exclusionary rule:
“As laid out in our cases, the exclusionary rule serves to deter
deliberate, reckless, or grossly negligent conduct, or in some
circumstances recurring or systemic negligence.” The open question
remains: How can defendants demonstrate sufficient recurring or
systemic negligence to warrant exclusion? The Supreme Court has
never answered the question, although the absence of systemic or
recurring problems has figured prominently in two recent exclusionary
rule decisions. Without the ability to document recurring failures or
patterns of police misconduct, courts can dismiss individual
constitutional violations merely as examples of “isolated negligence.”
But what if new data-driven surveillance technologies could
track police-citizen interactions and uncover recurring or systemic
problems? What if stops and arrests could be data mined to reveal
systemic racial bias? What if new surveillance technologies could
record police-citizen stops to monitor patterns of unconstitutional
practices? What if predictive analytics could identify at-risk officers in
order to predict future misconduct?
This Article looks to invert the big data surveillance gaze from
the citizen to the police. It asks whether the same big data policing
technologies built to track movements, actions, and patterns of
criminal activity could be redesigned to foster data-driven police
accountability. Tracking this “blue data” and studying the systemic
* Professor of Law, UDC David A. Clarke School of Law. Thank you for comments and
support from Professors Miriam Baer, Rachel Barkow, Paul Butler, Bennett Capers, Cynthia
Conti-Cook, Sharon Dolovich, Jeffrey Fagan, James Forman, Barry Freidman, Ben Grunwald,
Bernard Harcourt, David Harris, Eisha Jain, Orin Kerr, Adi Leibovitch, Kate Levine, Anna
Lvovsky, Shaun Ossei-Owusu, Tracey Meares, Erin Murphy, John Pfaff, Dan Richman, Alice
Ristrophe, Meghan Ryan, Sarah Seo, David Sklansky, Carol Steiker, Jenia Turner, and Crystal
Yang.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
562 VANDERBILT LAW REVIEW [Vol. 72:2:561
errors offers concrete answers to the open questions surrounding the
Supreme Court’s new exclusionary rule.
Unquestionably, the use of data mining, surveillance, and
predictive analytics to target police negligence will face resistance.
Police officers, administrators, and unions will likely protest the
invasion of personal and professional privacy it threatens. Yet, any
resistance is itself revealing and worth studying. This resistance offers
a provocative thought experiment: How could police objections to new
forms of surveillance inform community resistance to similar mass
surveillance technologies? This Article examines how police, courts,
and litigants will resist a push to police surveillance and what that
resistance means for current mass surveillance practices, law, and
policy.
INTRODUCTION ............................................................................... 563
I. THE LIMITS OF THE EXCLUSIONARY RULE AND POLICE
REFORM ............................................................................... 570 A. Roberts’ Rules of Exclusion ...................................... 572
1. The Question of Systemic Negligence:
Herring v. United States ............................... 573 2. A Lack of Recurring Violations:
Utah v. Strieff ............................................... 576 B. The Reality of Recurring Problems .......................... 580
1. The Fragmented Nature of Police Data ....... 582 2. Evidence of Recurring Problems .................. 585
C. Difficulties in Litigating Systemic or Recurring
Violations ................................................................. 591
II. BLUE DATA: INVERTING THE ARCHITECTURE OF BIG DATA
SURVEILLANCE .................................................................... 594 A. Data-Mining Technologies ....................................... 595
1. Mining Criminal Clues ................................. 596 2. Mining Policing Data .................................... 600 3. Mining Exclusion .......................................... 608
B. Monitoring Technologies .......................................... 611 1. Monitoring Crime ......................................... 612 2. Monitoring Police ......................................... 615 3. Monitoring Exclusion ................................... 619
C. Predictive Technologies ........................................... 621 1. Predicting Criminal Risk ............................. 621 2. Predicting Police Risk .................................. 626 3. Predicting Exclusion .................................... 632
D. Programmatic Benefit of Blue Data ........................ 634
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 563
III. THE REVEAL OF RESISTANCE ............................................... 635 A. Police Resistance ...................................................... 636 B. Legal Resistance ...................................................... 641
CONCLUSION ................................................................................... 645
INTRODUCTION
Digital technologies are transforming the daily practice of
policing.1 Big data surveillance technologies and predictive analytics
offer new methods for police to visualize otherwise hidden patterns of
criminal activity.2 Data mining assists law enforcement in gathering
intelligence.3 Predictive policing guides patrols.4 Pervasive
surveillance monitors the streets.5 Yet, in adopting this data-focused,
quantified approach to law enforcement, police have inadvertently
created equally revealing data-driven methods of police accountability.
The same surveillance technologies that can watch the citizenry can
also watch the police, and patterns of police misconduct can be
predicted and analyzed.
This technological change now holds significant constitutional
import because of how the Supreme Court has refashioned the
exclusionary rule, the suppression remedy for police wrongdoing.6 In
Herring v. United States, Chief Justice John Roberts reframed the
Supreme Court’s understanding of the exclusionary rule: “As laid out
in our cases, the exclusionary rule serves to deter deliberate, reckless,
or grossly negligent conduct, or in some circumstances recurring or
1. ANDREW GUTHRIE FERGUSON, THE RISE OF BIG DATA POLICING: SURVEILLANCE, RACE,
AND THE FUTURE OF LAW ENFORCEMENT 4 (2017) (detailing how big data surveillance
technologies will change the “who,” “where,” “when,” and “how” of the way in which law
enforcement addresses criminal risk).
2. See Andrew Guthrie Ferguson, Big Data and Predictive Reasonable Suspicion, 163 U.
PA. L. REV. 327, 329 (2015); Elizabeth E. Joh, Policing by Numbers: Big Data and the Fourth
Amendment, 89 WASH. L. REV. 35, 36 (2014).
3. See Tal Z. Zarsky, Governmental Data Mining and Its Alternatives, 116 PENN ST. L.
REV. 285, 287 (2011) (“[L]aw enforcement has shifted to ‘Intelligence Led Policing’ . . . . Rather
than merely reacting to events and investigating them, law enforcement is trying to preempt
crime. It does so by gathering intelligence, which includes personal information, closely
analyzing it, and allocating police resources accordingly . . . .”).
4. Andrew Guthrie Ferguson, Predictive Policing and Reasonable Suspicion, 62 EMORY
L.J. 259, 268–69 (2012) (describing the rise of place-based predictive policing).
5. Christopher Slobogin, Rehnquist and Panvasive Searches, 82 MISS. L.J. 307, 307–08
(2013) (describing panvasive searches arising from new technology).
6. See Illinois v. Gates, 462 U.S. 213, 254 (1983) (“The exclusionary rule is a remedy
adopted by this Court to effectuate the Fourth Amendment right of citizens ‘to be secure in their
persons, houses, papers, and effects, against unreasonable searches and seizures . . . .’ ”); Mapp v.
Ohio, 367 U.S. 643, 654–56 (1961) (“We hold that all evidence obtained by searches and seizures
in violation of the Constitution is, by that same authority, inadmissible in a state court.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
564 VANDERBILT LAW REVIEW [Vol. 72:2:561
systemic negligence.”7 Yet, despite the significance of “recurring or
systematic” problems in two recent Supreme Court cases,8 the Justices
did not explain how this could be proven. Equally limiting, the
ordinary practice of holding relatively brief suppression hearings
practically forecloses the ability to introduce evidence of systemic or
recurring policing problems.9 Without the ability to document
recurring patterns of police misconduct, courts can dismiss individual
constitutional violations merely as examples of “isolated negligence.”10
This Article looks to invert the big data surveillance gaze from
the citizen to the police. It asks whether the same law enforcement
technologies built to track movements, actions, and patterns of
criminal activity could also be repurposed to foster data-driven police
accountability. For example, what if stops and arrests could be data
mined to reveal systemic racial bias?11 What if predictive analytics
could identify at-risk officers or police units most likely to be involved
in recurring, future misconduct?12 What if new surveillance
technologies could record patterns of police-citizen stops to monitor
recurring unconstitutional practices?13 What if the entire architecture
of surveillance designed by law enforcement to surveil citizens could
be repurposed to identify recurring or systemic problems of police
violence, racial bias, and unconstitutional actions? Tracking this “blue
data”14 offers concrete answers to the open questions surrounding the
Supreme Court’s new application of the exclusionary rule.
Such futuristic surveillance technology already exists. Police
routinely search large datasets of biometric, geolocational, and
consumer information looking for patterns of recurring criminality.15
Communications, movements, or financial transactions can be
monitored to observe patterns of suspicious activities.16 In Los
Angeles, police track “chronic offenders” using social network analysis
7. 555 U.S. 135, 144 (2009) (emphasis added).
8. Id.; see also Utah v. Strieff, 136 S. Ct. 2056, 2063 (2016).
9. Andrew Guthrie Ferguson, Constitutional Culpability: Questioning the New
Exclusionary Rules, 66 FLA. L. REV. 623, 683 (2014) (discussing the resource constraints in
implementing a two-tiered suppression hearing after Herring).
10. The question of recurring violations also impacts private citizens’ ability to file civil
rights actions under 42 U.S.C. § 1983 (2012) and the federal government’s ability to investigate
patterns and practices of police abuse under 42 U.S.C. § 14141 (2012).
11. See infra Section II.A.2.
12. See infra Section II.B.2.
13. See infra Section II.C.2.
14. See FERGUSON, supra note 1, at 143–66 (detailing the concept of “blue data”).
15. See e.g., Wayne A. Logan, Policing Identity, 92 B.U. L. REV. 1561, 1575–78 (2012)
(describing biometric collection and searches).
16. Zarsky, supra note 3, at 287.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 565
technologies originally used to track international terrorists.17 These
growing social network systems link criminal associates in digital
webs of information that can be mined for investigatory clues.18
Patterns of criminal activities emerge from scraps of data, allowing
police to search through it to respond to community needs.
Video surveillance expands police capabilities to monitor
wrongdoing. In New York City, approximately nine thousand police
video cameras digitally record the streets in real time.19 Pattern
recognition software can automatically alert a central police command
center to a suspiciously placed bag or track all men wearing blue
sweatshirts.20 A single search query of the Domain Awareness
System—the New York City Police Department’s central command
center—can find all such blue sweatshirts in all locations recorded
over the last month.21 The city of Los Angeles has added facial
recognition software to a few police cameras, allowing those who pass
by to be matched with a database of active warrants.22 In both New
York City and Los Angeles, thousands of Automated License Plate
Readers (“ALPR”) record car licenses, marking location, time, and
direction of travel—all linked to details of the owner.23 Millions of
license plates are recorded every year and are included in local,
17. Sarah Brayne, Big Data Surveillance: The Case of Policing, 82 AM. SOC. REV. 977, 986–
87 (2017); Mark Harris, How Peter Thiel’s Secretive Data Company Pushed into Policing, WIRED
(Aug. 9, 2017), https://www.wired.com/story/how-peter-thiels-secretive-data-company-pushed-
into-policing [https://perma.cc/B686-QS2E].
18. See Harris, supra note 17.
19. Thomas H. Davenport, How Big Data is Helping NYPD Solve Crimes Faster, FORTUNE
(July 16, 2016), http://fortune.com/2016/07/17/big-data-nypd-situational-awareness
[https://perma.cc/TJA5-JMK5].
20. See TalkPolitix, New York City - Domain Awareness, YOUTUBE (June 7, 2013),
https://www.youtube.com/watch?v=ozUHOHAAhzg [https://perma.cc/N9ZP-SQB4]; see also
Michael L. Rich, Machine Learning, Automated Suspicion Algorithms, and the Fourth
Amendment, 164 U. PA. L. REV. 871, 896–901 (2016) (describing the capabilities of automated
search algorithms).
21. TalkPolitix, supra note 20.
22. CLARE GARVIE ET AL., THE PERPETUAL LINE-UP: UNREGULATED POLICE FACIAL
RECOGNITION IN AMERICA (Oct. 18, 2016), https://www.perpetuallineup.org/sites/default/
files/2016-12/The Perpetual Line-Up - Center on Privacy and Technology at Georgetown Law -
121616.pdf [https://perma.cc/5EZX-BPVW] [hereinafter PERPETUAL LINE-UP]; Clare Garvie &
Jonathan Frankle, Facial-Recognition Software Might Have a Racial Bias Problem, ATLANTIC
(Apr. 7, 2016), https://www.theatlantic.com/technology/archive/2016/04/the-underlying-bias-of-
facial-recognition-systems/476991 [https://perma.cc/QVR5-2B7A].
23. See Joh, supra note 2, at 48 (“The N.Y.P.D., for instance, has a database of 16 million
license plates captured from its license plate readers, along with the locations of where the plates
were photographed.”); Steven D. Seybold, Somebody’s Watching Me: Civilian Oversight of Data-
Collection Technologies, 93 TEX. L. REV. 1029, 1034 (2015) (“ALPR systems can photograph up to
1,800 license plates per minute, and approximately 10-12 million per day.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
566 VANDERBILT LAW REVIEW [Vol. 72:2:561
searchable databases.24 Citizens augment government surveillance
capabilities by carrying around trackable “smart” devices.25 GPS-
enabled cars record where we drive.26 Geotagging in photographs,
videos, and WiFi connections reveal where we have been (and what we
were doing there).27 Public social media accounts can be scraped and
studied to find patterns of movement and communications.28 Add in
the digital trails resulting from medical devices, financial applications,
and fitness trackers, and it is clear that a thick web of trackable self-
surveillance data exists.29
Predictive targeting allows police to narrow their surveillance
to specific individuals.30 Police in Chicago use an algorithm to identify
at-risk individuals in order to predict who might be the victim or
perpetrator of violence.31 In Manhattan, prosecutors and police
developed a data-driven “Moneyball” approach to incapacitate
24. Linda Merola & Cynthia Lum, Emerging Surveillance Technologies: Privacy and the
Case of License Plate Recognition (LPR) Technology, 96 JUDICATURE 119, 119–21 (2012).
25. Andrew Guthrie Ferguson, The Internet of Things and the Fourth Amendment of Effects,
104 CALIF. L. REV. 805, 818–23 (2016); Andrew Guthrie Ferguson, The “Smart” Fourth
Amendment, 102 CORNELL L. REV. 547, 548, 551 (2017); see also Scott R. Peppet, Regulating the
Internet of Things: First Steps Toward Managing Discrimination, Privacy, Security, and Consent,
93 TEX. L. REV. 85, 114–17 (2014) (discussing the varied capabilities of smartphone sensors).
26. Alex Hern, Florida Woman Arrested for Hit-and-Run After Her Car Calls Police,
GUARDIAN (Dec. 7, 2015), http://www.theguardian.com/technology/2015/dec/07/florida-woman-
arrested-hit-and-run-car-calls-police [https://perma.cc/MH8U-LF7Z]; Ned Potter, Privacy Battles:
OnStar Says GM Can Record Car’s Use, Even if You Cancel Service, ABC NEWS (Sept. 26, 2011),
https://abcnews.go.com/Technology/onstar-gm-privacy-terms-company-record-car-information/
story?id=14581571 [https://perma.cc/G4JQ-FZCC].
27. Rodolfo Ramirez et al., Location! Location! Location! Data Technologies and the Fourth
Amendment, CRIM. JUST., Winter 2016, at 19.
28. Aaron Cantú, #Followed: How Police Across the Country Are Employing Social Media
Surveillance, MUCKROCK (May 18, 2016), https://www.muckrock.com/news/archives/2016/may/
18/followed [https://perma.cc/KDX7-E3AR]; Matt Stroud, #Gunfire: Can Twitter Really Help Cops
Find Crime?, VERGE (Nov. 15, 2013), https://www.theverge.com/2013/11/15/5108058/gunfire-can-
twitter-really-help-cops-find-crime [https://perma.cc/RDV6-W4AR].
29. See Tony Danova, Morgan Stanley: 75 Billion Devices Will Be Connected to the Internet
of Things by 2020, BUS. INSIDER (Oct. 2, 2013), https://www.businessinsider.com/75-billion-
devices-will-be-connected-to-the-internet-by-2020-2013-10 [https://perma.cc/7RJF-SF8K]
(describing the sheer volume of devices that are, or in the future will be, connected to the
internet).
30. Andrew Guthrie Ferguson, Predictive Prosecution, 51 WAKE FOREST L. REV. 705, 724
(2016).
31. Monica Davey, Chicago Police Try to Predict Who May Shoot or Be Shot, N.Y. TIMES
(May 23, 2016), https://www.nytimes.com/2016/05/24/us/armed-with-data-chicago-police-try-to-
predict-who-may-shoot-or-be-shot.html [https://perma.cc/TR5H-9YN7]; Josh Kaplan, Predictive
Policing and the Long Road to Transparency, SOUTHSIDE WKLY. (July 12, 2017),
https://southsideweekly.com/predictive-policing-long-road-transparency [https://perma.cc/53L8-
27H2]; Nissa Rhee, Can Police Big Data Stop Chicago’s Spike in Crime?, CHRISTIAN SCI.
MONITOR (June 2, 2016), https://www.csmonitor.com/USA/Justice/2016/0602/Can-police-big-data-
stop-Chicago-s-spike-in-crime [https://perma.cc/RCV7-3F6F].
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 567
“primary targets” in particularized blocks or housing units.32 Police in
Los Angeles, Seattle, Miami, Atlanta, and dozens of other big cities
utilize predictive policing software to target forecasted high-crime
areas.33 Police patrol these predicted areas for additional physical
surveillance.34 These new data-driven technologies offer a blueprint
for a new type of policing. While not yet universally adopted, the
designs exist and have been growing in many cities.
This surveillance architecture unquestionably poses significant
liberty and privacy concerns. As I and others have written, these new
technologies undermine and distort traditional First and Fourth
Amendment freedoms in ways we are only just beginning to imagine.35
But these same technologies also offer a potential solution to the
current exclusionary rule puzzle. New data surveillance systems built
32. Chip Brown, The Data D.A., N.Y. TIMES, Dec. 6, 2014, at 22, 24–25; To Stem Gun Crime,
‘Moneyball,’ ST. LOUIS POST-DISPATCH, June 28, 2015, at A20; Heather Mac Donald, Prosecution
Gets Smart, CITY J. (Aug. 14, 2014), https://www.city-journal.org/html/prosecution-gets-smart-
13663.html [https://perma.cc/LL55-W276].
33. Ellen Huet, Server And Protect: Predictive Policing Firm PredPol Promises to Map
Crime Before It Happens, FORBES (Mar. 2, 2015), https://www.forbes.com/sites/ellenhuet/2015/02/
11/predpol-predictive-policing [https://perma.cc/TU4D-LT95]; Mara Hvistendahl, Can “Predictive
Policing” Prevent Crime Before It Happens?, SCIENCE (Sept. 28, 2016),
http://www.sciencemag.org/news/2016/09/can-predictive-policing-prevent-crime-it-happens
[https://perma.cc/66TH-TU4J]; David Robinson & Logan Koepke, Stuck in a Pattern: Early
Evidence on “Predictive Policing” and Civil Rights, UPTURN (Aug. 2016), https://www.upturn.org/
reports/2016/stuck-in-a-pattern/ [https://perma.cc/T456-5BLL].
34. Ferguson, supra note 4, at 267–69.
35. See sources cited supra notes 1–2, 4; see also, e.g., CHRISTOPHER SLOBOGIN, PRIVACY AT
RISK: THE NEW GOVERNMENT SURVEILLANCE AND THE FOURTH AMENDMENT 205 (2007)
(“[S]urveillance that is not regulated is unreasonable under the Constitution.”); Marc Jonathan
Blitz et al., Regulating Drones Under the First and Fourth Amendment, 57 WM. & MARY L. REV.
49, 60 (2015) (“It is clear . . . now is the time to understand the Fourth Amendment restrictions
of government flight, the First Amendment protections for private flight, and the
interdependency of between the two.”); Marc Jonathan Blitz, Video Surveillance and the
Constitution of Public Space: Fitting the Fourth Amendment to a World that Tracks Image and
Identity, 82 TEX. L. REV. 1349, 1383 (2004) (“It is not only the expansion of video surveillance
itself that poses a challenge to the viability of the Katz test but also the dramatic changes
occurring in technologies that supplement and enhance such surveillance.”); David Gray &
Danielle Citron, The Right to Quantitative Privacy, 98 MINN. L. REV. 62, 66 (2013) (highlighting
certain surveillance technologies in use across the country); Stephen Henderson, Fourth
Amendment Time Machines (and What They Might Say About Police Body Cameras), 18 U. PA. J.
CONST. L. 933, 936 (2016) (asking, given the advancements in surveillance technology, how our
constitutional jurisprudence should respond to bulk capture of information via technology);
Elizabeth E. Joh, Privacy Protests: Surveillance Evasion and Fourth Amendment Suspicion, 55
ARIZ. L. REV. 997, 1002 (2013) (discussing “privacy protects,” defined as “actions individuals may
take to block or thwart surveillance . . . for reasons unrelated to criminal wrongdoing”); Neil
Richards, The Dangers of Surveillance, 126 HARV. L. REV. 1934, 1953 (2013) (“Even in democratic
societies, the blackmail threat of surveillance is a real one.”); Steven D. Seybold, Note,
Somebody’s Watching Me: Civilian Oversight of Data-Collection Technologies, 93 TEX. L. REV.
1029, 1034 (2015) (“Combining surveillance technologies not only allows for more information to
be collected but also allows for powerful inferences to be drawn from that information; inferences
that may not have been readily apparent from each individual piece of information by itself.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
568 VANDERBILT LAW REVIEW [Vol. 72:2:561
by the police can also be used to monitor systemic and recurring police
practices. In every Big Brother–esque example discussed above,
technology also captures police-citizen interactions in new and
revealing ways that can help expose existing police abuses. The great
irony of the modern surveillance state is that law enforcement
accidently designed a system that can monitor the police better than
its citizens.
The need to reimagine police accountability is a contested
national issue. The open secret of minority distrust and fear of police
has loudly revealed itself in a series of self-reinforcing, cascading
scandals and events.36 The protests arising from the deaths of
unarmed African American men at the hands of police sparked an
ongoing national debate over inadequate police accountability.37 Black
lives, made visible by a pattern of Black deaths, turned police reform
into a national movement.38 This movement exposed a lack of police
accountability, made worse by the parallel judicial weakening of
deterrence-based remedies like the exclusionary rule.39 More
immediately, the need to reimagine accountability has grown stronger
still, as the Department of Justice (“DOJ”) Civil Rights Division has
backed away from prioritizing police accountability investigations
under its new leadership.40
36. Monica Davey & Julie Bosman, Protests Flare After Ferguson Police Officer Is Not
Indicted, N.Y. TIMES (Nov. 24, 2014), https://www.nytimes.com/2014/11/25/us/ferguson-darren-
wilson-shooting-michael-brown-grand-jury.html [https://perma.cc/YG3N-CT9K]; Dana Ford et
al., Protests Erupt in Wake of Chokehold Death Decision, CNN (Dec. 8, 2014, 8:14 PM),
http://www.cnn.com/2014/12/04/justice/new-york-grand-jury-chokehold [https://perma.cc/H2WV-
AGJ5]; see also MICHELLE ALEXANDER, THE NEW JIM CROW: MASS INCARCERATION IN THE AGE OF
COLORBLINDNESS (2010); PAUL BUTLER, LET’S GET FREE: A HIP-HOP THEORY OF JUSTICE (2009);
DAVID COLE, NO EQUAL JUSTICE: RACE AND CLASS IN THE AMERICAN CRIMINAL JUSTICE SYSTEM
(1999).
37. Alan Blinder, Walter Scott Shooting Seen as Opening for Civil Suits Against North
Charleston’s Police Dept., N.Y. TIMES (Apr. 13, 2015), https://www.nytimes.com/2015/04/14/us/
walter-scott-shooting-turns-michael-slager-into-litigant-as-north-charleston-braces-for-suits.html
[https://perma.cc/4BEZ-ZBPQ]; Shaila Dewan & Richard A. Oppel, Jr., In Tamir Rice Case, Many
Errors by Cleveland Police, Then a Fatal One, N.Y. TIMES (Jan. 22, 2015),
https://www.nytimes.com/2015/01/23/us/in-tamir-rice-shooting-in-cleveland-many-errors-by-
police-then-a-fatal-one.html [https://perma.cc/439U-7HPE]; Roger Parloff, Two Deaths: The
Crucial Difference Between Eric Garner’s Case and Michael Brown’s, FORTUNE (Dec. 5, 2014),
http://fortune.com/2014/12/05/two-deaths-the-crucial-difference-between-eric-garners-case-and-
michael-browns/ [https://perma.cc/C5YR-M69H].
38. Ferguson Unrest: From Shooting to Nationwide Protests, BBC NEWS (Aug. 10, 2015),
https://www.bbc.com/news/world-us-canada-30193354 [https://perma.cc/JX87-YGUK].
39. See infra Section I.A.
40. Eric Lichtblau, Sessions Indicates Justice Department Will Stop Monitoring Troubled
Police Agencies, N.Y. TIMES (Feb. 28, 2017), https://www.nytimes.com/2017/02/28/us/politics/jeff-
sessions-crime.html [https://perma.cc/6X89-82EZ]; Sheryl Gay Stolberg & Eric Lichtblau,
Sweeping Federal Review Could Affect Consent Decrees Nationwide, N.Y. TIMES (Apr. 3, 2017),
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 569
The time has come to examine new data-driven forms of
accountability, as law enforcement is beginning to embrace a mass
surveillance mindset.41 Technology has made it temptingly easy to
monitor and target citizens. But the current technological capabilities
have also outpaced citizen awareness, providing a moment to stop and
reflect on the potential impacts before ubiquitous adoption. Serious
examination of the dual nature of surveillance—on the public and
police—may help frame a more cautious approach to big data policing
in the future.
Part I of this Article examines the question left open by the
Supreme Court’s recent exclusionary rule cases, Herring v. United
States42 and Utah v. Strieff.43 Namely, how can defendants
demonstrate recurring or systemic police negligence? The short
answer is that under traditional Fourth Amendment law and practice,
litigants cannot (and in practice do not) regularly meet this burden.44
Building a record of systemic violations is time-consuming, expensive,
and taxes the abilities of both lawyers and courts and thus has not
been a focus of suppression litigation. Yet systemic and recurring
problems exist in many police forces.45 As seen in media reports,
scholarly articles, lawsuits, and federal investigations, the problem of
police violence, racial bias, and constitutional violations must be
remedied.46
Part II of the Article examines how big data surveillance tools
can be redesigned to develop a record of police accountability useful for
this new exclusionary rule regime. This is the promise of “blue data.”47
The rise of new technologies to mine data and analyze criminal
activity can also identify patterns of constitutional violations or police
misconduct. Additionally, new video and audio surveillance
technologies can not only monitor the streets but also monitor police
activities. Finally, new predictive analytics can flag at-risk criminals
and at-risk police officers with equal ease. By quantifying police
activities, litigants can begin to visualize patterns of systemic and
https://www.nytimes.com/2017/04/03/us/justice-department-jeff-sessions-baltimore-police.html
[https://perma.cc/3QXD-U5UJ].
41. See Stephen Rushin, The Judicial Response to Mass Surveillance, 2011 U. ILL. J.L.
TECH. & POL’Y 281, 285–86 (“[M]any departments across the country are using [certain
technologies] not just for observational comparison, but also for indiscriminate data collection.”).
42. 555 U.S. 135 (2009).
43. 136 S. Ct. 2056 (2016).
44. See infra Section I.C.
45. See infra Section I.B.2.
46. See infra Section I.B.
47. See Ferguson, supra note 1, at 143–66 (detailing the concept of “blue data”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
570 VANDERBILT LAW REVIEW [Vol. 72:2:561
recurring issues and introduce them in Fourth Amendment
suppression hearings.
Part III examines the revealing nature of police resistance to
blue data collection. Obviously, new surveillance technologies will be
resisted by police officers and administrators concerned about how
they might impact professional autonomy and criminal investigations.
These arguments offer a provocative thought experiment: How could
police objections to new forms of surveillance inform citizen and
community resistance to similar surveillance technology? It might be
the case that police resistance to self-surveillance informs citizen
resistance to mass surveillance. This Part examines how police,
courts, and litigants will resist a push to police surveillance and what
that resistance says about current practice, law, and policy priorities.
In redirecting the target of surveillance from the citizen to the
police, this Article explores how to meet the Supreme Court’s new
burden for exclusion. These “blue data” systems—already in
development—offer a solution to the long-standing problem of police
accountability. They offer new ways to visualize the recurring and
systemic gaps in the existing policing system and thus to close the
widening gap between the Supreme Court’s standards for exclusion
and the ability to offer proof to meet those standards.
I. THE LIMITS OF THE EXCLUSIONARY RULE AND POLICE REFORM
Since its creation, the exclusionary rule has been criticized by
judges and scholars.48 The remedy of suppressing evidence recovered
as a result of a constitutional violation has divided the Supreme Court
for decades.49 In recent years, a conservative majority has limited the
48. See Herring v. United States, 555 U.S. 135, 151 (2009) (Ginsburg, J., dissenting)
(recognizing “a more majestic conception” of the Fourth Amendment (quoting Arizona v. Evans,
514 U.S. 1, 18 (1995)); People v. Defore, 150 N.E. 585, 587–89 (N.Y. 1926) (highlighting the
scrutiny surrounding the doctrine by some courts and pondering why the criminal should “go free
because the constable has blundered”); Henry J. Friendly, The Bill of Rights as a Code of
Criminal Procedure, 53 CALIF. L. REV. 929, 951 (1965) (“Another imperative which in my view
has been too quickly assumed is that the Constitution demands that convictions be automatically
set aside in every instance in which material evidence obtained in violation of some ‘specific’ of
the Bill of Rights was received.”); see also Craig M. Bradley, The “Good Faith Exception” Cases:
Reasonable Exercise in Futility, 60 IND. L.J. 287 (1985); John M. Burkoff, Bad Faith Searches, 57
N.Y.U. L. REV. 70 (1982); Tracey Maclin, When the Cure for the Fourth Amendment is Worse than
the Disease, 68 S. CAL. L. REV. 1, 49–50 (1994); Carol S. Steiker, Second Thoughts About First
Principles, 107 HARV. L. REV. 820, 847–52 (1994); Silas Wasserstrom & William J. Mertens, The
Exclusionary Rule on the Scaffold: But Was It a Fair Trial?, 22 AM. CRIM. L. REV. 85 (1984).
49. Fourth Amendment—Exclusionary Rule—Deterrence Costs and Benefits—Utah v.
Strieff—Leading Case, 130 HARV. L. REV. 337 (2016) [hereinafter Utah v. Strieff—Leading Case]
(“Over the next forty years, the Court stripped away the exclusionary rule’s justification either as
an individual right or as a means of ensuring judicial integrity.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 571
availability of the remedy, first through the creation of a patchwork of
exceptions,50 and later by reconceptualizing the purpose of the
exclusionary rule as only to deter police misconduct.51 In a series of
cases—Hudson v. Michigan,52 Herring v. United States,53 Davis v.
United States,54 and Utah v. Strieff55—the Court created a new
framework with two primary considerations: first, whether the
officer’s actions were deliberate, reckless, grossly negligent, or the
result of systemic or recurring negligence,56 and second, whether the
actions were attenuated from the original constitutional harm.57
Scholars have ably critiqued the Court’s reasoning, challenging
the logic, interpretation, and even constitutional theory underpinning
these decisions.58 In a prior article, I addressed the complexities of
taking seriously the terms “deliberate,” “reckless,” and “gross
50. See, e.g., Murray v. United States, 487 U.S. 533, 542 (1988) (“independent source”
doctrine); United States v. Leon, 468 U.S. 897, 920–21 (1984) (good faith exception); Nix v.
Williams, 467 U.S. 431, 444 (1984) (inevitable discovery doctrine); Wong Sun v. United States,
371 U.S. 471, 491 (1963) (attenuation or causation exception); see also Tonja Jacobi, The Law
and Economics of the Exclusionary Rule, 87 NOTRE DAME L. REV. 585, 656 (2011) (“For advocates
of the exclusionary rule, the great tragedy of recent jurisprudence has been the erosion of the
strength of the rule: courts have developed numerous exceptions, a process which has arguably
steadily eroded Fourth Amendment protections over time.”).
51. Utah v. Strieff, 136 S. Ct. 2056, 2059 (2016) (“[E]ven when there is a Fourth
Amendment violation, this exclusionary rule does not apply when the costs of exclusion outweigh
its deterrent benefits.”); id. at 2071 (Kagan, J., dissenting) (“The exclusionary rule serves a
crucial function—to deter unconstitutional police conduct. By barring the use of illegally
obtained evidence, courts reduce the temptation for police officers to skirt the Fourth
Amendment’s requirements.”); Utah v. Strieff—Leading Case, supra note 49, at 343 (“Over the
next forty years, the Court stripped away the exclusionary rule’s justification either as an
individual right or as a means of ensuring judicial integrity.”).
52. 547 U.S. 586 (2006).
53. 555 U.S. 135.
54. 131 S. Ct. 2419 (2011).
55. 136 S. Ct. 2056.
56. See Herring, 555 U.S. at 137 (noting that arrests based on incorrect beliefs or negligence
can still constitute Fourth Amendment violations).
57. Utah v. Strieff—Leading Case, supra note 49, at 338.
58. See, e.g., Albert W. Alschuler, Herring v. United States: A Minnow or a Shark?, 7 OHIO
ST. J. CRIM. L. 463, 501–07, 510–11 (2009) (describing history of the exclusionary rule beginning
before the Revolutionary War); Jeffrey Fagan, Terry’s Original Sin, 2016 U. CHI. LEGAL F. 43, 66
(“[T]he attenuation doctrine applied by the Strieff Court essentially scrubs out reasonableness
from the Terry formula.”); Wayne R. LaFave, The Smell of Herring: A Critique of the Supreme
Court’s Latest Assault on the Exclusionary Rule, 99 J. CRIM. L. & CRIMINOLOGY 757, 758 (2009)
(critiquing the Court’s decision in Herring for complicating Fourth Amendment analyses); David
Alan Sklansky, Is the Exclusionary Rule Obsolete?, 5 OHIO ST. J. CRIM. L. 567, 578 (2008)
(discussing ongoing informal reforms in police departments); Steiker, supra note 48, at 848
(arguing for a modernized interpretation of what constitutes reasonableness under the Fourth
Amendment); James J. Tomkovicz, Hudson v. Michigan and the Future of Fourth Amendment
Exclusion, 93 IOWA L. REV. 1819, 1832–33, 1848–49, 1880–81 (2008) (describing the variety of
interpretations lower courts can make in response to the Court’s ruling in Hudson); Craig M.
Bradley, Red Herring or the Death of the Exclusionary Rule?, TRIAL, Apr. 2009, at 53 (noting the
difficulty in applying Herring to a broad collection of cases).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
572 VANDERBILT LAW REVIEW [Vol. 72:2:561
negligence” when it comes to litigating suppression issues.59 But the
latter part of that test was left unexamined—the problem of proving
systemic and recurring negligence.60
This Part proceeds in three steps. First, it examines two recent
cases involving unlitigated, but arguably controlling, examples of
systemic or recurring police errors. In Herring v. United States, the
Court relied heavily on the fact that no evidence of systemic database
errors had been included in the trial record.61 In Utah v. Strieff, the
Court dismissed Strieff’s claim, in part, because no recurring pattern
of misconduct had been demonstrated.62 In these cases, the Supreme
Court provided a new framework for exclusion, but little guidance on
how to prove that patterns of misconduct exist. Second, this Part
situates the Supreme Court’s focus on systemic or recurring problems
within a larger national conversation about police reform in America.
The problem of police reform and deterring police misconduct,
including unconstitutional stops, racial bias, and excessive force, has
been demonstrated though a growing collection of investigations, court
decisions, and media reports.63 Finally, this Part examines why proof
of systemic and recurring violations rarely makes it into ordinary
Fourth Amendment suppression hearings. Both law and practice
conspire to limit the trial record, minimizing the opportunity to
develop proof of systemic problems. This Part lays out a framework for
why a new, data-driven, surveillance-oriented approach may better
respond to the challenge of the modern exclusionary rule doctrine.
A. Roberts’ Rules of Exclusion
This Section briefly examines two recent Supreme Court cases
which suggest that systemic negligence or recurring violations could
59. Ferguson, supra note 9, at 683.
60. Jennifer E. Laurin, Trawling for Herring: Lessons in Doctrinal Borrowing and
Convergence, 111 COLUM. L. REV. 670, 684 (2011):
A second innovation of Herring, and a corollary to its culpability focus, was its
adoption of an exclusionary rule test expressly aimed at institutional, in addition to
individual, misconduct. The Court allowed that even in the face of apparently
blameless action by a law enforcement officer, evidence of “systemic error” or, phrased
differently, “systemic negligence,” would justify application of the exclusionary rule.
Beyond incantation of these apparent terms of art, virtually no explanation is
provided as to their meaning. Nor, despite the Court’s allusion to precedent, can the
meaning of these phrases be discerned from prior exclusionary rule decisions, since no
case prior to Herring had held that systemic Fourth Amendment misconduct could
provide the basis for a motion to suppress.
61. See infra Section I.A.1.
62. See infra Section I.A.2.
63. See infra Section I.A.2.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 573
be a trigger for exclusion. While neither case directly involved
systemic or recurring problems, the Court acknowledged that proof of
such a pattern could alter the analysis and thus the outcome of the
suppression argument.
1. The Question of Systemic Negligence: Herring v. United States
Herring v. United States involved a police-database error which
resulted in the unconstitutional stop and search of Bennie Dean
Herring.64 Mr. Herring, it appears, had gotten on the wrong side of
investigator Mark Anderson by informing the local district attorney
that Anderson had been involved in a recent murder.65 When Herring
was visiting the Coffee County Sheriff’s Department’s impound lot,
Anderson decided to determine whether Herring had any open arrest
warrants.66 First, Anderson asked the Coffee County warrant clerk to
see if any open warrants existed.67 When none were found, he asked
the clerk to check with the neighboring Dale County Sheriff’s
Department.68 The Dale County computer database erroneously
reported that Herring had an open arrest warrant.69 Apparently, the
warrant had been recalled, but the computer did not record this fact.70
Based on that mistaken information, Anderson stopped and searched
Herring.71 Methamphetamine and a handgun were recovered and
Herring was arrested.72
In the subsequent criminal prosecution, Herring moved to
suppress the evidence, arguing that his Fourth Amendment rights had
been violated since he had been arrested without a valid arrest
warrant. Factually, at the time of his stop, there had been no valid
warrant, and so, as a legal matter, Herring had been arrested without
justification. On appeal to the Supreme Court, the parties agreed that
a Fourth Amendment violation had occurred, but focused on whether
investigator Anderson’s good faith reliance on the Dale County
database required suppression of the evidence.73
64. Herring v. United States, 555 U.S. 135, 137 (2009).
65. Id. at 149 (Ginsburg, J., dissenting).
66. Id.
67. Id.
68. Id.
69. Id.
70. Id.
71. Id. at 150.
72. Id. at 137 (majority opinion).
73. Id. at 139 (“[W]e accept the parties’ assumption that there was a Fourth Amendment
violation.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
574 VANDERBILT LAW REVIEW [Vol. 72:2:561
In a sweeping opinion, Chief Justice John Roberts used
Herring to reinterpret the rationale for the exclusionary rule. Writing
for the majority, Chief Justice Roberts canvassed the history of the
exclusionary rule, examining its utility in exposing flagrant or
purposeful police violations. He explained that exclusion should not be
considered an automatic remedy for a constitutional violation:74 “We
have repeatedly rejected the argument that exclusion is a necessary
consequence of a Fourth Amendment violation. Instead we have
focused on the efficacy of the rule in deterring Fourth Amendment
violations in the future.”75 The key to determining whether the
exclusionary rule applies, according to Chief Justice Roberts, is to ask
whether the exclusion will deter future misconduct.76 To further that
deterrent focus, the Court established a new test:
To trigger the exclusionary rule, police conduct must be sufficiently deliberate that
exclusion can meaningfully deter it, and sufficiently culpable that such deterrence is
worth the price paid by the justice system. As laid out in our cases, the exclusionary rule
serves to deter deliberate, reckless, or grossly negligent conduct, or in some
circumstances recurring or systemic negligence.77
Because the database error in the case appeared to be an isolated
mistake, the Court found no need to suppress the evidence recovered
on Herring.78
Critical to the Court’s decision in Herring was the lack of
demonstrated systemic error in the database.79 In fact, the Court
made this point explicit, stating that it might be reckless to rely on an
unreliable warrant system if systematic errors were shown.80
This concern with systemic error animated Justice Ginsburg’s
dissent. As she explained, the fact that the erroneous warrant existed
for five months without correction, and that there was “no routine
practice of checking the database for accuracy,” undermined the
74. Id. at 137 (pointing out that “suppression is not an automatic consequence of a Fourth
Amendment violation”).
75. Id. at 141 (citations omitted).
76. Id. at 137.
77. Id. at 144.
78. Id. at 137 (holding that “the error was the result of isolated negligence attenuated from
the arrest”).
79. Id. at 147:
But there is no evidence that errors in Dale County’s system are routine or
widespread. Officer Anderson testified that he had never had reason to question
information about a Dale County warrant, . . . and both Sandy Pope and Sharon
Morgan testified that they could remember no similar miscommunication ever
happening on their watch . . . .
80. Id. at 146 (“If the police have been shown to be reckless in maintaining a warrant
system, or to have knowingly made false entries to lay the groundwork for future false arrests,
exclusion would certainly be justified under our cases should such misconduct cause a Fourth
Amendment violation.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 575
isolated nature of the error.81 More importantly, Justice Ginsburg
argued that proven errors in arrest databases across the nation
required the Court to address systemic threats to the Fourth
Amendment:
Inaccuracies in expansive, interconnected collections of electronic information raise
grave concerns for individual liberty. The offense to the dignity of the citizen who is
arrested, handcuffed, and searched on a public street simply because some bureaucrat
has failed to maintain an accurate computer data base is evocative of the use of general
warrants that so outraged the authors of our Bill of Rights.82
Broadly focusing on the grave consequences of law enforcement’s
recordkeeping errors,83 Justice Ginsburg concluded that these
databases pose a considerable risk because they are frequently out of
date or filled with mistakes.84
Finally, Justice Ginsburg pointed out the practical problem
with defendants—mostly indigent—litigating these issues. As she
wrote, “even when deliberate or reckless conduct is afoot, the Court’s
assurance will often be an empty promise: How is an impecunious
defendant to make the required showing?”85 Justice Ginsburg noted
that discovery would place a substantial administrative burden on
both the court and law enforcement, and might even include an audit
of police databases.86 Discovery or any required police-database audit
would have to be carried out in ordinary, trial court level suppression
hearings because only in such hearings could defendants evaluate the
extent of systemic or recurring problems under a negligence theory.
In addition, there is a more fundamental question at the heart
of this new requirement. Chief Justice Roberts did not explain what
the Court meant by the term “negligence” in the context of the
exclusionary rule, and there has been little judicial commentary on
the subject.87 Oddly, for such a seemingly sweeping doctrinal change,
81. Id. at 154 (Ginsburg, J., dissenting).
82. Id. at 155–56 (internal quotation marks omitted) (quoting Arizona v. Evans, 514 U.S. 1,
23 (1995) (Stevens, J., dissenting)).
83. See id. at 150 (referring to the gravity of recordkeeping errors in law enforcement).
84. Id. at 155 (“Herring’s amici warn that law enforcement databases are insufficiently
monitored and often out of date. Government reports describe, for example, flaws in [National
Criminal Information Center (“NCIC”)] databases, terrorist watchlist databases, and databases
associated with the Federal Government’s employment eligibility verification system.” (citation
and footnotes omitted)). See generally Wayne A. Logan & Andrew Guthrie Ferguson, Policing
Criminal Justice Data, 101 MINN. L. REV. 541, 542–43 (2016) (detailing how there are
“significant quality problems with criminal justice databases” and a “blasé acceptance of data
error and its negative consequences for individuals”).
85. Herring, 555 U.S. at 157 (Ginsburg, J., dissenting).
86. Id.
87. Kit Kinports, Culpability, Deterrence, and the Exclusionary Rule, 21 WM. & MARY BILL
RTS. J. 821, 836 (2013) (detailing the flawed conception of negligence revealed in Supreme Court
exclusionary rule cases). A few clues can be divined from Herring itself. Examining the language
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
576 VANDERBILT LAW REVIEW [Vol. 72:2:561
the Supreme Court did not provide much detail directing lawyers how
to actually implement the standard.88
2. A Lack of Recurring Violations: Utah v. Strieff
Recurring constitutional violations also played a role in Utah v.
Strieff.89 At issue was the “normal” or apparently “common practice” of
Salt Lake City police officers to detain pedestrians without reasonable
suspicion in order to run warrant checks.90 This unconstitutional (if
routine) practice played out with the stop and search of Edward
Joseph Strieff.
chosen shows that Chief Justice Roberts used both the term “grossly negligent” and “negligence”
in the same sentence, the former being understood to be a higher standard than mere negligence.
Herring, 555 U.S. at 144. In addition, in characterizing the error in Herring, the Court used the
term “isolated negligence” as a different and seemingly lesser standard than the other terms of
art (i.e., “deliberate,” “reckless,” and “grossly negligent”). Id. at 137. Using the traditional
methods of interpretation, courts could borrow from civil negligence precedents or criminal
negligence cases, with each standard possibly resulting in a different outcome. See Ferguson,
supra note 9, at 653 (outlining different definitions of criminal negligence from California courts
and the Model Penal Code); Ronen Perry, Re-Torts, 59 ALA. L. REV. 987, 989 (2008):
Section 3 of the Third Restatement defines “negligence” in cost-benefit terms. The
initial clause provides that a person acts negligently if he or she does not exercise
reasonable care under all the circumstances. The next clause stipulates that
“[p]rimary factors to consider in ascertaining whether the person’s conduct lacks
reasonable care are the foreseeable likelihood that the person’s conduct will result in
harm, the foreseeable severity of any harm that may ensue, and the burden of
precautions to eliminate or reduce the risk of harm.
(alteration in original) (footnotes omitted).
88. For example, with any negligence standard, courts must define a duty of care, and in
other contexts courts have been quite protective of police in limiting these duties. Police do not
ordinarily have an affirmative duty to protect individual citizens on the streets. See Town of
Castle Rock v. Gonzalez, 545 U.S. 748, 768 (2005) (finding no police liability for failing to enforce
a restraining order that resulted in violence); DeShaney v. Winnebago Cty. Dep’t of Soc. Servs.,
489 U.S. 189, 196–202 (1989) (requiring a special relationship before finding a constitutional
right to government protection in the social services protection context). Yet, the Supreme Court
also seems to acknowledge that police do have a duty to not violate constitutional rights. This is,
after all, the theory behind Section 1983 litigation where an individual must show a deprivation
of constitutional rights. See 42 U.S.C § 1983 (2012). Further, in Herring, the Court implied that
ignoring systemic negligence would be reckless and open police up to the exclusionary rule
remedy. Is there a duty of care not to stop or frisk a suspect without adequate legal justification?
At least at a systemic level, would such violations be sufficient to find negligence? This
conclusion would create a measure of symmetry with the pattern and practice violations under
42 U.S.C. § 14141 (2012) and municipal liability under 42 U.S.C § 1983.
89. 136 S. Ct. 2056 (2016).
90. Id. at 2068 (Sotomayor, J., dissenting) (“The Utah Supreme Court described as
‘ “routine procedure” or “common practice” ’ the decision of Salt Lake City police officers to run
warrant checks on pedestrians they detained without reasonable suspicion.” (quoting State v.
Topanotes, 2003 UT 30, ¶ 2, 76 P.3d 1159, 1160)); see id. at 2073 (Kagan, J., dissenting) (“As
Fackrell testified, checking for outstanding warrants during a stop is the ‘normal’ practice of
South Salt Lake City police.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 577
In December 2006, narcotics detective Douglas Fackrell
received an anonymous tip that a particular house was the source of
drug dealing.91 Fackrell monitored the house over the course of a
week.92 Fackrell’s observations confirmed his suspicion of illegal
behavior, and he subsequently stopped Edward Joseph Strieff as he
left the house.93 At the time of the stop, Fackrell did not know Strieff,
did not know how long Strieff had been at the targeted house, and had
no suspicion of Strieff personally.94 After seizing Strieff pursuant to
Salt Lake City’s common practice of detaining people in order to
search for open warrants, Fackrell had a police dispatcher run Strieff’s
name through a database and found he had an existing arrest warrant
for a traffic violation.95 Fackrell arrested Strieff and in a search
incident to that arrest recovered methamphetamine and drug
paraphernalia.96
Strieff moved to suppress the drug evidence, arguing that
Fackrell seized him without reasonable suspicion.97 On appeal, the
State of Utah conceded that Fackrell did not have adequate
reasonable suspicion to stop Strieff.98 The United States Supreme
Court assumed without deciding that Strieff was stopped in violation
of the Fourth Amendment and instead focused on the appropriateness
of the suppression remedy.99 In an opinion written by Justice Clarence
Thomas, the Court held that the existence of a valid arrest warrant
served to attenuate the constitutional violation from the subsequent
recovery of the drugs.100 In other words, the preexisting lawful
warrant severed the connection between the constitutional violation
91. Id. at 2059 (majority opinion).
92. Id.
93. Id. at 2060.
94. See id. (explaining how Fackrell needed to identify himself to Streiff when Fackrell
made the stop).
95. Id.
96. Id.
97. Id.
98. Id. at 2060.
99. Id. at 2062 (assuming without deciding that “Fackrell lacked reasonable suspicion to
initially stop Strieff”); see id. at 2072 (Kagan, J., dissenting):
At the suppression hearing, Fackrell acknowledged that the stop was designed for
investigatory purposes—i.e., to “find out what was going on [in] the house” he had
been watching, and to figure out “what [Strieff] was doing there.” . . . And Fackrell
frankly admitted that he had no basis for his action except that Strieff “was coming
out of the house.”
100. Id. at 2064 (majority opinion) (“We hold that the evidence Officer Fackrell seized as part
of his search incident to arrest is admissible because his discovery of the arrest warrant
attenuated the connection between the unlawful stop and the evidence seized from Strieff
incident to arrest.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
578 VANDERBILT LAW REVIEW [Vol. 72:2:561
and the remedy of excluding evidence.101 Applying an attenuation
analysis first developed in Brown v. Illinois, the Court held that the
evidence should not be suppressed.102
Underlying the Court’s ultimate attenuation theory were two
arguments. First, that detective Fackrell’s constitutional error was at
worst only negligent.103 Second, that the error was not part of any
recurring pattern of police misconduct.104 This framing tracked the
logic in Herring—without purposeful error or recurring negligence,
suppression would not deter future misconduct, and thus the
exclusionary rule should not apply.105
As explained by the majority, detective Fackrell made a good
faith mistake.106 While he had no particularized or individualized
suspicion of Strieff, the existing—fortuitous—arrest warrant cleansed
his constitutional error.107 But critical to this reasoning was the
absence of evidence of any recurring unconstitutional practice. Justice
Thomas emphasized that the stop was merely an “isolated incident of
negligence” stemming from a legitimate investigation, rather than
101. Id. at 2061 (“Evidence is admissible when the connection between unconstitutional
police conduct and the evidence is remote or has been interrupted by some intervening
circumstance, so that ‘the interest protected by the constitutional guarantee that has been
violated would not be served by suppression of the evidence obtained.’ ” (quoting Hudson v.
Michigan, 547 U.S. 586, 593 (2006))).
102. Id. at 2061–64 (“First, we look to the ‘temporal proximity’ between the unconstitutional
conduct and the discovery of evidence to determine how closely the discovery of evidence followed
the unconstitutional search. Second, we consider ‘the presence of intervening circumstances.’
Third, . . . we examine ‘the purpose and flagrancy of the official misconduct.’ ” (citations omitted)
(quoting Brown v. Illinois, 422 U.S. 590, 603–04 (1975))).
103. Id. at 2063 (“Officer Fackrell was at most negligent.”). But see id. at 2068 (Sotomayor,
J., dissenting) (“[T]he Fourth Amendment does not tolerate an officer’s unreasonable searches
and seizures just because he did not know any better. Even officers prone to negligence can learn
from courts that exclude illegally obtained evidence.”).
104. Id. at 2063 (majority opinion) (“Moreover, there is no indication that this unlawful stop
was part of any systemic or recurrent police misconduct.”).
105. See Herring, 555 U.S. at 147 (explaining that no evidence of widespread errors existed
in the warrant database).
106. “Good faith” here is a term of art borrowed from United States v. Leon, 468 U.S. 897
(1984), which borrowed the term from civil Section 1983 cases. See, e.g., Davis v. United States,
564 U.S. 229, 238–39 (2011) (describing the use of good faith in the Supreme Court’s
exclusionary rule cases); see also Laurin, supra note 60, at 739–42 (discussing the concept of
constitutional borrowing in the context of the exclusionary rule’s adoption of civil tort
terminology).
107. Strieff, 136 S. Ct. at 2063:
In stopping Strieff, Officer Fackrell made two good-faith mistakes. First, he had not
observed what time Strieff entered the suspected drug house, so he did not know how
long Strieff had been there. Officer Fackrell thus lacked a sufficient basis to conclude
that Strieff was a short-term visitor who may have been consummating a drug
transaction. Second, because he lacked confirmation that Strieff was a short-term
visitor, Officer Fackrell should have asked Strieff whether he would speak with him,
instead of demanding that Strieff do so.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 579
part of a systematic problem.108 Justice Thomas’s statement implicitly
recognized that had there been a systemic or recurring pattern of
police misconduct, the result might have been different.109
In dissent, Justice Sotomayor made this focus on recurring
practices explicit. Challenging the majority’s interpretation of the
evidence and narrow frame of analysis, she asserted that the incident
was not isolated at all.110 To support the argument that Fackrell’s
actions were both part of a local practice of unconstitutional stops111
and part of a national practice of exploiting low-level arrest
warrants,112 Justice Sotomayor broadened the focus to look at the
national context detailing recurring and systemic problems of
unconstitutional stops:
I do not doubt that most officers act in “good faith” and do not set out to break the law.
That does not mean these stops are “isolated instance[s] of negligence,” however. Many
are the product of institutionalized training procedures. The New York City Police
Department long trained officers to, in the words of a District Judge, “stop and question
first, develop reasonable suspicion later.”113
Equally important, Justice Sotomayor faulted the majority for
failing to articulate how any indigent litigant like Strieff could prove a
108. Id.
109. Id. at 2064:
Strieff argues that, because of the prevalence of outstanding arrest warrants in many
jurisdictions, police will engage in dragnet searches if the exclusionary rule is not
applied. We think that this outcome is unlikely. Such wanton conduct would expose
police to civil liability. And in any event, the Brown factors take account of the
purpose and flagrancy of police misconduct. Were evidence of a dragnet search
presented here, the application of the Brown factors could be different. But there is no
evidence that the concerns that Strieff raises with the criminal justice system are
present in South Salt Lake City, Utah.
(citations omitted).
110. See id. at 2068 (Sotomayor, J., dissenting) (highlighting the prevalence of outstanding
warrants across the United States).
111. See id. at 2069.
112. Id. at 2068 (“Justice Department investigations across the country have illustrated how
these astounding numbers of warrants can be used by police to stop people without cause.”).
113. Id. at 2069 (Sotomayor, J., dissenting) (citation omitted) (quoting Ligon v. City of New
York, 925 F. Supp. 2d 478, 537–38 (S.D.N.Y. 2013), stay granted on other grounds, 736 F.3d 118
(2d Cir. 2013)); see also id. at 2068:
The States and Federal Government maintain databases with over 7.8 million
outstanding warrants, the vast majority of which appear to be for minor offenses. . . .
The county in this case has had a “backlog” of such warrants. . . . Justice Department
investigations across the country have illustrated how these astounding numbers of
warrants can be used by police to stop people without cause.
(citations omitted); id. at 2073 (Kagan, J., dissenting) (“In other words, the department’s
standard detention procedures—stop, ask for identification, run a check—are partly designed to
find outstanding warrants. And find them they will, given the staggering number of such
warrants on the books.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
580 VANDERBILT LAW REVIEW [Vol. 72:2:561
systemic problem.114 She pointed out that there were “countless other
examples” of situations like Strieff’s, and yet the majority insisted it
was an isolated incident.115 “Surely,” she asserted, “it should not take
a federal investigation of Salt Lake County before the Court would
protect someone in Strieff’s position.”116 Echoing Justice Ginsburg’s
concern in Herring that traditional Fourth Amendment motions
practice makes proving repeated error too difficult, Justice Sotomayor
questioned how anyone could prove a systemic violation without
outside assistance.117
While clearly diverging in result and reasoning, both the
majority and dissent appear to recognize that systemic or recurring
unconstitutional violations would make exclusion more likely. In cases
of proven recurring police misconduct, exclusion would be an
appropriate remedy. In fact, even without a deliberate, reckless, or
grossly negligent act, if a systemic or recurring problem is proven, all
subsequent negligent unconstitutional actions will warrant
exclusion.118
B. The Reality of Recurring Problems
The Supreme Court’s openness to considering systemic and
recurring police negligence can be understood, in part, as a response to
a developing national awareness about police misconduct. During oral
argument in Strieff, the Justices explicitly brought up facts from the
Department of Justice Civil Rights Division’s Ferguson Report
exposing the pattern and practice of unconstitutional policing
practices in Ferguson, Missouri.119 In dissent, Justice Sotomayor
specifically referenced the federal litigation declaring the New York
114. Id. at 2069 (Sotomayor, J., dissenting).
115. Id.
116. Id.
117. Justice Sotomayor’s dissent concluded with impassioned language citing W.E.B.
DuBois, James Baldwin, Michelle Alexander, Ta’nehisi Coates, Lani Guinier, and Gerald Torres
that this absence of police accountability will further racial discrimination and justify
unconstitutional practices. Id. at 2069–71:
We must not pretend that the countless people who are routinely targeted by police
are “isolated.” They are the canaries in the coal mine whose deaths, civil and literal,
warn us that no one can breathe in this atmosphere. They are the ones who recognize
that unlawful police stops corrode all our civil liberties and threaten all our lives.
Until their voices matter too, our justice system will continue to be anything but.
(citation omitted) (citing LANI GUINIER & GERALD TORRES, THE MINER’S CANARY 274–83 (2002)).
118. See Laurin, supra note 60, at 687 (detailing how the exclusionary rule serves to deter
recurring or systemic negligence in some circumstances).
119. Transcript of Oral Argument at 6, Strieff, 136 S. Ct. 2056 (No. 14-1373) (discussing the
fact that approximately eighty percent of the minority population in Ferguson had an
outstanding municipal warrant, making the stop-and-identify practice quite tempting).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 581
Police Department’s stop and frisk policy unconstitutional.120 In
Herring, Justice Ginsberg raised the growing problem of systemic data
error.121 Obviously, these national issues and well-publicized
investigations involved particular police departments at particular
moments in time, but the documented problems of unconstitutional
stops, racial bias, and excessive use of force suggest a reason for
concern.122
Much has been written on police reform in the last few years.123
This Section briefly examines two related but opposing problems with
police accountability: first, a lack of data about policing in general124
and, second, the repeated findings of systemic and recurring problems
in specific investigations into particular police departments.125 Both
120. Strieff, 136 S. Ct. at 2069 (Sotomayor, J., dissenting) (citing Ligon v. City of New York,
925 F. Supp. 2d 478, 537–38 (S.D.N.Y. 2013)).
121. See Herring v. United States, 555 U.S. 135, 155 (2009) (Ginsburg, J., dissenting)
(“Herring’s amici warn that law enforcement databases are insufficiently monitored and often
out of date.”).
122. See generally Kami Chavis Simmons, Cooperative Federalism and Police Reform: Using
Congressional Spending Power to Promote Police Accountability, 62 ALA. L. REV. 351, 357 (2011)
(“Despite the many efforts to reform local police departments and to increase police
accountability, police misconduct and corruption persist in the United States.”); Samuel Walker,
Institutionalizing Police Accountability Reforms: The Problem of Making Police Reforms Endure,
32 ST. LOUIS U. PUB. L. REV. 57, 76–77 (2012) (discussing early intervention (“EI”) systems,
which provide data to identify officers with performance problems, as emerging as a powerful
police accountability tool).
123. See, e.g., Bernard E. Harcourt & Jens Ludwig, Broken Windows: New Evidence from
New York City and a Five-City Social Experiment, 73 U. CHI. L. REV. 271, 272–73 (2006)
(discussing empirical studies of broken window policing); Tracey L. Meares, The Law and Social
Science of Stop and Frisk, 10 ANN. REV. L. & SOC. SCI. 335, 341 (2014) (concluding that police
stops in New York City correlated more to racial composition of a neighborhood than crime rate);
Kami Chavis Simmons, The Politics of Policing: Ensuring Stakeholder Collaboration in the
Federal Reform of Local Law Enforcement Agencies, 98 J. CRIM. L. & CRIMINOLOGY 489, 496
(2008) (arguing that police violence in the United States is a systematic problem); David Alan
Sklansky, Not Your Father’s Police Department: Making Sense of the New Demographics of Law
Enforcement, 96 J. CRIM. L. & CRIMINOLOGY 1209, 1213–15 (2006) (discussing how racial
minorities compose a significantly increased share of urban police forces since the 1960s); David
Alan Sklansky, Police and Democracy, 103 MICH. L. REV. 1699, 1742 (2005) [hereinafter
Sklansky, Police and Democracy] (providing a history of police reform); Samuel Walker,
Governing the American Police: Wrestling with the Problems of Democracy, 2016 U. CHI. LEGAL F.
615, 618 (detailing how law enforcement is organized in the United States).
124. See, e.g., Rachel Harmon, Why Do We (Still) Lack Data on Policing?, 96 MARQ. L. REV.
1119, 1121 (2013) (arguing that “today we still lack enough information about what the police do
to shape their conduct effectively”).
125. See CIVIL RIGHTS DIV., U.S. DEP’T OF JUSTICE, INVESTIGATION OF THE FERGUSON POLICE
DEPARTMENT 2–3 (Mar. 15, 2015), https://www.justice.gov/sites/default/files/opa/press-releases/
attachments/2015/03/04/ferguson_police_department_report.pdf [https://perma.cc/W7NS-9CSB]
[hereinafter DOJ FERGUSON REPORT] (finding the Ferguson Police Department’s policy to have
been geared toward aggressive enforcement, with officers demanding compliance when they lack
legal authority); CIVIL RIGHTS DIV., U.S. DEP’T OF JUSTICE, INVESTIGATION OF THE BALTIMORE
CITY POLICE DEPARTMENT 24 (Aug. 10, 2016), https://www.justice.gov/crt/file/883296/download
[https://perma.cc/U4CT-49ZN] [hereinafter DOJ BALTIMORE REPORT] (finding that the Baltimore
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
582 VANDERBILT LAW REVIEW [Vol. 72:2:561
this general lack of knowledge and the specific concern about
recurring problems inspire this Article’s attempt to find new
technological ways to expose, identify, and monitor police misconduct,
which is the subject of Part II.
1. The Fragmented Nature of Police Data
As a constitutional matter, policing is studied in fragments. We
know what detective Fackrell did when he stopped Edward Strieff, but
not what he did before, or after, or any other day of his career. The
record established facts without context about routine practice,
training, or comparative circumstances.126 This fragmented moment of
time is further split by the localized nature of policing. There are
approximately eighteen thousand separate police forces in the United
States, each with different protocols, rules, and cultures.127 What one
detective does in Salt Lake City may not be done in Miami, or
Minneapolis, or Missoula.
If courts cannot track what police do on the streets, one might
think that governments might systemically monitor police practices.
But federal and state efforts to collect data have been similarly
fragmented. Professor Rachel Harmon has expressed dismay at the
lack of state and federal data on police.128 Particularly in the states,
which extensively regulate police, one might expect more information
about police actions. But that is not the case. Even federal data is far
too limited to provide any meaningful assistance to the government in
its oversight of police activity.129 Police leaders do not always
encourage transparency,130 and police unions131 and other employment
Police Department (“BPD”) “engages in a pattern or practice of making stops, searches, and
arrests in violation of the Fourth and Fourteenth Amendments and Section 14141”).
126. This myopic approach has been exposed by scholars who understand policing as a
product of systemic choices and strategies and not isolated incidents. See Tracey L. Meares,
Programming Errors: Understanding the Constitutionality of Stop-and-Frisk as a Program, Not
an Incident, 82 U. CHI. L. REV. 159, 162 (2015) (“[W]hile the Court in Terry authorized police
intervention in an individual incident when a police officer possesses less than probable cause to
believe that an armed individual is involved in a crime, in reality stop-and-frisk is more typically
carried out by a police force en masse as a program.”).
127. Barry Friedman & Maria Ponomarenko, Democratic Policing, 90 N.Y.U. L. REV. 1827,
1843 (2015) (“Policing in the United States is a diffuse business. . . . [M]uch of policing occurs at
the local level. There are just under 18,000 separate police forces in the United States, and some
765,000 sworn officers.”).
128. Harmon, supra note 124, at 1129.
129. See id. at 1122, 1132–33 (“[W]hile existing federal law and agency efforts provide for
some data collection about policing, those efforts are flawed, stymied by institutional and legal
limitations.”).
130. Id. at 1129:
In practice, police chiefs and other local government actors often limit rather than
promote information availability. Cities and police departments sometimes actively
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 583
or privacy laws further restrict access to information regarding alleged
officer misconduct.132 Municipalities that suffer the financial liability
for police misconduct remain unenthused about collecting data as it
could be used against them in court.133 And even when certain
jurisdictions do collect data, or better yet, restructure their police force
to focus on data collection and analysis,134 this data is not integrated
or compared nationally to other police departments.135
In the context of addressing allegations of racial profiling, a few
data collection systems have been imposed by court order.136 In terms
of systemic abuse of unconstitutional stops and frisks, some data has
been revealed through civil rights lawsuits.137 Regarding structural
inhibit the collection of information about police by, for example, requiring secrecy
when they settle civil suits for police misconduct or discouraging citizens from filing
complaints about officer conduct.
131. Stephen Rushin, Using Data to Reduce Police Violence, 57 B.C. L. REV. 117, 153 (2016)
(“[C]ollective bargaining and civil service protections inadvertently discourage police
management from responding forcefully to misconduct.”); Walker, supra note 122, at 72:
Collective bargaining agreements, for example, contain provisions related to the
investigation of alleged officer misconduct (whether on the basis of a citizen complaint
or an internally generated complaint) that impede a timely and thorough
investigation. Officer appeals of discipline, meanwhile, may involve procedures that
tend to increase the likelihood of disciplinary sanctions being mitigated or overturned.
(footnote omitted).
132. Harmon, supra note 124, at 1133 (2013) (“[S]tates not only do little to encourage police
departments to produce information about policing that does exist, they also often restrict public
access to it through privacy laws and exemptions from open records statutes.”).
133. See Myriam E. Gilles, Breaking the Code of Silence: Rediscovering “Custom” in Section
1983 Municipal Liability, 80 B.U. L. REV. 17, 31 (2000) (“Municipalities generally write off the
misconduct of an individual officer to the ‘bad apple theory,’ under which municipal governments
or their agencies attribute misconduct to aberrant behavior by a single ‘bad apple,’ thereby
deflecting attention from systemic and institutional factors contributing to recurring
constitutional deprivations.”); Joanna C. Schwartz, Myths and Mechanics of Deterrence: The Role
of Lawsuits in Law Enforcement Decisionmaking, 57 UCLA L. REV. 1023, 1063–64 (2010)
(explaining that in many jurisdictions police departments will suspend internal review of citizen
complaints if the department is sued).
134. See James J. Willis et al., Making Sense of COMPSTAT: A Theory-Based Analysis of
Organizational Change in Three Police Departments, 41 LAW & SOC’Y REV. 147, 148 (2007)
(introducing a data collection tool employed by the New York City Police Department to reduce
crime by keeping officers accountable for crime reduction).
135. Harmon, supra note 124, at 1129 (“Even when departments collect information, they
may do so in ways that make it impossible to aggregate the records or compare them with data
from other departments. Departments often, for example, keep only paper files and use
anomalous report forms and categories . . . .”).
136. Mary D. Fan, Panopticism for Police: Structural Reform Bargaining and Police
Regulation by Data-Driven Surveillance, 87 WASH. L. REV. 93, 127 (2012) (“Many of the reforms
in cases involving recurrent problems such as excessive force or racial targeting call for police to
report uses of force, demographic information, and bases for investigative stops and searches.
The methods of regulation and remedies are shifting to information and data-driven surveillance
of police practices.” (footnotes omitted)).
137. Bailey v. City of Philadelphia, No. 10-5952 (E.D. Pa. June 21, 2011) (order approving
settlement agreement, class certification, and consent decree), https://www.aclupa.org/download
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
584 VANDERBILT LAW REVIEW [Vol. 72:2:561
change and consent degrees, the data collection piece has been
imposed by federal sanction.138 Even when it comes to law
enforcement’s ultimate power—use of deadly force—no national
system exists to track police use of force or killings.139 This absence
forced the United States Attorney General140 and the Director of the
FBI141 to separately admit embarrassment at not being able to provide
the statistics to interested parties.142 Instead, the stories of police
violence—both tragic and justified—become part of an anecdotal and
fragmented policing landscape.143
_file/view_inline/744/198 [https://perma.cc/8Q5S-VZ55]; Plaintiffs’ First Report to Court and
Master on Stop and Frisk Practices at 7, Bailey, No. 10-5952 (Nov. 4, 2010),
https://www.law.columbia.edu/sites/default/files/microsites/contract-economic-organization/
files/Bailey First Report_final version.docx [https://perma.cc/T4HB-XQR7]; David A. Harris,
Across the Hudson: Taking the Stop and Frisk Debate Beyond New York City, 16 N.Y.U. J. LEGIS.
& PUB. POL’Y 853, 865 (2013):
But in point of fact, data collection on stops and frisks in the U.S. has been relatively
rare. . . . All in all, in most police departments there has been virtually no systematic,
organized effort to collect information on the practice in a way that gives big-picture
insight into what police are doing.
138. Fan, supra note 136, at 127–28 (2012); Stephen Rushin, Structural Reform Litigation in
American Police Departments, 99 MINN. L. REV. 1343, 1347 (2015) (“[T]he federal government
can now use equitable relief to force problematic police agencies to adopt significant structural,
procedural, and policy reforms aimed at curbing misconduct.”).
139. Matthew J. Hickman, Alex R. Piquero & Joel H. Garner, Toward a National Estimate of
Police Use of Nonlethal Force, 7 CRIMINOLOGY & PUB. POL’Y 563, 565 (2008) (“[L]ocal, state, and
federal governments actually collect and report very little information about police use of force,
much less than about police behavior in general.”); Rob Barry & Coulter Jones, Hundreds of
Police Killings Are Uncounted in Federal Stats, WALL STREET J. (Dec. 3, 2014, 11:26 AM),
http://www.wsj.com/articles/hundreds-of-police-killings-are-uncounted-in-federal-statistics-
1417577504 [https://perma.cc/PLB2-E9AM]; Wesley Lowery, How Many Police Shootings a Year?
No One Knows, WASH. POST (Sept. 8, 2014), http://www.washingtonpost.com/news/post-nation/
wp/2014/09/08/how-many-police-shootings-a-year-no-one-knows [https://perma.cc/4439-C492].
140. Jon Swaine, Eric Holder Calls Failure to Collect Reliable Data on Police Killings
Unacceptable, GUARDIAN (Jan. 15, 2015), http://www.theguardian.com/us-news/2015/jan/15/eric-
holder-no-reliable-fbi-data-police-related-killings [http://perma.cc/3YUB-CFFP] (reporting Eric
Holder as stating that “[t]he troubling reality is that we lack the ability right now to
comprehensively track the number of incidents of either uses of force directed at police officers or
uses of force by police,” and saying “[t]his strikes many – including me – as unacceptable”).
141. Michael S. Schmidt, F.B.I. Director Speaks Out on Race and Police Bias, N.Y. TIMES
(Feb. 12, 2015), http://www.nytimes.com/2015/02/13/us/politics/fbi-director-comey-speaks-frankly
-about-police-view-of-blacks.html [https://perma.cc/HJ5T-6BVQ] (quoting James Comey as
saying, “It’s ridiculous that I can’t tell you how many people were shot by the police last week,
last month, last year.”).
142. Rushin, supra note 131, at 126 (“It seems incongruent for the federal government to
keep detailed records on the number of law enforcement officers killed or assaulted in the line of
duty, but not keep comparable records on citizens killed or assaulted by law enforcement.”
(footnote omitted)).
143. Editorial Board, One Thing the U.S. Government Doesn’t Count: How Often Police Kill
Civilians, L.A. TIMES (Dec. 16, 2014), http://www.latimes.com/opinion/editorials/la-ed-police-
statistics-20141217-story.html [http://perma.cc/9W9R-4LTC] (“But one thing the government
doesn’t count, as was spotlighted this summer amid the fallout from Michael Brown’s shooting
death in Ferguson, Mo., is how often police kill civilians.”); The Counted: People Killed by Police
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 585
2. Evidence of Recurring Problems
In the absence of good data about policing practices, one could
hope that recurring patterns of police misconduct would not be
prevalent. Yet evidence indicates otherwise.144 For example, in civil
rights lawsuits challenging unconstitutional stops in New York City
and Philadelphia, repeated Fourth Amendment violations were
documented.145
Rather famously, Judge Shira Scheindlin declared the New
York City Police Department’s (“NYPD”) stop and frisk practice
unconstitutional, finding that New York City was liable for violating
plaintiffs’ Fourth and Fourteenth Amendment rights.146 The court
found that the City acted with deliberate indifference toward the
NYPD’s practice of making unconstitutional stops and conducting
unconstitutional frisks.147 At trial, police data became key to
establishing the racially discriminatory caste of constitutional
violations. The trial record showed that the NYPD “made 4.4 million
stops between January 2004 and June 2012. Over 80% of these 4.4
million stops were of blacks or Hispanics.”148 Of those stops, 52%
involved frisks, but a weapon was only recovered 1.5% of the time,
in the US, GUARDIAN, http://www.theguardian.com/us-news/ng-interactive/2015/jun/01/the-
counted-police-killings-us-database (last visited Oct. 19, 2018) [https://perma.cc/3XW6-UMKY]
(providing a list of each individual killed by police in 2015 and 2016).
144. Andrew Gelman, Jeffrey Fagan & Alex Kiss, An Analysis of the New York City Police
Department’s “Stop-and-Frisk” Policy in the Context of Claims of Racial Bias, 102 J. AM. STAT.
ASS’N 813, 821 (2007) (“In the period for which we had data, the NYPD’s records indicate that
they were stopping blacks and Hispanics more often than whites, in comparison to both the
populations of these groups and the best estimates of the rate of crimes committed by each
group.”); K. Babe Howell, Broken Lives From Broken Windows: The Hidden Costs of Aggressive
Order-Maintenance Policing, 33 N.Y.U. REV. L. & SOC. CHANGE 271, 276–80 (2009) (discussing
the various downfalls associated with zero-tolerance policing).
145. Floyd v. City of New York, 959 F. Supp. 2d 540, 562 (S.D.N.Y. 2013) (“The City acted
with deliberate indifference toward the NYPD’s practice of making unconstitutional stops and
conducting unconstitutional frisks.”); Ligon v. City of New York, 925 F. Supp. 2d 478, 492–510
(S.D.N.Y. 2013) (providing evidence of nine independent police stops illustrating misconduct);
Davis v. City of New York, 902 F. Supp. 2d 405, 412–30 (S.D.N.Y. 2012) (providing seven unique
instances of NPYD misconduct); Bailey v. City of Philadelphia, No. 10-5952 (E.D. Pa. June 21,
2011) (order approving settlement agreement, class certification, and consent decree),
https://www.aclupa.org/download_file/view_inline/744/198 [https://perma.cc/8Q5S-VZ55]; Daniels
v. City of New York, 198 F.R.D. 409, 412 (S.D.N.Y. 2001) (documenting allegations of misconduct
in as many as 18,000 instances).
146. Floyd, 959 F. Supp. 2d at 562 (finding that out of nineteen instances, nine stops were
unconstitutional, five frisks were unconstitutional, and the rest were constitutional stop and
frisks).
147. Id.; see id. at 660 (“The NYPD’s practice of making stops that lack individualized
reasonable suspicion has been so pervasive and persistent as to become not only a part of the
NYPD’s standard operating procedure, but a fact of daily life in some New York City
neighborhoods.”).
148. Id. at 556.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
586 VANDERBILT LAW REVIEW [Vol. 72:2:561
meaning “in 98.5% of the 2.3 million frisks, no weapon was found.”149
In only 6% of the stops was a suspect arrested.150 Despite being
stopped more often, contraband was found less often on Blacks and
Hispanics compared to Whites.151 While the collected stop data made
the lawsuit possible, it also demonstrated a pattern of systemic and
recurring constitutional violations.152
In Philadelphia, a lawsuit challenged the practice of
Philadelphia police officers stopping individuals without constitutional
justification.153 The lawsuit documented a recurring pattern of Fourth
Amendment violations and resulted in a consent decree requiring
further monitoring.154 Somewhat troublingly, despite being under
court-ordered monitoring, “one-half of all stops were made without the
requisite reasonable suspicion and . . . over one-half of all frisks were
made without reasonable suspicion.”155 These recurring patterns of
constitutional violations continued in 2011, 2012, and 2013.156
149. Id. at 558.
150. Id. at 558–59 (“6% of all stops resulted in an arrest, and 6% resulted in a summons. The
remaining 88% of the 4.4 million stops resulted in no further law enforcement action.”).
151. Id. at 559 (“In 52% of the 4.4 million stops, the person stopped was black, in 31% the
person was Hispanic, and in 10% the person was white. . . . Contraband other than weapons was
seized in 1.8% of the stops of blacks, 1.7% of the stops of Hispanics, and 2.3% of the stops of
whites.”).
152. See generally Jeffrey Bellin, The Inverse Relationship Between the Constitutionality and
Effectiveness of New York City “Stop and Frisk,” 94 B.U. L. REV. 1495, 1541 (2014) (“The NYPD’s
use of stop-and-frisk to deter people from carrying weapons runs afoul of another constitutional
provision: the Fourteenth Amendment’s Equal Protection Clause.”); Jeffrey Fagan & Amanda
Geller, Following the Script: Narratives of Suspicion in Terry Stops in Street Policing, 82 U. CHI.
L. REV. 51, 69 (2015) (providing empirical data to demonstrate violations of the Fourth
Amendment and Fourteenth Amendment); Harris, supra note 137, at 854–57 (highlighting the
case law establishing the NYPD’s systemic violations of the Constitution).
153. Complaint, Bailey v. City of Philadelphia, No. 10-5925 (E.D. Pa. Nov. 4, 2010), 2010 WL
4662865.
154. Bailey, No. 10-5952 (order approving settlement agreement, class certification, and
consent decree), https://www.aclupa.org/download_file/view_inline/744/198 [https://perma.cc/
8Q5S-VZ55] (requiring a “data base [that] shall have the capability to retrieve Information by
DC number, district, date, race, officer’s actions, and other relevant characteristics necessary to
effective monitoring of stop and frisk practices”).
155. See Plaintiffs’ First Report to Court and Master on Stop and Frisk Practices, supra note
137, at 7 (emphasis omitted); see id. at 8:
In sum, over the first six months of 2011, based on the 1426 75-48a forms reviewed by
counsel (a larger number were reviewed by law students with similar findings), 713
pedestrian stops were made with reasonable suspicion and 713 were made without
reasonable suspicion. Of 355 frisks, 165 were with reasonable suspicion and 190
without reasonable suspicion.
156. See sources cited supra note 155; see also Plaintiffs’ Fourth Report to Court and Monitor
on Stop and Frisk Practices at 7, Bailey, No. 10-5952 (Dec. 3, 2013), https://www.aclupa.org/
download_file/view_inline/1529/198 [https://perma.cc/7YZ9-WAE5] (“43% of all stops and over
50% of all frisks were made without the requisite reasonable suspicion. These results are not
appreciably different from the data reviews for 2011 and 2012, as set forth in the First, Second,
and Third Reports.”); Plaintiffs’ Third Report to Court and Monitor on Stop and Frisk Practices
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 587
Additionally, a series of investigations by the Department of
Justice Civil Rights Division uncovered systemic problems touching on
unconstitutional stops, use of force, and racial discrimination.157
During 2015 through 2017, the DOJ Civil Rights Division investigated
the Chicago Police Department (“CPD”),158 the City of Baltimore Police
Department,159 and the Ferguson Police Department160 and ultimately
uncovered systemic and recurring constitutional violations that led to
ongoing federal oversight.161 Read in total, these lengthy, in-depth
reports offer a devastating critique of local policing practices and an
equally damning account of the lack of accountability of police
administrators.
For example, the DOJ found that Chicago police officers
“engaged in a pattern or practice of unreasonable force in violation of
the Fourth Amendment and that the deficiencies in CPD’s training,
supervision, accountability, and other systems have contributed to
that pattern or practice.”162 This force was not the product of
individual “bad apples,” but “largely attributable to systemic
deficiencies.”163 The misconduct was routine,164 largely ignored by the
at 8, Bailey, No. 10-5952, https://www.aclupa.org/download_file/view_inline/1015/198
[https://perma.cc/XHL2-LHYN]:
It is remarkable that 43-47% of all stops and over 45% of all frisks were made without
the requisite reasonable suspicion. These results are not appreciably different from
the data reviews for 2011, as set forth in the First and Second Reports. Thus, tens of
thousands of persons in Philadelphia continue to be stopped each year (and a
significant number frisked) without reasonable suspicion.
157. Rachel Moran, Ending the Internal Affairs Farce, 64 BUFF. L. REV. 837, 847–48 (2016):
Recent investigations by the DOJ’s Civil Rights Division have revealed that officers in
many cities use unconstitutionally excessive force during their encounters with
minorities, stop and frisk minorities without any legal justification, systematically
arrest and charge minorities for nonviolent crimes far more aggressively than they
enforce similar crimes in white communities, and arrest poor minorities—subjecting
many of them to jail time—for minor unpaid fines.
158. CIVIL RIGHTS DIV., U.S. DEP’T OF JUSTICE, INVESTIGATION OF THE CHICAGO POLICE
DEPARTMENT (Jan. 13, 2017), https://www.justice.gov/opa/file/925846/download [https://perma.cc/
U8W6-6C9G] [hereinafter DOJ CHICAGO REPORT].
159. DOJ BALTIMORE REPORT, supra note 125, at 24.
160. DOJ FERGUSON REPORT, supra note 125, at 23.
161. Sunita Patel, Toward Democratic Police Reform: A Vision for “Community Engagement”
Provisions in DOJ Consent Decrees, 51 WAKE FOREST L. REV. 793, 794 (2016) (“[T]he Obama
administration has invigorated the Civil Rights Division of the Department of Justice, with
particular emphasis placed on the Special Litigation Section’s involvement in police reform. The
Special Litigation Section has opened thirty-six investigations and signed approximately twenty-
one agreements or intent to reach agreements with various localities.”).
162. DOJ CHICAGO REPORT, supra note 158, at 23.
163. Id. at 5.
164. Id. (“The pattern of unlawful force we found resulted from a collection of poor police
practices that our investigation indicated are used routinely within CPD.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
588 VANDERBILT LAW REVIEW [Vol. 72:2:561
city,165 endemic to the force,166 and directed predominantly at people of
color.167 Practices involving both deadly force168 and nondeadly force169
violated the Constitution. Traditional accountability mechanisms
failed to remedy misconduct,170 and police failed to develop training or
supervision systems to improve the problem.171
In Baltimore, the DOJ revealed a recurring pattern of
unconstitutional stops, frisks, and arrests in violation of the Fourth
Amendment.172 Police stopped citizens without reasonable
suspicion.173 Police frisked individuals without a belief that the person
was armed and dangerous.174 Police arrested people without cause.175
In fact, similar to Utah v. Strieff, the DOJ discovered a pattern of
unconstitutional stops to run warrant checks, finding that officers
regularly approached, detained, and questioned individuals on the
sidewalk without reasonable suspicion.176 These unconstitutional
165. Id. at 7 (“The City received over 30,000 complaints of police misconduct during the five
years preceding our investigation, but fewer than 2% were sustained, resulting in no discipline in
98% of these complaints. This is a low sustained rate.”).
166. Id. at 8 (“We discovered numerous entrenched, systemic policies and practices that
undermine police accountability.”).
167. Id. at 145 (“Blacks, Latinos, and whites make up approximately equal thirds of the
population in Chicago, but the raw statistics show that CPD uses force almost ten times more
often against blacks than against whites.”).
168. Id. at 5 (“CPD officers engage in a pattern or practice of using force, including deadly
force, that is unreasonable.”).
169. Id. at 32 (“Although CPD documents generally include insufficient detail of when and
how officers use force, particularly less-lethal force, our review of CPD records made clear that
CPD’s pattern of unreasonable force includes unreasonable less-lethal force.”).
170. Id. at 47 (“Our investigation confirmed that CPD’s accountability systems are broadly
ineffective at deterring or detecting misconduct, and at holding officers accountable when they
violate the law or CPD policy.”).
171. Id. at 10 (“CPD’s pattern of unlawful conduct is due in part to deficiencies in CPD’s
training and supervision. CPD does not provide officers or supervisors with adequate training
and does not encourage or facilitate adequate supervision of officers in the field.”).
172. DOJ BALTIMORE REPORT, supra note 125, at 24 (“We find that BPD engages in a pattern
or practice of making stops, searches, and arrests in violation of the Fourth and Fourteenth
Amendments and Section 14141.”).
173. Id. at 6 (“BPD’s stops often lack reasonable suspicion.”).
174. Id. at 30 (“BPD officers commonly frisk people during stops without reasonable
suspicion that the subject of the frisk is armed and dangerous.”); see id. at 6 (“During stops, BPD
officers frequently pat-down or frisk individuals as a matter of course, without identifying
necessary grounds to believe that the person is armed and dangerous. And even where an initial
frisk is justified, we found that officers often violate the Constitution by exceeding the frisk’s
permissible scope.”).
175. Id. at 34 (“Our investigation likewise found reasonable cause to believe that BPD’s
approach to street-level crime suppression has contributed to officers making thousands of
unlawful arrests over the past five years.”).
176. Id. at 28 (“Many of the unlawful stops we identified appear motivated at least in part by
officers’ desire to check whether the stopped individuals have outstanding warrants that would
allow officers to make an arrest or search individuals in hopes of finding illegal firearms or
narcotics.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 589
stops exposed a racial bias.177 As the DOJ found, “[r]acially disparate
impact is present at every stage of [the Baltimore Police
Department’s] enforcement actions, from the initial decision to stop
individuals on Baltimore streets to searches, arrests, and uses of
force.”178 Again, the DOJ did not find that these recurring
unconstitutional actions were isolated but instead found them to be
part of systemic, structural problems.179 Accountability systems
largely failed,180 and training181 and internal disciplinary systems
were demonstrated to be woefully lacking.182
Much has already been written about the Ferguson Police
Department,183 as the DOJ’s investigation revealed a practice of
177. Id. at 7 (“BPD disproportionately searches African Americans during stops. BPD
searched African Americans more frequently during pedestrian and vehicle stops, even though
searches of African Americans were less likely to discover contraband.”).
178. Id.; see also id. (“Citywide, BPD stopped African-American residents three times as
often as white residents after controlling for the population of the area in which the stops
occurred.”); id. (“African Americans accounted for 91 percent of the 1,800 people charged solely
with ‘failure to obey’ or ‘trespassing’; 89 percent of the 1,350 charges for making a false
statement to an officer; and 84 percent of the 6,500 people arrested for ‘disorderly conduct.’ ”).
179. Id. at 10 (“BPD’s systemic constitutional and statutory violations are rooted in
structural failures. BPD fails to use adequate policies, training, supervision, data collection,
analysis, and accountability systems, has not engaged adequately with the community it polices,
and does not provide its officers with the tools needed to police effectively.”).
180. Id. (“BPD lacks meaningful accountability systems to deter misconduct. The
Department does not consistently classify, investigate, adjudicate, and document complaints of
misconduct according to its own policies and accepted law enforcement standards.”); see also id.
at 134 (“Moreover, BPD conducts minimal pattern analysis of officer activities. The Department
does not generate any reports or otherwise track patterns in officers’ stops, searches, arrests,
uses of force, or community interactions.”).
181. Id. at 43 (“BPD exacerbates the risk that its aggressive street enforcement tactics will
lead to constitutional violations by failing to use effective policies, training, oversight, and
accountability systems.”).
182. Id. at 135:
Despite BPD’s longstanding notice of concerns about its policing activities and
problems with its internal accountability systems, the Department has failed to
implement an adequate EIS or other system for tracking or auditing information
about officer conduct. Rather, BPD has an early intervention system in name only;
indeed, BPD commanders admitted to us that the Department’s early intervention
system is effectively nonfunctional.
183. See, e.g., John Felipe Acevedo, Restoring Community Dignity Following Police
Misconduct, 59 HOW. L.J. 621, 633 (2016) (“The shortcomings of the Ferguson Police Department
came to public attention following the killing of eighteen year old Michael Brown by police officer
Darren Wilson.”); Devon W. Carbado, Blue-on-Black Violence: A Provisional Model of Some of the
Causes, 104 GEO. L.J. 1479, 1502 (2016) (“Ferguson, Missouri presents a concrete example of the
ease with which predatory policing can become an institutional feature of everyday policing.”);
S. David Mitchell, Ferguson: Footnote or Transformative Event?, 80 MO. L. REV. 943, 944 (2015)
(“ ‘Ferguson.’ No longer does this name simply represent the geographical boundaries of a city in
St. Louis County formed initially by white flight from St. Louis City and that has become
increasingly African American over time. It has come to represent so much more.” (footnote
omitted)); Michael Pinard, Poor, Black and “Wanted”: Criminal Justice in Ferguson and
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
590 VANDERBILT LAW REVIEW [Vol. 72:2:561
prioritizing low-level arrests to generate revenue for the municipal
government.184 These arrests primarily impacted the African
American population185 and did little to reduce crime. But the DOJ
report also demonstrated a systemic bias in the use of force186 and in
recurring violations of the Fourth Amendment.187 Police stopped
people without reasonable suspicion as part of a larger system of
revenue collection.188 As in Chicago and Baltimore, the culture in
Ferguson created a system that allowed recurring police misconduct.
These federal investigations offer a deep dive into a few specific
departments. Because of poor data collection and the limitations of
civil rights lawsuits and federal investigations, however, we do not
know the extent of the national problem. But we do know that since
1994, with the enactment of 42 U.S.C. § 14141, the DOJ has opened 69
investigations and entered into 40 reform agreements.189 Just since
2012, the DOJ has “opened 11 new pattern-or-practice investigations
and negotiated 19 new reform agreements since 2012.”190
These investigations confirm that systemic and recurring
problems of racial discrimination, unconstitutional stops, and
excessive force remain issues to be addressed.191 The DOJ revelations
Baltimore, 58 HOW. L.J. 857, 862 (2015) (“Pathetically, at the time of the DOJ investigation, only
four out of fifty-four police officers in Ferguson were Black.”).
184. DOJ FERGUSON REPORT, supra note 125, at 15 (“FPD’s approach to law enforcement,
shaped by the City’s pressure to raise revenue, has resulted in a pattern and practice of
constitutional violations.”).
185. Id. at 4:
Ferguson’s law enforcement practices overwhelmingly impact African Americans.
Data collected by the Ferguson Police Department from 2012 to 2014 shows that
African Americans account for 85% of vehicle stops, 90% of citations, and 93% of
arrests made by FPD officers, despite comprising only 67% of Ferguson’s population.
African Americans are more than twice as likely as white drivers to be searched
during vehicle stops even after controlling for non-race based variables such as the
reason the vehicle stop was initiated, but are found in possession of contraband 26%
less often than white drivers, suggesting officers are impermissibly considering race
as a factor when determining whether to search.
186. Id. at 5 (“Nearly 90% of documented force used by FPD officers was used against
African Americans.”).
187. Id. at 15 (“Officers violate the Fourth Amendment in stopping people without
reasonable suspicion, arresting them without probable cause, and using unreasonable force.”).
188. Id. at 16 (“Frequently, officers stop people without reasonable suspicion or arrest them
without probable cause. Officers rely heavily on the municipal ‘Failure to Comply’ charge, which
appears to be facially unconstitutional in part, and is frequently abused in practice.”).
189. CIVIL RIGHTS DIV., U.S. DEP’T OF JUSTICE, THE CIVIL RIGHTS DIVISION’S PATTERN AND
PRACTICE POLICE REFORM WORK: 1994-PRESENT 3 (Jan. 2017), https://www.justice.gov/crt/
file/922421/download [https://perma.cc/QC3S-A792]; see also id. at 15 (“Of 69 total investigations
since Section 14141’s enactment, the Division has closed 26 investigations without making a
formal finding of a pattern or practice.”).
190. Id. at 1.
191. See I. Bennett Capers, Crime, Legitimacy, and Testilying, 83 IND. L.J. 835, 852 (2008)
(“Regardless of whether this race-based policing is intentional or not, there is the continuing
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 591
offer a glimpse of the promise and potential of better data collecting
mechanisms as a path to expose systemic or recurring patterns of
misconduct. And at least in cities with documented patterns of
unconstitutional stops, like Baltimore, Chicago, or Ferguson, such
data could be used for individual suppression hearings.192
C. Difficulties in Litigating Systemic or Recurring Violations
In the face of recurring police problems, and under the
Supreme Court’s constitutional command to be concerned with such
systematic transgressions, the question remains: Why do these issues
not manifest themselves in ordinary Fourth Amendment suppression
hearings?
One answer is that the Supreme Court did not explain how
these systemic or recurring patterns should be proven in court. The
trial records in Herring and Strieff offered few clues. In Herring, the
Dale County clerk had testified—somewhat imprecisely—that
communication problems had arisen “several times.”193 But because
Chief Justice Robert’s new test had not yet been written into law,
there was no reason to expend effort to prove systemic or recurring
problems at the trial level. The fact that database errors may have
occurred in prior cases or in other counties did not become part of the
record because it had not been identified as an important factor
relevant for exclusion.
In Strieff, the suppression hearing did not include any
testimony outside of the arresting officer’s.194 The arguments of the
parties focused on attenuation due to a lawful warrant, not principles
of error or deterrence.195 In fact, the motions hearing in Strieff offers a
revealing example of the sparse nature of these types of hearings.196
The Strieff suppression hearing consisted of one witness—detective
perception, supported by evidence, that police treat citizens differently based on their race.”);
Simmons, supra note 122, at 365 (“Empirical evidence supports the view that racial profiling is a
widespread practice of police officers in many communities.”).
192. As detailed in Part III, the utility of collecting such data is that it demonstrates
systemic or recurring problems. In a particular case involving a particular suppression issue,
this demonstrated pattern should be admissible to prove the systemic and recurring negligence
required under Herring.
193. United States v. Herring, 451 F. Supp. 2d 1290, 1292 (M.D. Ala. 2005), aff’d, 492 F.3d
1212 (11th Cir. 2007), aff’d, 555 U.S. 135 (2009) (“To be sure, during the first of two suppression
hearings, Morgan testified as follows: ‘Q. All right. Ma’am, how many times have you had or has
Dale County had problems, any problems with communicating about warrants?’ ‘A. Several
times.’ ”).
194. Joint Appendix, Utah v. Strieff, 136 S. Ct. 2056 (2016) (No. 14-1373), 2015 WL 8146388.
195. See id. at *29 (“The issue at this point is going to rest on attenuation.”).
196. Id.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
592 VANDERBILT LAW REVIEW [Vol. 72:2:561
Fackrell—and a quite limited legal argument consisting of a few
paragraphs of written text.197 Such practice is quite common with
most pretrial suppression hearings in relatively low-level criminal
cases.
This practical reality creates tension with the Supreme Court’s
stated concern about systemic or recurring problems. In traditional
practice, judges might well discourage building a record about
incidents or problems not germane to the case at hand. In fact, one
might speculate that if Strieff’s counsel asked questions about other
times detective Fackrell stopped other people without a warrant, the
lawyer would have been shut down on relevance grounds.198 One
might imagine a judge would have been reluctant to grant discovery
requests for department-wide practices, or internal training materials
about stops or searches, or even detective Fackrell’s own practice. Yet,
as the United States Supreme Court and the Utah Supreme Court
acknowledged, this unconstitutional practice was a normal police
practice in Salt Lake City and would have been ripe for inquiry
because negative exposure presumably would have had a future
deterrent effect.199
In addition to practical reality, such a broadening of the
inquiry to systemic problems requires time and resources. As both
Justice Sotomayor and Justice Ginsberg warned, such a burden
negatively impacts indigent defendants.200 In public defense systems
already underfunded and overwhelmed with cases and starved of
proper investigative resources or expert funding, any requirement
that defense counsel challenge patterns and practices of police
misconduct would ordinarily be unrealistic.201 Without better sources
197. Id.
198. Relevance is defined in terms of whether the evidence has “any tendency to make a fact
more or less probable than it would be without the evidence.” FED. R. EVID. 401. Despite the
term’s broad construct, many judges view evidence about events not related to the defendant in
court as irrelevant.
199. See Strieff, at 2069 (describing running warrant checks as a “ ‘routine procedure’ or
‘common practice’ ” (Sotomayor, J., dissenting) (quoting State v. Topanotes, 2003 UT 30, ¶¶ 2, 76
P.3d 1159, 1160)); see id. at 2073 (Kagan, J., dissenting) (“As Fackrell testified, checking for
outstanding warrants during a stop is the ‘normal’ practice of South Salt Lake City police.”).
200. See supra notes 117, 121 and accompanying text.
201. See generally Mary Sue Backus & Paul Marcus, The Right to Counsel in Criminal
Cases, A National Crisis, 57 HASTINGS L.J. 1031, 1031–36 (2006) (discussing deficiencies in the
right to counsel for poor people in criminal cases); JUST. POL’Y INST., SYSTEM OVERLOAD: THE
COSTS OF UNDER-RESOURCING PUBLIC DEFENSE 6–16 (July 2011), http://www.justicepolicy.org/
uploads/justicepolicy/documents/system_overload_final.pdf [https://perma.cc/YHE6-GV7Z]
(highlighting the ways in which underresourcing hinders adequate public defense); NAT’L RIGHT
TO COUNSEL COMM., JUSTICE DENIED 49–99 (Apr. 2009), http:/www.constitutionproject.org/
pdf/139.pdf [https://perma.cc/W6TN-76XX] (exploring impediments to competent defense
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 593
of police data, one simply cannot expect ordinary, overworked defense
attorneys to conduct full-scale investigations into pattern and practice
problems in local police departments for low-level motions hearings.
Post-Herring there have been a handful of federal cases that
attempted to take seriously the Supreme Court’s interest in systemic
or recurring negligence.202 In United States v. Esquivel-Rios, the Court
of Appeals for the Tenth Circuit decided a case involving an
incomplete (and misleading) license plate database that resulted in
the traffic stop of the defendant.203 Before deciding the case, the
appellate court sent it back to the trial court in order to develop a
factual record on the scope of negligent recordkeeping.204 At issue was
the extent of errors in the computerized database.205 The appellate
court recognized that in order to decide the suppression issue, it
needed to understand the type and magnitude of errors in the
database.206
In other cases, testimony about systemic practices of
misconduct resulted in the suppression of evidence.207 In yet other
cases, the lack of evidence of recurring violations allowed the court to
avoid suppression.208 But the reported cases have thus far been rather
services, such as insufficient funding, excessive caseloads, and lack of performance standards,
training, and oversight).
202. Claire Angelique Nolasco et al., What Herring Hath Wrought: An Analysis of Post-
Herring Cases in the Federal Courts, 38 AM. J. CRIM. L. 221, 233–36 (2011).
203. 786 F.3d 1299, 1301–03 (10th Cir. 2015), cert. denied, 136 S. Ct. 280 (2015).
204. Id. at 1301. (“[W]e concluded that the record lacked the quantity and quality of
information necessary for us to determine whether Mr. Esquivel–Rios’s Fourth Amendment
rights had been violated. We remanded to allow the district court to reconsider its Fourth
Amendment ruling in light of our discussion.” (citation omitted)).
205. Id. at 1301–03.
206. Id. at 1306 (“Whether the rule applies in any given case, however, is context-dependent.
In other words, ‘suppression is not an automatic consequence of a Fourth Amendment
violation.’ ” (quoting Herring v. United States, 555 U.S. 135, 139 (2009))).
207. See United States v. Edwards, 666 F.3d 877, 886 (4th Cir. 2011) (“[T]he circumstances
under which Edwards was searched are likely to recur. Indeed, the evidence in this case showed
that Baltimore City police officers conduct searches inside the underwear of about 50 percent of
arrestees, in the same general manner as the strip search performed on Edwards.”); see also id.
at 886 n.7:
Detective Bailey testified on cross-examination at the suppression hearing, in
pertinent part, as follows: “Question: So is it customary for Baltimore City police
officers to search the underwear area or the dip areas of people that are arrested?
Detective Bailey: I would say it’s about 50 percent of the time, because nobody likes to
do that search. . . . But if you have reason to believe that there might be something,
then it’s a good idea to check, because often they do hide things down there.”
208. United States v. Davis, 690 F.3d 226, 256 (4th Cir. 2012):
We have no proof before us showing that victims’ DNA profiles or individuals cleared
of suspicion in an investigation are routinely entered into the local database
by . . . [Prince George’s County Police Department], or have been entered into the
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
594 VANDERBILT LAW REVIEW [Vol. 72:2:561
few in number, and the practical and legal barriers have proven
prohibitive to litigation.209
Therefore, in the absence of a full federal investigation and in
recognition of the practical realities of trial practice, a new solution to
expose systemic misconduct must be conceived. Fortuitously, a
solution potentially exists in the form of new surveillance technologies
developed to police the citizenry. These big data technologies involving
monitoring, predictive analytics, and data mining offer new ways to
visualize and prove systemic and recurring problems of policing. This
is the subject of the next Part.
II. BLUE DATA: INVERTING THE ARCHITECTURE OF BIG DATA
SURVEILLANCE
In choosing the language “systemic or recurring negligence,”
the Supreme Court invited defendants to prove a certain type of
policing problem.210 To know whether police are negligent, one needs
data on policing practices at both a systemic and individual-officer
level. This is the promise of “blue data”—quantified information of
actual police practice in searchable, sortable, and usable formats.
To envision the potential of blue data, one needs to understand
the existing big data policing capacities being developed in major U.S.
cities. These technologies will fundamentally reshape criminal
investigation by using a combination of data mining, social network
analysis, and video, audio, sensory, and predictive analytics to
identify, track, and monitor citizens living in certain neighborhoods.211
Currently adopted in piecemeal fashion in different cities, the
technologies exist, have proven effective, and will likely expand in
sophistication, integration, and reach.
This Part examines how digital surveillance technologies have
been used to track those suspected of criminal activities; how these
same technologies could be used to address the accountability
database in any other instance. There is nothing in the record to suggest that the acts
here are likely to reoccur.;
United States v. Campbell, 603 F.3d 1218, 1235 (10th Cir. 2010) (“Defendant has demonstrated
at most a single instance of an arguably negligent breakdown in communication among the
WPD. He has not demonstrated what the Supreme Court appears to have indicated is required—
‘recurring or systemic negligence.’ ” (quoting Herring, 555 U.S. at 144)).
209. As discussed in Section I.C, these barriers involve caseload, cost, and procedural rules
that limit the development of a factual record to show systemic misconduct. See supra notes 194–
201.
210. Herring, 555 U.S. at 144.
211. See FERGUSON, supra note 1, at 4.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 595
problems of police violence, racial bias, and unconstitutional practices
raised in Part I; and how proof of those problems could be introduced
in suppression hearings to show systemic or recurring police
negligence. While not comprehensive to the vast array of new
technologies being developed, this Part looks at three widely adopted
new policing tools: (1) data-mining technologies, (2) monitoring
technologies, and (3) predictive technologies.
A. Data-Mining Technologies
The revolution commonly known as “big data computing”
involves new capabilities to collect, store, and sort through vast
quantities of data using sophisticated analytical and machine-learning
tools.212 The quantity of data being created defies comprehension—the
only way this amount of information can become practically useful is
because computing power and analytics have matched its growth.213
Within this expanding data stream, data mining offers new ways to
search. In broad terms, data mining offers the ability to target
particular items of information and visualize patterns of both expected
and unexpected insights.214
As to targeting, data mining allows researchers (or
investigators) to locate a particular data point out of an overwhelming
amount of information. For example, only digital automation and
search capabilities could help the FBI match a suspect using facial
recognition technology from a collection of fifty million mugshots.215
Without the ability to quickly sort through images, the number of
photographs would overwhelm traditional, human-matching
capabilities.
212. VIKTOR MAYER-SCHÖNBERGER & KENNETH CUKIER, BIG DATA: A REVOLUTION THAT
WILL TRANSFORM HOW WE LIVE, WORK, AND THINK 2 (2013); Kenneth Cukier, Data, Data
Everywhere, ECONOMIST (Feb. 25, 2010), http://www.economist.com/node/15557443
[https://perma.cc/SU2U-PPV8]; Steve Lohr, How Big Data Became So Big, N.Y. TIMES (Aug. 11,
2012), http://www.nytimes.com/2012/08/12/business/how-big-data-became-so-big-unboxed.html
[https://perma.cc/NVX8-YM84].
213. See EXEC. OFF. OF THE PRESIDENT, BIG DATA: A REPORT ON ALGORITHMIC SYSTEMS,
OPPORTUNITY, AND CIVIL RIGHTS 5–10 (May 2016), https://obamawhitehouse.archives.gov/sites/
default/files/microsites/ostp/2016_0504_data_discrimination.pdf [https://perma.cc/H3BJ-PXS6]
(introducing the concept of big data and its ramifications).
214. See generally Zarsky, supra note 3, at 287 (discussing data mining’s role in “clos[ing] the
intelligence gap constantly deepening between governments and their new targets”); Note, Data
Mining, Dog Sniffs, and the Fourth Amendment, 128 HARV. L. REV. 691, 693–94 (2014) (defining
data mining as the process by which people or algorithms examine data for patterns of useful
information).
215. U.S. GOV’T ACCOUNTABILITY OFF., GAO-16-267, FACE RECOGNITION TECHNOLOGY: FBI
SHOULD BETTER ENSURE PRIVACY AND ACCURACY 10 (Aug. 3, 2016), https://www.gao.gov/
assets/680/677098.pdf [https://perma.cc/W6E6-4S3G]; PERPETUAL LINE-UP, supra note 22.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
596 VANDERBILT LAW REVIEW [Vol. 72:2:561
As to pattern recognition, algorithms can be created to find
suspected criminal activity. For example, most credit card fraud
warnings and insider trading tips arise because investigators have
programmed computer algorithms to identify unusual patterns of
behavior that correlate with criminal activity.216 But sometimes big
data searches can uncover entirely unexpected correlations.217 In the
criminal justice space, for example, investigators in Richmond,
Virginia, found that certain burglaries were more predictive of sexual
assaults than were prior sexual assaults218 and that sex trafficking
could be discovered by looking at unusual credit card transactions in
nail salon operations.219 Whether used to predict consumer or criminal
activities, the same technologies could be used to crunch, and thus
comprehend, the accumulated big data.
1. Mining Criminal Clues
Law enforcement routinely mines databases for investigatory
purposes.220 Large criminal justice databases include criminal
histories and identifying data for millions of people.221 Biometric
databases with DNA samples, fingerprints, palm prints, photographs,
and even iris scans allow police to identify suspects with relative
216. Philip K. Chan et al., Distributed Data Mining in Credit Card Fraud Detection, 14 IEEE
INTELLIGENT SYS. 67 (1999), http://cs.fit.edu/~pkc/papers/ieee-is99.pdf [https://perma.cc/2ULD-
N3JU]; Peter P. Swire, Privacy and Information Sharing in the War on Terrorism, 51 VILL. L.
REV. 951, 964 (2006).
217. In the consumer space, Walmart discovered that impending hurricanes result in an
uptick of purchases of strawberry Poptarts and Target learned to predict pregnant women from a
combination of common household purchases. Cathy O’Neil, WEAPONS OF MATH DESTRUCTION:
HOW BIG DATA INCREASES INEQUALITY AND THREATENS DEMOCRACY 98 (2016); Charles Duhigg,
How Companies Learn Your Secrets, N.Y. TIMES MAG. (Feb. 16, 2012), https://www.nytimes.com/
2012/02/19/magazine/shopping-habits.html [https://perma.cc/T2LK-UX89]; Constance L. Hayes,
What Wal-Mart Knows About Customers’ Habits, N.Y. TIMES (Nov. 14, 2004),
https://www.nytimes.com/2004/11/14/business/yourmoney/what-walmart-knows-about-customers
-habits.html [https://perma.cc/J3QC-U7RW].
218. Colleen McCue & Andre Parker, Connecting the Dots: Data Mining and Predictive
Analytics in Law Enforcement and Intelligence Analysis, 10 POLICE CHIEF 115, 122 (2003).
219. Tierney Sneed, How Big Data Battles Human Trafficking, U.S. NEWS (Jan. 14, 2015),
https://www.usnews.com/news/articles/2015/01/14/how-big-data-is-being-used-in-the-fight-
against-human-trafficking [https://perma.cc/NB3A-PSG9].
220. See Daniel J. Steinbock, Data Matching, Data Mining, and Due Process, 40 GA. L. REV.
1, 4 (2005) (“Data mining’s computerized sifting of personal characteristics and behaviors
(sometimes called ‘pattern matching’) is a more thorough, regular, and extensive version of
criminal profiling, which has become both more widespread and more controversial in recent
years.”).
221. See generally Logan & Ferguson, supra note 84, at 541 (discussing the growing scope of
criminal justice data collection).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 597
ease.222 Court records, gang and sex offender registries, and other
digital collections of criminal justice information provide new ways to
investigate and monitor individuals.223 As traditionally used, these
databases can be searched for a particular piece of information. If
police need to match a fingerprint found at a crime scene, they can
compare the lifted and digitized print with millions of similar
prints.224 If police need to run a name through the National Criminal
Information Center to find an address, they have access to millions of
addresses with a single query.225
Big data policing rests on the idea of creating databases out of
the almost incalculable number of variables that determine the
“where,” “when,” and “what” of criminal activity. In the past few
decades, crime analysts have been plotting and mapping these crime
patterns.226 Particular hotspots can be identified by address.227 Crime
patterns can be identified by neighborhood or block. In doing so,
recurring problem areas can be visualized and environmental factors
222. See Laura K. Donohue, Technological Leap, Statutory Gap, and Constitutional Abyss:
Remote Biometric Identification Comes of Age, 97 MINN. L. REV. 407, 415 (2012) (“[Remote
Biometric Identification] technologies present capabilities significantly different from that which
the government has held at any point in U.S. history.”); Logan, supra note 15, at 1575 n.91
(“ ‘Biometrics’ refers either to biological or physiological characteristics usable for automatic
recognition of individuals on the basis of such characteristics.” (citing NAT’L SCI. & TECH.
COUNCIL, PRIVACY & BIOMETRICS: BUILDING A CONCEPTUAL FOUNDATION 4 (Apr. 14, 2006),
http://www.biometrics.gov/Documents/privacy.pdf [https://perma.cc/W9QH-JYGH])); Daniel J.
Steinbock, National Identity Cards: Fourth and Fifth Amendment Issues, 56 FLA. L. REV. 697,
704 (2004) (“Biometrics are identification techniques based on some unique, physiological, and
difficult-to-alienate characteristic.”).
223. WAYNE A. LOGAN, KNOWLEDGE AS POWER CRIMINAL REGISTRATION AND COMMUNITY
NOTIFICATION LAWS IN AMERICA 22–30 (2009) (discussing local criminal registration laws).
224. Erin Murphy, Databases, Doctrine & Constitutional Criminal Procedure, 37 FORDHAM
URB. L.J. 803, 806–08 (2010).
225. David M. Bierie, National Public Registry of Active-Warrants: A Policy Proposal, 79
FED. PROB. 27, 28 (2015) (“[NCIC] is the central transactional data system that tracks the
nation’s warrants. All police agencies can enter their warrants in the system and check the
system to identify whether a given individual has a warrant.”).
226. See generally Andrew Guthrie Ferguson, Crime Mapping and the Fourth Amendment:
Redrawing “High-Crime Areas,” 63 HASTINGS L.J. 179, 184–86 (2011) (discussing the history of
crime mapping programs).
227. James G. Cameron, Spatial Analysis Tools for Identifying Hotspots, in MAPPING CRIME:
UNDERSTANDING HOT SPOTS 35, 35 (John E. Eck et al. eds., 2005):
A central concern of hot spot analyses of crime is assessing the degree of spatial
randomness observed in the data. Most of the available tools provide different ways of
determining whether the underlying pattern is uniform over space or whether
significant clusters or other spatial patterns exist, which are not compatible with
spatial randomness.;
John E. Eck, Crime Hot Spots: What They Are, Why We Have Them, and How to Map Them, in
MAPPING CRIME, supra, at 1, 4 (“The most basic form of a hot spot is a place that has many
crimes. A place can be an address, street corner, store, house, or any other small location, most of
which can be seen by a person standing at its center.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
598 VANDERBILT LAW REVIEW [Vol. 72:2:561
identified.228 As computer power grew, and as data collection
expanded, so did the capabilities of crime analysts and the
development of data-crunching systems to help target and predict
criminal activity in particular places.229 Now crime maps include not
only the place, time, and type of crime but also a whole host of other
environmental factors that might increase the risk of future crime.230
Out of the seeming chaos of individual crimes, a pattern of activity can
be visualized and addressed.
In addition to seeing where crime is occurring, a few police
organizations are proactively trying to create large-scale social
network analysis datasets that can be queried for clues or
investigatory leads about who is committing those crimes.231 A good
example of this data creation is found in the Los Angeles Police
Department (“LAPD”). The LAPD has partnered with a private
company—Palantir—to begin collecting data about chronic offenders
in the city.232 This collection process involves three distinct steps.
First, police identify particular chronic offenders who are suspected to
be involved in recurring criminal activity.233 Second, the police
proactively contact these offenders in an effort to collect personal data
about them.234 These contacts, recorded on “field interview cards,”
228. Anthony A. Braga et al., The Relevance of Micro Places to Citywide Robbery Trends: A
Longitudinal Analysis of Robbery Incidents at Street Corners and Block Faces in Boston, 48 J.
RES. CRIME & DELINQ. 7, 11 (2011) (“Studies of the spatial distribution of robbery in urban
environments have also revealed that a small number of micro places generate a
disproportionate number of robberies. Certain high-risk facilities, such as bars, convenience
stores, and banks, at particular places also tend to experience a disproportionate amount of
robbery . . . .”); Lisa Tompson & Michael Townsley, (Looking) Back to the Future: Using Space-
Time Patterns to Better Predict the Location of Street Crime, 12 INT’L J. POLICE SCI. & MGMT. 23,
24 (2010) (“Research has repeatedly demonstrated that offenders prefer to return to a location
associated with a high chance of success instead of choosing random targets.”).
229. See Andrew Guthrie Ferguson, Policing Predictive Policing, 94 WASH. U. L. REV. 1109,
1126–32 (2017) (“Predictive Policing . . . involved the collection of historical crime data (time,
place, and type) and the application of an experimental computer algorithm that used data to
predict likely areas of criminal activity.”).
230. Id.
231. CITY OF NEW ORLEANS, NOLA FOR LIFE: COMPREHENSIVE MURDER REDUCTION
STRATEGY (Apr. 2016), http://nolaforlife.org/files/n4l-2016-comprehensive-murder-reduction-
strategy-b/ [https://perma.cc/UU4N-Q6NM]; Jason Sheuh, New Orleans Cuts Murder Rate Using
Data Analytics, GOVTECH.COM (Oct. 22, 2014), http://www.govtech.com/data/New-Orleans-Cuts-
Murder-Rate-Using-Data-Analytics.html [https://perma.cc/9DBJ-FV9T] (explaining how the city
of New Orleans monitors criminal social networks in an effort to reduce crime).
232. Sarah Brayne, Big Data Surveillance: The Case of Policing, 82 AM. SOC. REV. 977, 987
(2017); Matt McFarland, A Rare Look Inside LAPD’s Use of Data, CNN MONEY (Sept. 11, 2017),
https://money.cnn.com/2017/09/11/technology/future/lapd-big-data-palantir [https://perma.cc/
B8QN-W7PA].
233. See Brayne, supra note 232, at 987 (“[Chronic offender field identification cards] are key
intelligence tools for law enforcement and were one of the first data sources integrated into
Palantir.”).
234. Id.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 599
involve collecting information about where they were spotted, who
they were with, the type of car they were driving, and any other
identifiable information about the target or his associates.235 Third,
police load this data into a large central database so that each field
can be queried.236 As a result, if police want to search for the
whereabouts of a particular gang member, they are able to access
information related to where he has been spotted and who he has been
spotted with over the last year and may then develop a network of
social and geolocational connections. Each variable can be connected
to any other variable in the database.
These data points then serve as the building blocks for a more
ambitious networked system. The inputted data is combined with
other crime data, some public government data, and even some
consumer data to create a searchable network of criminal offenders.237
Again, each variable can be separately sorted and ordered. If police
want to track a phone number or address across different groups of
people, they can find a connecting phone number.238 If police want to
track all visitors to a suspected house, they can geofence the area to
identify any car that drives through.239 All of the data is inputted and
connected through network analysis.240 The computer model for this
social network technology evolved from developments tracking
international terrorists who needed to be linked, monitored, and
watched across different jurisdictions.241
In terms of search capabilities, police can target a particular
person (or phone number, or license plate) for investigation. For
example, a partial license plate number, a partial description, and a
235. Id.
236. Id. at 992–93.
237. See id. at 993–96 (explaining how Palantir located a public database on foreclosures and
added the public information to its system).
238. See id. at 994 (describing how the police acquire and connect people to various phone
numbers).
239. Chris Hackett & Michael Grosinger, The Growth of Geofence Tools Within the Mapping
Technology Sphere, PDVWIRELESS (Dec. 15, 2014), https://www.pdvwireless.com/the-growth-of-
geofence-tools-within-the-mapping-technology-sphere [https://perma.cc/P3VA-K29M].
240. Jenna McLaughlin, L.A. Activists Want to Bring Surveillance Conversation Down to
Earth, INTERCEPT (Apr. 6, 2016, 8:22 AM), https://theintercept.com/2016/04/06/l-a-activists-want-
to-bring-surveillance-conversation-down-to-earth [https://perma.cc/252Z-94A8].
241. Mark Harris, How Peter Thiel’s Secretive Data Company Pushed into Policing, WIRED
(Aug. 9, 2017), https://www.wired.com/story/how-peter-thiels-secretive-data-company-pushed-
into-policing [https://perma.cc/B686-QS2E]; Peter Waldman, Lizette Chapman & Jordan
Robertson, Palantir Knows Everything About You, BLOOMBERG (Apr. 19, 2008),
https://www.bloomberg.com/features/2018-palantir-peter-thiel [https://perma.cc/7B4M-NPDZ];
see also Palantir, Palantir at the Los Angeles Police Department, YOUTUBE (Jan. 25, 2013),
http://www.youtube.com/watch?v=aJ-u7yDwC6g [https://perma.cc/P8WB-BSLC].
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
600 VANDERBILT LAW REVIEW [Vol. 72:2:561
tattoo can transform into an actual human target by querying the
database.
In terms of patterns, the same network analysis can be used to
track gang members or others involved in large-scale criminal activity.
In Chicago, for example, network analysis found that most homicides
involve rival gangs of young men.242 The circles of retaliatory killings
can be studied using social network analysis.243 Police can develop a
database that tracks where, why, and with whom their targets are
associating. Police can see crime not only as a series of individual acts
but as part of a larger pattern of relationships and connections. As
might be expected from the name, “social network analysis” reveals
hidden connections that otherwise would not be identified.
As a technical matter, the innovation for policing is the ability
to break down ordinary life into discrete and searchable variables.
Repeated problem actors can be identified. Recurring crimes can be
linked with associated groups. The general point is that these types of
technologies allow data to be queried in unusual ways to find new
insights to identify and study crime patterns. Whatever the subject
matter of the database, the technology allows for new mechanisms to
manipulate and study the data. While hard questions remain about
the cost of these systems, the interoperability of linking different
datasets and the willingness of police to embrace a data-driven
strategy combine to offer new ways to reduce crime.
2. Mining Policing Data
Police officers generate data during every single shift. Location,
contacts, actions, observations, and arrests are all data points.244 For
242. TRACEY MEARES, ANDREW V. PAPACHRISTOS & JEFFREY FAGAN, HOMICIDE AND GUN
VIOLENCE IN CHICAGO: EVALUATION AND SUMMARY OF THE PROJECT SAFE NEIGHBORHOODS
PROGRAM (2009), https://www.flintridge.org/newsresources/documents/HomicideandGun
ViolenceinChicago-EvaluationandSummaryoftheProjectSafeNeighborhoodsProgram-2009.pdf
[https://perma.cc/TMM5-8HFB]; see also David M. Kennedy, Pulling Levers: Chronic Offenders,
High-Crime Settings, and a Theory of Prevention, 31 VAL. U. L. REV. 449, 459 (1997) (“Finally,
much crime–violent, drug, property, and domestic–is concentrated in certain
neighborhoods . . . .”); Tracey Meares, Andrew V. Papachristos & Jeffrey Fagan, Attention Felons:
Evaluating Project Safe Neighborhoods in Chicago, 4 J. EMPIRICAL LEGAL STUD. 223 (2007);
Andrew V. Papachristos et al., Social Networks and the Risk of Gunshot Injury, 89 J. URB.
HEALTH 992, 993 (2012) [hereinafter Papachristos et al., Social Networks]; Andrew V.
Papachristos, Commentary: CPD’s Crucial Choice: Treat Its List As Offenders or as Potential
Victims?, CHI. TRIB. (July 29, 2016, 10:00 AM), http://www.chicagotribune.com/
news/opinion/commentary/ct-gun-violence-list-chicago-police-murder-perspec-0801-jm-20160729-
story.html [https://perma.cc/SR7C-852Q].
243. See sources cited supra note 242.
244. Amy Feldman, How Mark43’s Scott Crouch, 25, Built Software to Help Police
Departments Keep Cops on the Street, FORBES (Oct. 19, 2016, 10:00 AM), https://www.forbes.com/
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 601
supervisors or communities concerned with police accountability, this
data is incredibly valuable. Do police stop more people of color? Do
police stop people more in particular neighborhoods? Is police
suspicion justified by successful outcomes (e.g., recovering weapons or
contraband)? With data, police are able to determine if there are
particular persons or patterns that raise concerns.
As with collected crime data, police could mine data they
routinely collect to identify recurring violations of the Constitution
and study systemic violations through social network analysis. One
example of data-mining police practices arises from the NYPD stop
and frisk litigation. As discussed previously, this data analysis
ultimately led to the stop and frisk practice being declared
unconstitutional.245 But it also inspired a team of researchers led by
Professors Sharad Goel, Ravi Shroff, and David Sklansky to examine
the data to see if they could predict which types of stop and frisks
would more likely result in the recovery of contraband.246 As they
explained, data could predict the likelihood that a stop and frisk would
uncover contraband or other evidence based on the officer’s prior
knowledge, such as time, location, characteristics of the suspect, and
the suspicious circumstance at hand.247 The researchers called this
prediction a “stop-level hit rate,” which can be operationalized to
predict the probability of recovering a weapon.248 According to analysis
from the actual NYPD data, “43% of the Terry stops carried out by the
NYPD based on suspicion of CPW [criminal possession of a weapon]
sites/amyfeldman/2016/10/19/how-mark43s-scott-crouch-25-built-software-to-help-police-do-
their-jobs-better [https://perma.cc/99AN-72PF].
245. See Floyd v. City of New York, 959 F. Supp. 2d 540, 562 (S.D.N.Y. 2013) (“In conclusion,
I find that the City is liable for violating plaintiffs’ Fourth and Fourteenth Amendment rights.”);
Declaration of Jeffrey Fagan, Ph.D. at 2, Floyd, 959 F. Supp. 2d 540 (No. 08 Civ. 01034(SAS)),
2011 WL 7552634; see also CTR. FOR CONST. RTS., STOP AND FRISK: THE HUMAN IMPACT (2012),
https://ccrjustice.org/sites/default/files/attach/2015/08/the-human-impact-report.pdf
[https://perma.cc/8RRX-P9AN]; CTR. FOR CONST. RTS., RACIAL DISPARITY IN NYPD STOPS-AND-
FRISKS 1, 10, 15 (Jan. 15, 2009), https://ccrjustice.org/sites/default/files/assets/Report-CCR-
NYPD-Stop-and-Frisk_3.pdf [https://perma.cc/N8NC-2ZUV].
246. Sharad Goel et al., Combatting Police Discrimination in the Age of Big Data, 20 NEW
CRIM. L. REV. 181 (2017).
247. Id. at 187 (“The data can be used to compute the likelihood that any particular stop-
and-frisk will result, for example, in the discovery of particular kinds of evidence . . . .”).
248. Id. at 187–88:
[T]his information is recorded in what the NYPD calls a “UF-250” report, and it can be
used to estimate a “stop-level hit rate”—the ex ante probability of discovering a
weapon, based on all the factors that were known to the officer before the Terry stop.
The stop-level hit rate, or “SHR,” can be thought of as a measure of the strength of the
evidence supporting the suspicion that the individual to be stopped and frisked has a
gun.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
602 VANDERBILT LAW REVIEW [Vol. 72:2:561
had less than a 1% chance of actually resulting in the discovery of a
weapon.”249
The researchers came to this determination through
sophisticated data analysis of the information already collected by the
NYPD. Essentially, researchers collected data from NYPD UF-250
forms, only focusing on information that would have been available to
officers when making a decision about whether to stop or not.250 They
then built a computer model that incorporated numerous variables,
including:
[D]emographic information about the suspect (sex, race, age, height, weight, and build);
location of the stop (precinct; inside or outside; and on public transit, in public housing,
or neither), date and time of the stop (year, month, day of week, and time of day); the
recorded reasons for the stop (e.g., “furtive movements” or “high crime area”); whether
the stop was the result of a radio run; whether the officer was in uniform; how long the
officer observed the suspect before initiating the stop; and the “local hit rate” of stops at
that location.251
Then, utilizing this model, researchers examined which of the 472,344
stops from 2008 through 2010 recovered a weapon.252 By considering
recovery of a weapon a successful stop, researchers were able to isolate
the variables that might contribute to successful stops and those
variables that likely do not. The model included 7,705 predictive
features.253 This model was then applied to stops for 2011 and 2012,
under the reasoning that if the predictive model could isolate those
variables that mattered to effective stops in 2008 through 2010, then
the researchers should be able to predict the outcome for stops in 2011
and 2012.254
The results were impressively accurate. The model predicted
eighty-three percent of successful stops.255 Equally helpful, the model
could predict which types of stops would not be successful. For
example, a stop-hit rate analysis showed some standard police
249. Id. at 187.
250. Id. at 211–12. UF-250 forms are NYPD documents that record the type of stop, the
justifications for the stop, and the time and place of the stop.
251. Id. at 212.
252. Id. at 211–12.
253. Id. at 212.
254. Id.
255. Id. at 212–13:
The results produced by the SHR method are dramatic. First, the model turns out to
be highly accurate. To evaluate the model, we selected random pairs of cases from
among the 2011 and 2012 stops where a weapon was ultimately found in exactly one
stop of the pair. Presented only with the stop-level predictors (and not the outcomes),
a completely uninformative model would do no better than chance at determining in
which one of the two stops a weapon was found. In contrast, we found that our SHR
model correctly picked out the stop with the weapon 83 percent of the time, indicating
good predictive performance.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 603
justifications for a stop, such as “the suspect made furtive
movements,” did not correlate with the recovery of a weapon.256 The
ability to predict unsuccessful stops enables the model to assist in
determining how stop and frisk practices could be redirected to be
more effective257 and less discriminatory.258
Three conclusions arise from this experiment in data-mining
stop and frisk statistics. First, the stop-hit rate could be used to show
that whole categories of police practice—that is, systemic patterns—
were both ineffective and racially discriminatory.259 Second, the stop-
hit rate could be used retrospectively to determine in particular cases
whether the officer did in fact have reasonable suspicion to conduct
the stop and thus empirically support a court’s Fourth Amendment
conclusion.260 Finally, the stop-hit rate could be used to create a
predictive tool to assist officers in deciding whether to stop a
suspect.261 All of these possibilities are now recognized because
scholars saw the value in studying blue data.
As another example of the potential to quantify and mine police
practices, Stanford University Professor Jennifer Eberhardt led a
team of researchers in a two-year data-driven study of the Oakland
256. Id. at 188 (“SHR analysis reveals that some of the standard justifications for pedestrian
stops that the UF-250 has employed—‘furtive movements,’ for example—are unhelpful in
identifying suspects who actually have weapons; avoiding the use of those factors would make
stops less discriminatory and more successful.” (citation omitted)).
257. Id. (“The SHRs can thus provide a road map for redirecting stop-and-frisk practices to
make them, simultaneously, less racially lopsided in their impact and more effective at finding
what the police say they are looking for.”).
258. Id. (“But the SHR method does more than that. It pinpoints particular categories of
Terry stops for CPW that both (a) are relatively unlikely to actually find a weapon, and (b)
impose an especially disproportionate burden on racial minorities.”).
259. Id. (“And these low-odds stops had a heavy racial tilt: 49 percent of the stops of blacks
fell below the 1 percent probability threshold, as did 34 percent of the stops of Hispanics,
compared with only 19 percent of the stops of whites.”); see also id. at 215 (“Third, the SHR
method provides strong, numerical support for the conclusion reached in Floyd: that the stop-
and-frisk practices of the NYPD discriminated against racial minorities, particularly blacks.”).
260. Id. at 217:
Fourth, the SHR method not only allows one to estimate the aggregate number of
stops that fall below a specified probability threshold, but also yields a quantitative
measure of the evidence supporting a stop-and-frisk in each particular case, which can
in turn be used to determine whether “reasonable articulable suspicion” existed.
261. Id. at 188 (“More ambitiously, SHR analysis could be used to craft a simple heuristic for
officers to use on the street to determine which suspects to stop and frisk, drastically reducing
the disparate impact and increasing the ‘efficiency’ of the searches.”); see also id. at 218:
Finally—and more speculatively—SHR analysis can be used not just to assess an
officer’s decision to conduct a Terry stop after the fact, but also to guide that decision
in the first place. Because the SHR is calculated from information available to the
officer at the time the decision is made to carry out a stop-and-frisk, the method also
could be used, in theory, to inform the stop decision.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
604 VANDERBILT LAW REVIEW [Vol. 72:2:561
Police Department.262 Initiated by a federal court order, the project
sought to examine whether racial bias impacted policing in
Oakland.263 The methodology involved an intensive data dive into the
records of the police department, examining stops, arrests, use of
handcuffs, narrative scripts in police reports, and the language used
in police stops (obtained from body camera footage).264 In many cases,
this data is similar to the data described in Professor Sarah Brayne’s
study, but with the focus inverted from tracking civilians to tracking
police officers.265 The data showed that Oakland police treated people
of different races in different and seemingly discriminatory ways.266
For example, in studying police contacts, the researchers
examined 28,119 self-initiated police stops over a thirteen-month
period.267 Each of these stops generated a Field Interview/Stop Data
Report (“FI/SDR”) which could be broken down into different data
fields, including “encounter variables,” “officer variables,” and “census
track variables.”268 Encounter variables included the reason for the
encounter; the justification for the stop (reasonable suspicion,
probable cause, traffic violation, probation or parole status, or
consensual encounter); the time, date, and day of the week that the
stop occurred; the type of stop (vehicle, pedestrian, or other); and the
policing area in which the stop occurred, as well as the gender, age,
262. REBECCA C. HETEY ET AL., DATA FOR CHANGE: A STATISTICAL ANALYSIS OF POLICE,
STOPS, SEARCHES, HANDCUFFINGS, AND ARRESTS IN OAKLAND, CALIF., 2013-2014 (June 23, 2016),
https://stacks.stanford.edu/file/druid:by412gh2838/Data%20for%20Change%20%28June%2023%
29.pdf [https://perma.cc/8MSM-5Z4B].
263. Id. at 11 (“In May 2014, the City of Oakland contracted with our team of Stanford
University researchers to assist the Oakland Police Department (OPD) in complying with a
federal order to collect and analyze data on OPD officers’ self-initiated stops of pedestrians and
vehicles by race.” (citation omitted)).
264. Id. at 12–26 (providing an overview of the data that the study analyzed); see also id. at 9
(referencing the researchers’ “[d]evelopment of computational tools to analyze linguistic data
from body-worn cameras”).
265. See supra notes 232–237 and accompanying text (describing the Los Angeles Police
Department’s collection of data on particular chronic offenders).
266. HETEY, supra note 262, at 9:
Across our research programs, we indeed uncovered evidence that OPD officers treat
people of different races differently. At the same time, we found little evidence that
this disparate treatment arose from overt bias or purposeful discrimination. Instead,
our research suggests that many subtle and unexamined cultural norms, beliefs, and
practices sustain disparate treatment.
267. Id. at 12 (“During this 13-month time period, 28,119 stops were recorded by 510 sworn
OPD officers.”); id. at 16 (“Members of the OPD are required to complete a stop data form for all
self-initiated encounters that involve one or more persons subject to detention, arrest, search, or
request to search.” (citation omitted)).
268. Id. at 11 (“Our task was to analyze the reports that OPD officers completed after every
stop they initiated between April 1, 2013, and April 30, 2014. These reports are called Field
Interview/Stop Data Reports (FI/SDR), and the information they contain is called stop data.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 605
and race of the suspect.269 The system lists officer variables, including
the officers’ race, gender, age, experience level, type of assignment,
squad, and a link to the individual employee ID.270 Census tract
variables included information about the area of the stop, including
the address where the stop occurred; the area’s crime rate, population
demographics, total population, population density, and land area; the
age and race demographics of the population; and the percentage of
the population living in poverty in the area, among other
socioeconomic variables.271
Using this data, researchers were able to study patterns of
police stops and control for other potential influences. Researchers
found that police officers stopped more African Americans than
Whites, even after controlling “for neighborhood crime rates and
demographics; officer race, gender, and experience; and other factors
that shape police actions.”272 Despite African Americans only making
up twenty-eight percent of the population, they were stopped sixty
percent of the time, nearly three times more than Hispanics, who
made up the next most common racial group.273
The same racial disparity could be observed in how police
treated suspects after they had been stopped. For example, the data
showed that African American men were handcuffed in one out of
every four stops, compared to one out of every fifteen stops for White
men.274 Again, even controlling for neighborhood crime rate, African
Americans were more likely to be placed in handcuffs.275 Similarly,
African American men were searched in one out of five stops compared
to one out of twenty stops of White men.276 Again, even controlling for
crime rate and the racial makeup of neighborhood, African Americans
were searched more with no increase in recovered contraband.277
269. Id. at 49–51.
270. Id. at 52–53.
271. Id. at 54–60.
272. Id. at 10.
273. Id. at 14 (“African Americans were the racial group most often stopped . . . . Sixty
percent of stops, or nearly 17,000 stops, were of African Americans. Stops of African Americans
were made at a rate of more than three times that of the next most common group, Hispanics.”).
274. Id. at 90 (“Excluding arrests, African American men were handcuffed in 1 out of every 4
stops vs. 1 in every 15 stops for White men.”).
275. Id. (“Even controlling for multiple covariates like neighborhood crime rate, African
Americans were still significantly more likely to be handcuffed (excluding arrests) than Whites
in 4 out of 5 of Oakland’s policing areas.”).
276. Id. at 109 (“Excluding incident to arrest, inventory, and probation/parole searches,
Black men were searched in 1 out of 5 stops, vs. 1 out of 20 stops for White men.”).
277. Id.:
Even after controlling for a host of factors, including the crime rate and the racial
demographics of the neighborhood where the stop was made, African Americans were
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
606 VANDERBILT LAW REVIEW [Vol. 72:2:561
Finally, African Americans were arrested more than one out of every
six stops versus arrests for only one out of every fourteen White men
stopped, with the arrest discrepancy most obvious in arrests for traffic
violations.278
At a more granular level, researchers found that female officers
made fewer stops,279 as did more experienced officers.280 Similarly,
more senior officers arrested less and used handcuffs less, but
seniority did not have an impact on the number of searches
conducted.281 The analysis also demonstrated that “Asian officers
show[ed] less of an African American–White gap in searches.”282
African American officers, on the other hand, “show[ed] more of an
African American-White gap in arrests.”283
This data-mining approach was also applied to the narratives
of the police reports. As one of the pilot programs, the Stanford
researchers developed a machine-learning technique to sort through
the narratives of the FI/SDR.284 This model could quickly sort through
the different justifications for a traffic stop to see if racial bias could be
detected in the outcome. Again, racial bias was detected: “These
analyses uncovered racial disparities in both type and severity of
stops, with [Oakland Police Department] officers disproportionately
still significantly more likely than Whites to be the subject of such high-discretion
searches in 3 of Oakland’s 5 policing areas. The African American–White race
difference was especially pronounced for vehicle stops, stops made because of traffic
violations, and stops made by officers working special assignments, other than
violence suppression. We found no race differences in search recovery rates.
278. Id. at 140:
Overall, more than 1 in 6 African American men stopped was arrested vs. only 1 in 14
White men stopped. Even when controlling for other variables, African Americans
were still significantly more likely than Whites to be arrested in 2 of Oakland’s 5
policing areas. The African American-White arrest gap was most pronounced for
vehicle stops, stops made because of traffic violations, and stops made by officers
working violence suppression.
279. Id. at 158.
280. Id.
281. Id.
282. Id.
283. Id.
284. SPARQ, STANFORD UNIV., STRATEGIES FOR CHANGE: RESEARCH INITIATIVES AND
RECOMMENDATIONS TO IMPROVE POLICE-COMMUNITY RELATIONS IN OAKLAND, CALIF. 24 (Jennifer
L. Eberhardt ed., 2016), https://stanford.app.box.com/v/Strategies-for-Change [https://perma.cc/
2DGB-3GYG] [hereinafter STRATEGIES FOR CHANGE] (“[W]e then developed advanced natural-
language-processing and machine-learning techniques for coding the narratives in the stop data
forms. Once refined, these techniques will eliminate the need for human coders, and allow the
OPD and other law enforcement agencies to analyze large quantities of narrative data cheaply,
quickly, and reliably.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 607
stopping African Americans for all types of violations, as well as for
very minor violations.”285
The Stanford study focused on racial disparities in policing,
concluding that the Oakland Police Department’s practices produced
racially discriminatory outcomes.286 The conclusion is useful to
demonstrate the potential for data-driven accountability mechanisms
for policing. The study showed that traditional police data, normally
used to monitor criminal activity, could be mined to encourage police
accountability and improve training and oversight.287
For purposes of this Article, the real insight is in the
technological capacity to mine police data. While Oakland was just an
experiment (and one mandated by federal court order), the fact that a
major U.S. city with thousands of police-citizen interactions could
collect, sort, and study its policing patterns to find disparate racial
impacts shows the potential for obtaining other information reflecting
policing patterns. In addition, social network analysis can show which
officers are involved in what types of stops, where the officers are
making those stops, and against whom the stops are made. If one
wanted to query all police stops in a neighborhood, all police stops
against a certain gang, or whether a particular unit caused more
complaints, social network analysis makes that possible. Knowing
who, where, how, and why suspects were stopped opens up new
research avenues to understand the choices police make on a daily
basis. Particular police officers could be targeted for study, and
patterns relating to experience, gender, or other variables could be
examined.
285. Id. at 20:
To enrich our exploration of police-community relations in Oakland, we first
developed a coding scheme to analyze these narratives. We then recruited experts to
use our coding scheme to sort some 1,000 traffic violations from April 2014 by type
(e.g., moving violations vs. equipment violations) and severity (from minor to severe).
286. See HETEY ET AL., supra note 262, at 179:
To be clear, though: our results do not suggest that OPD officers are “racists.” Our
mission is not to point fingers at specific individuals, but to explore an institution’s
effects on its communities, particularly its communities of color. Our exploration
revealed that racial disparities in the OPD’s activities are widespread and systemic.
287. See id.:
These findings are not evidence of a few or even many bad apples, but of pervasive
cultural norms—the unwritten rules of how to behave—about how to police people of
different races. Focusing on individual officers, rather than on the culture as a whole,
will likely allow racial disparities in policing to persist. Put another way, focusing on
the individual officer may let law enforcement agencies, especially their leaders, off
the hook too easily. Instead, to combat racial disparities in the treatment of
community members, law enforce[ment] agencies must challenge the cultural beliefs,
policies, practices, and norms that encourage disparate treatment.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
608 VANDERBILT LAW REVIEW [Vol. 72:2:561
While two experiments in data-rich policing environments
cannot be used to predict the future of blue data, they do show the
potential of data-driven accountability. While limited in scope and
purpose, the ability to mine datasets for insights can be adapted to
other police departments. While still early in its development, new
data-mining techniques can provide new ways to visualize these
constitutionally problematic practices. In fact, as will be discussed in
the next Section, these data-driven insights might provide evidence of
the systemic and recurring problems needed to fulfill the Supreme
Court’s new exclusionary test.
3. Mining Exclusion
Mined data can reveal patterns of racial discrimination and
unconstitutional stops. If the Supreme Court’s new test requires
defendants to show systemic or recurring negligence, imagine how
suppression hearings might play out in Oakland, New York City, or
Baltimore, where systemic problems have been documented. After all,
in a five-year period, Baltimore police stopped over three hundred
thousand people.288 Almost half of the stops took place in two small,
predominantly African-American districts that contained only eleven
percent of the city’s population.289 According to the DOJ, many of
those stops took place in violation of the Fourth Amendment with
stops based on less than reasonable suspicion.290 More than ninety-
four percent of the stops did not result in a citation or an arrest,
meaning no contraband was recovered from suspicionless stops.291
Again, following the Supreme Court’s guidance in Herring, an
unconstitutional stop connected to a systemic or recurring pattern
warrants suppression.
288. DOJ BALTIMORE REPORT, supra note 125, at 5 (“BPD officers recorded over 300,000
pedestrian stops from January 2010–May 2015, and the true number of BPD’s stops during this
period is likely far higher due to under-reporting.”).
289. Id. at 6 (“BPD’s pedestrian stops are concentrated on a small portion of Baltimore
residents. BPD made roughly 44 percent of its stops in two small, predominantly African-
American districts that contain only 11 percent of the City’s population.”).
290. Id. (“BPD’s stops often lack reasonable suspicion. Our review of incident reports and
interviews with officers and community members found that officers regularly approach
individuals standing or walking on City sidewalks to detain and question them and check for
outstanding warrants, despite lacking reasonable suspicion to do so.”).
291. Id. (“Only 3.7 percent of pedestrian stops resulted in officers issuing a citation or
making an arrest. And, as noted below, many of those arrested based upon pedestrian stops had
their charges dismissed upon initial review by either supervisors at BPD’s Central Booking or
local prosecutors.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 609
The Stanford study focused on racial bias, a subject of
relevance, if not a direct relation to Fourth Amendment practice.292
But the same type of data analysis could focus on the justifications for
why Oakland police stopped an individual, as the stop-hit-ratio
researchers did with the NYPD data.293 Once collected, the data could
be queried to show the events that transpired, the justifications for the
stop, and the result of the stop. Following Herring, this data could be
introduced in a Fourth Amendment suppression hearing to effectuate
the exclusionary rule.
Take as an example facts from one of the plaintiffs in the
NYPD Floyd stop and frisk litigation.294 Devin Almonor, a thirteen-
year-old teenager, was stopped, frisked, detained in handcuffs, and
taken to a police station without reasonable suspicion.295 Almonor’s
stop arose from a series of 911 calls reporting a fight in progress with
the potential of armed juveniles in a particular geographic location.296
When police arrived at that location, Almonor and a friend were seen
walking up the street.297 There was no description of the suspects
except that the juveniles were Black youth.298 Police forcefully put
Almonor over the hood of a police car, handcuffed him, searched him,
and eventually took him to the police station. Ultimately, no
contraband was recovered and the case was dismissed.299
But now consider the analysis if marijuana had been recovered
from Almonor and, as with tens of thousands of other narcotics busts,
the case required a Fourth Amendment suppression hearing. In
Almonor’s case, Judge Scheindlin found that the stop violated the
Fourth Amendment—a necessary but, under Herring, not sufficient
292. Whren v. United States, 517 U.S. 806, 813 (1996) (“[T]he Constitution prohibits
selective enforcement of the law based on considerations such as race. But the constitutional
basis for objecting to intentionally discriminatory application of laws is the Equal Protection
Clause, not the Fourth Amendment.”); Milton Heumann & Lance Cassak, Profiles in Justice?
Police Discretion, Symbolic Assailants, and Stereotyping, 53 RUTGERS L. REV. 911, 956 (2001)
(“Under the Court’s reasoning in Whren, race is irrelevant to any issues raised under the Fourth
Amendment.”).
293. See supra notes 246–261 and accompanying text (describing the work of Professors
Goel, Shroff, and Sklansky).
294. Floyd v. City of New York, 959 F. Supp. 2d 540, 628–30 (S.D.N.Y. 2013).
295. Id.
296. Id. at 628.
297. Id.
298. Id.
299. Id. at 628–29. Interestingly and relevant for data-driven policing, despite being
innocent, Almonor’s personal information was entered into the police database accusing him of
being suspected of possessing a weapon and resisting arrest. Id. at 629–30 (“Almonor was never
arrested. The next morning, Officer Dennis filled out a computerized UF–250 and another
juvenile report worksheet, both of which noted a suspicious bulge.” (citation omitted)).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
610 VANDERBILT LAW REVIEW [Vol. 72:2:561
condition for suppression.300 The second step would be to evaluate the
deterrent value of suppressing the evidence, with one consideration
being the systemic or recurring nature of the practice. In a traditional
suppression hearing, all Almoror’s lawyer would be able to show is
that this case involved an unconstitutional stop. But with evidence of
a 911 call, a flimsy but matching description, and close proximity to
“the crime scene,” one could hypothesize a judge finding that the error
was one of isolated negligence. The stop was unconstitutional, yes, but
not deliberate, reckless, grossly negligent, or part of a systemic or
recurring pattern. As a result, the marijuana-possessing defendant
would lose.
But now imagine that a data-driven system existed that could
be mined for police practices. Using a data-focused stop-hit rate, one
could determine the actual likelihood that a stop would be successful.
As the researchers concluded after examining the Almonor case,
“[stop-hit rate] analysis indicates that there was a 3% chance that
Devin Almonor—a thin, 5 foot, 10 inch 13-year-old black teenager in
Harlem who ‘fits description’ and was behaving ‘furtively’—would be
found to have a weapon.”301 This low rate of successful stops might be
sufficient to show a recurring constitutional problem, since ninety-
seven percent of stops resulted in no finding of contraband.302 Or one
could examine the rates of all officers recovering contraband in similar
stops in the city—data on the number of stops, the location of the
stops, the type of stops, the outcome of those stops, and some figure
about whether such stops were successful would all be available. In an
individual case, one could target the rates of a particular officer’s
successful stops, the particular unit, or the particular police district.
Similarly, using mined data, one could track patterns of types
of stops. While of course every stop would need to be evaluated
individually, it might be possible to show in the aggregate that the
practice resulted in largely ineffectual searches and thus
demonstrates a systemic practice of unreasonable stops.303 And if the
police administrators knew about this practice and did nothing, a
300. Id. at 630 (“Almonor’s Fourth Amendments rights [were] violated at the inception of
both the stop and the frisk . . . .”).
301. Goel et al., supra note 246, at 217.
302. Id.
303. See Floyd, 959 F. Supp. 2d at 559 (“[T]he analysis of the UF–250 database reveals
that at least 200,000 stops were made without reasonable suspicion.”); Second Supplemental
Report of Jeffrey Fagan, Ph.D. at 10 tbl.1, Floyd, 959 F. Supp. 2d 540 (No. 08 Civ. 01034(SAS));
Jeffrey Fagan & Garth Davies, Street Stops and Broken Windows: Terry, Race, and Disorder in
New York City, 28 FORDHAM URB. L.J. 457, 496 (2000); Meares, supra note 126, at 161; Meares,
supra note 123, at 342.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 611
defendant might be able to show that the police department acted
negligently in not fixing training, policy, or practices.
Or consider a scenario similar to Utah v. Strieff but where
data-mining technologies existed to monitor the patterns of police
stops. If the question became whether detective Fackrell’s actions
were part of a systemic or recurring pattern of unconstitutional stops,
the data could now support that theory. If Utah police recorded all of
the times they followed the “stop, ask for identification, run a check”304
tactic, courts would know whether the practice was part of a systemic
or recurring practice as alleged by the defendant and dissent.
Independent of detective Fackrell, if others in the Salt Lake City
Police Department repeatedly engaged in the tactic, this too would
show a recurring problem warranting suppression. If collected and
proved, data might change the outcome of the Strieff suppression
hearing.
The point is that data mining can offer new insights into
recurring police problems or systemic practices that fill the proof gap
under the exclusionary rule, which requires some demonstration of
systemic or recurring police negligence. Again, the technology exists.
The data exists. All that is needed is to redirect the focus of the
technology toward police accountability.
B. Monitoring Technologies
Law enforcement is in the information business. Police need
information about what is happening on the streets, who is
committing crimes, and where they are taking place, as well as data
about the patterns of criminal activity and potential threats to the
community. New surveillance technologies with video, audio, tracking,
and automated alert capabilities dramatically expand the potential to
watch what happens on the streets.305 This Section looks at how
monitoring technologies provide potential mechanisms to surveil
citizens and the police and establish patterns of systemic or recurring
misconduct by police.
304. Utah v. Strieff, 136 S. Ct. 2056, 2073 (2016) (Kagan, J., dissenting) (“[T]he department’s
standard detention procedures—stop, ask for identification, run a check—are partly designed to
find outstanding warrants.”).
305. See, e.g., SLOBOGIN, supra note 35, at 3 (“What is new about today’s surveillance is the
ease with which it can be conducted; over the past several decades, technological advances have
vastly expanded the government’s monitoring ability.”); Blitz et al., supra note 35, at 56 (“To
opponents and skeptics . . . [drones] threaten to usher in Orwellian, ubiquitous surveillance.”
(citation omitted)); Blitz, supra note 35, at 1383 (describing the expansion of video surveillance
and the dramatic changes occurring in technologies that supplement and enhance such
surveillance).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
612 VANDERBILT LAW REVIEW [Vol. 72:2:561
1. Monitoring Crime
In the not so distant future, police will go on patrol in the
following surveillance environment: A network of linked cameras will
record public activity on the streets.306 Thousands of camera feeds will
relay live footage to a central command center.307 All of the video will
be digitally recorded and thus watchable after the fact to track or
investigate a crime.308 Automated algorithms programmed to spot
specific activities (for instance, an abandoned bag or a hand-to-hand
transaction) will flag particular actions for human observers.309
Particular objects, people, or activities will remain searchable for
several weeks after the fact. If a crime is later reported, the entire
incident, including the path of both perpetrator and victims, can be
replayed through the series of linked cameras. This data can be
viewed in real time or saved, creating a perfect investigative “time
machine”310 to solve the crime.
306. Gray & Citron, supra note 35, at 66 (“DAS will ensure the surveillance of New Yorkers
and the city as a whole, twenty-four hours a day, seven days a week.”); Cara Buckley, Police Plan
Web of Surveillance for Downtown, N.Y. TIMES (July 9, 2007), https://www.nytimes.com/
2007/07/09/nyregion/09ring.html [https://perma.cc/Y6P3-2PM9]; Paul Harris, NYPD and
Microsoft Launch Advanced Citywide Surveillance System, GUARDIAN (Aug. 8, 2012, 4:20 PM),
http://theguardian.com/world/2012/aug/08/nypd-microsoft-surveillance-system [https://perma.cc/
V36V-NWG7?type=image].
307. Davenport, supra note 19 (“[The NYPD’s DAS system] collects and analyzes data from
sensors—including 9,000 closed circuit TV cameras . . . .”).
308. See Amitai Etzioni, A Cyber Age Privacy Doctrine: A Liberal Communitarian Approach,
10 I/S 641, 659 (2014):
[Microsoft’s Domain Awareness System] collates thousands of pieces of information
about the same person from public sources—such as that from the city’s numerous
CCTV cameras, arrest records, 911 calls, license plate readers, and radiation
detectors—and makes them easily and instantly accessible to the police. While the
system does not yet utilize facial recognition, it could be readily expanded to include
such technology.;
Joh, supra note 2, at 49:
The N.Y.P.D. claims that the DAS can track where a car associated with a suspect is
located, and where it has been in the past days, weeks, or months. The DAS can also
check license plate numbers, compare them to watch lists, and provide the police with
immediate access to any criminal history associated with the car owner.
(citation omitted).
309. See Joh, supra note 2, at 49 (“This system gives the police real-time access to
information that can reveal connections between persons, items, and places in ways that may not
be obvious to individual crime analysts. The DAS employs video analytic software designed to
detect threats, such as unattended bags.”); Associated Press, NJ City Leading Way in Crime-
Fighting Tech, CBS NEWS (June 19, 2010, 9:30 AM), https://www.cbsnews.com/news/nj-city-
leading-way-in-crime-fighting-tech [https://perma.cc/PU29-JFGW]; Digital Justice, AOL,
Digisensory Technologies Avista Smart Sensors, YOUTUBE (Sept. 14, 2012),
https://www.youtube.com/watch?v=JamGobiS5wg [https://perma.cc/659A-V35Y].
310. Stephen E. Henderson, Fourth Amendment Time Machines (and What They Might Say
About Police Body Cameras), 18 U. PA. J. CONST. L. 933, 937 (2016).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 613
Overhead, another sort of visual time machine will be
recording public movements.311 A small plane hovers with
sophisticated cameras capable of recording entire neighborhoods.
Multiple cameras, infrared sensors, and night vision track all visible
objects.312 If a shooting should occur in a local park, the video footage
can show not only the violence involved but also the paths of the
participants before and after the incident. All public movement can be
recorded and saved for future investigative use.
Police officers entering this surveillance space will wear
attached body cameras that will record the sights and sounds of their
interactions.313 If turned on, every statement and scene will be
recorded for future prosecution.314 But, equally useful, a daily record of
contacts, conversations, and the community will be recorded for
investigators. Facial recognition can mark people by place and time.315
Search capabilities will allow particular faces to be found amid the
multitudes. Combined, facial recognition and GPS capabilities on
body-camera systems and department-issued smartphones will track
311. Monte Reel, Secret Cameras Record Baltimore’s Every Move from Above, BLOOMBERG
BUSINESSWEEK (Aug. 23, 2016), https://www.bloomberg.com/features/2016-baltimore-secret-
surveillance [https://perma.cc/8GQK-R95T].
312. See Amitai Etzioni, A Cyber Age Privacy Doctrine: More Coherent, Less Subjective, and
Operational, 80 BROOK. L. REV. 1263, 1297 (2015) (“The planes also carry infrared cameras that
can track people and cars under foliage and in some buildings.”); Ian Duncan, New Details
Released About High-Tech Gear FBI Used on Planes to Monitor Freddie Gray Unrest, BALT. SUN
(Oct. 30, 2015), http://www.baltimoresun.com/news/maryland/freddie-gray/bs-md-ci-fbi-
surveillance-flights-20151030-story.html [https://perma.cc/KS4E-ZZGC] (discussing the FBI’s
aerial surveillance operation in Baltimore, which captured thirty-six hours of video and infrared
images).
313. See Barak Ariel et al., The Effect of Police Body-Worn Cameras on Use of Force and
Citizen’s Complaints Against the Police: A Randomized Controlled Trial, 31 J. QUANTITATIVE
CRIMINOLOGY 509 (2015) (explaining the results of a controlled trial in which officers wore body
cameras during interactions with the public); David A. Harris, Picture This: Body-Worn Video
Devices (Head Cams) as Tools for Ensuring Fourth Amendment Compliance by Police, 43 TEX.
TECH. L. REV. 357 (2010) (advocating for the use of police body cameras during search and
seizure incidents); Vivian Ho, San Francisco Cops Expected to Get Body-Worn
Cameras, SFGATE (Apr. 30, 2015, 8:47 AM), http://www.sfgate.com/crime/article/San-Francisco-
cops-expected-to-get-body-worn-6232517.php [https://perma.cc/59ZA-X3H2] (describing San
Francisco Mayor Ed Lee’s efforts to equip city police with body cameras).
314. See Mary D. Fan, Justice Visualized: Courts and the Body Camera Revolution, 50 U.C.
DAVIS L. REV. 897, 908 (2017) (characterizing the modern world as a “toutveillance society”
wherein “everybody is watching everybody” and “everyone has incentive to record or control the
narrative”).
315. See Julia Angwin, Dragnet Nation: A Quest for Privacy, Security, and Freedom in a
World of Relentless Surveillance: Chapter 1: Hacked, 12 COLO. TECH. L.J. 291, 294 (2014) (“And
new tracking technologies are just around the corner: companies are building facial recognition
technology into phones and cameras, technology to monitor your location is being embedded into
vehicles . . . .”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
614 VANDERBILT LAW REVIEW [Vol. 72:2:561
the location of police patrols and citizens in granular detail.316 Anyone
who comes into contact with police officers will be caught in this
recorded web.
Police-issued computers—including handheld smartphones—
will provide updated information about the neighborhood.317 Criminal
incidents, calls for service, historic crime patterns, gang rivalries, and
predictive assessments about the crime forecast of the area will be
updated in real time and sent to officers trying to assess risk.318 As
officers patrol, new information providing the context of the places and
the people they interact with can be instantaneously retrieved. Facial
recognition technology will augment police identification and allow
automatic alerts from open arrest warrants in police databases.
Each of these technologies exists today in some form or another
in major U.S. cities. They do not all exist together, and may not for
some time, but the surveillance architecture is real and technically
possible. In New York City, the Domain Awareness System links more
than almost ten thousand cameras in a real-time surveillance net.319
In West Baltimore, Persistent Surveillance System planes flew and
recorded entire portions of the city.320 In Los Angeles, facial
recognition cameras record people near Skid Row321 and LAPD officers
316. See Sidney Fussell, The New Tech That Could Turn Police Body Cams into Nightmare
Surveillance Tools, GIZMODO (Mar. 9, 2017), https://gizmodo.com/new-ai-could-turn-police-body-
cams-into-nightmare-surve-1792224538 [https://perma.cc/W36W-6WJZ] (describing novel body
camera technology with surveillance abilities, including facial and object recognition).
317. Tim Fleischer, Officers Embrace New Smartphones as Crime Fighting Tools, ABC7NY
(Aug. 13, 2015), https://abc7ny.com/news/exclusive-officers-embrace-new-smartphones-as-crime-
fighting-tools-/928007 [https://perma.cc/7D8K-N5G3] (discussing the NYPD’s new smartphone
technology, which provides its thirty-five thousand officers with access to numerous department
databases, including the Domain Awareness System); Palantir, Palantir Mobile Prototype for
Law Enforcement, YOUTUBE (Oct. 20, 2010), https://www.youtube.com/watch?v=aRDW_A8eG8g
[https://perma.cc/Z66B-H3SX] (demonstrating software that allows law enforcement to search
numerous police databases through a mobile device).
318. See Justin Jouvenal, The New Way Police Are Surveilling You: Calculating Your Threat
“Score,” WASH. POST (Jan. 10., 2016), https://www.washingtonpost.com/local/public-safety/the-
new-way-police-are-surveilling-you-calculating-your-threat-score/2016/01/10/e42bccac-8e15-11e5-
baf4-bdf37355da0c_story.html [https://perma.cc/CQG9-R7ZV] (noting the availability of police
software that compiles data and scores a suspect’s potential for violence); Maurice Chammah,
Policing the Future, MARSHALL PROJECT (Feb. 3, 2016), https://www.themarshallproject.org/
2016/02/03/policing-the-future [https://perma.cc/PHX6-BJ83] (describing an officer’s use of
predictive policing technology in Missouri).
319. See Davenport, supra note 19; see also I. Bennett Capers, Crime, Surveillance, and
Communities, 40 FORDHAM URB. L.J. 959, 962 (2013) (describing the surveillance capabilities of
major urban cities).
320. The same company flew planes over the city of Compton in a “secret test of mass
surveillance.” Conor Friedersdorf, Eyes over Compton: How Police Spied on a Whole City,
ATLANTIC (Apr. 21, 2014), http://www.theatlantic.com/national/archive/2014/04/sheriffs-deputy-
compares-drone-surveillance-of-compton-to-big-brother/360954 [https://perma.cc/PR6U-P6LA].
321. See Garvie & Frankle, supra note 22 (“In 16 ‘undisclosed locations’ across northern Los
Angeles, digital eyes watch the public. . . . Using facial-recognition software, the cameras can
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 615
input those contact cards into the Palantir-designed social network
tracking system.322 Body cameras have been adopted in dozens of
jurisdictions, and data-driven patrols are part of even more policing
strategies.323 In fact, in April 2017, Axon (the company formally
known as Taser) offered all police officers free body cameras for a year.
This surveillance state raises obvious and poignant privacy
concerns. Scholars, including myself, have examined the costs and
problems of this new reality and even proposed constitutional
solutions to the growing danger.324 But this Article examines the silver
lining of such comprehensive surveillance as it relates to police
accountability: all of this observational data is searchable and thus
usable to visualize recurring patterns of police misconduct.
2. Monitoring Police
Imagine the same police patrol in the same surveillance state,
but with a focus on tracking police officers and targeting police
accountability.325 Police administrators want to know what particular
officers are doing on the streets as well as patterns of police activity.
Police officers drive into the networked camera field.
Automated license plate readers identify the patrol car. When the
officer gets out of her car, every single interaction can be recorded by
recognize individuals from up to 600 feet away.”); Stop LAPD Spying Coalition Visits the
Regional Fusion Center, PRIVACYSOS (Dec. 17, 2012), https://privacysos.org/blog/stop-lapd-
spying-coalition-visits-the-regional-fusion-center [https://perma.cc/WK9N-5TXQ] (spotlighting a
Los Angeles–based coalition’s efforts to end dragnet spying within the city).
322. See Brayne, supra note 232, at 992 (discussing how Palantir’s technology is used to
track “person[s] of interest” by the LAPD).
323. See supra notes 313–316.
324. See e.g., A. Michael Froomkin, Regulating Mass Surveillance As Privacy Pollution:
Learning from Environmental Impact Statements, 2015 U. ILL. L. REV. 1713, 1721 (2015):
Creating a database recording everyone’s movements allows the state to learn who
associates with whom. It chills the freedom of association no less than requiring
organizations to publish their membership lists. A government that has access to 24/7
information about the movements and habits of people is one that, even when acting
within the law, has the power to investigate people for their political activities.;
Gray & Citron, supra note 35, at 66 (noting the comparison between surveillance technology
used in New York City and “Orwell’s ‘Big Brother’ ”); Steve Mann & Joseph Ferenbok, New
Media and the Power Politics of Sousveillance in a Surveillance-Dominated World, 11
SURVEILLANCE & SOC’Y 18, 26 (2013) (“Foucault’s prisoner metaphor is no longer sufficient to
describe power relationships mediated by mobile computing and ubiquitous computing enabled
by new media.”); Richards, supra note 35, at 1953 (“The power effects of surveillance illustrate
three additional dangers of surveillance: blackmail, discrimination, and persuasion.”); see also
Andrew Guthrie Ferguson, Personal Curtilage: Fourth Amendment Security in Public, 55 WM. &
MARY L. REV. 1283, 1287 (2014) (“The question remains: does a space, constitutionally protected
from technologically enhanced surveillance, exist in public?”).
325. Capers, supra note 319, at 986; see also Fan, supra note 136, at 102–03 (positing that
“police panopticism” could increase both visibility and accountability in law enforcement).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
616 VANDERBILT LAW REVIEW [Vol. 72:2:561
surveillance cameras in real time. Supervisors can watch the complete
pattern of interactions, every stop and every search, from the
command center. Police cars can be tracked from overhead flights.
GPS tracking can watch where police officers drive or walk or chase. If
there should be an incident, a complaint, or a lawful arrest at any
point, supervisors can rewind the video to watch the entire event
occur. In fact, any of the contacts, stops, or arrests recorded can be
studied with the ease of replaying a video.
The same incident will also be recorded on body-worn cameras,
providing a more officer-centered perspective. This less structured but
equally revealing footage can track the specific details of each stop or
arrest.326 If supervisors wanted to search for all of an officer’s past
arrests, they could pull up each event. If supervisors wanted to study
each traffic stop, they could review each stop. If they wanted to
identify patterns of how frisks were conducted, when weapons were
drawn, when handcuffs were used, or the types of physical contact
initiated, they need only replace the existing automated search
capabilities (for example, targeting an abandoned bag) for the type of
event they wish to review (for example, a protective frisk). All
searches at a particular corner, all frisks by a particular officer, or all
stops by a particular unit could be identified and studied with
algorithmic ease.
The surveillance capacities of body-worn cameras will increase
with an increased capacity to search the footage. One company,
Dextro—recently purchased by Axon/Taser, one of the leading body-
camera companies—has debuted technology that can scan for any
particular object in the footage.327 As a result, police can, for example,
search for all Nike swoosh symbols, all baseball caps, or all hand-to-
hand transactions observed over a day or a week.328 The company has
explained that the process begins once the body camera identifies
objects and movements. Once identified, the footage creates a timeline
of when each action or object appears, including timestamps and
frequency data. This allows law enforcement to reduce the footage to
the exact time at which the object or motion in question appears and
add these moments to a searchable database. For instance, law
326. See, e.g., David A. Harris, How Accountability-Based Policing Can Reinforce—or
Replace—the Fourth Amendment Exclusionary Rule, 7 OHIO ST. J. CRIM. L. 149, 177–78 (2009)
(presciently discussing the evolution of police body-worn cameras as a method of police
accountability).
327. Fussell, supra note 316.
328. See id. (“Dextro scans and pinpoints objects in footage that users are looking for, for
example, a book, a Nike shoe, lines of text, or a gun. Dextro can also pick up motion information,
like handshakes or a punch.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 617
enforcement could search for “officer foot chase” or “traffic stop.”329
The result would be a timeline of all foot chases, searchable with
relative ease. This technology can also help with police training, so
officers can review their decisionmaking strategies.330 This same
technology will also record place, time, location, and conversations,
thus limiting the amount of paperwork officers need to complete on a
daily basis. In doing so, the technology will be responsible for a
massive database of all police-citizen contacts. With almost four
thousand police departments using body cameras, this technology will
potentially offer a game-changing ability to track particular things,
people, or patterns.331
The available crime and neighborhood data also provides
context for the police officer’s actions. In the same way police officers
can learn to better understand an area because of the reported data,
so too can supervisors better understand the officers’ decisions
because of the information provided to officers before the stop.332
Supervisors will know what the officers knew, what information they
checked or failed to check, and the reasonableness of their reaction.
Beyond video footage, audio surveillance capabilities can also
reveal policing patterns and practices. Professor Eberhardt’s
investigation into the Oakland Police Department involved monitoring
the language spoken between police and civilians.333 Because the
Oakland Police Department used body-worn cameras and because
those cameras recorded sound, the researchers could create a
searchable database of audio recordings of police-citizen
interactions.334
By tracking data on the tone, content, quality, and types of
phrases chosen, researchers could observe language patterns that
differed by race.335 In fact, by studying the use of “apologies” (words
329. Id.
330. See id.:
An officer’s body camera records an incident in which a cop mistook a cell phone for a
gun; the software helps pinpoint the precise moments when the cop made a mistake;
and the video is later used for training. Police departments could potentially analyze
and compile hundreds of videos for similar purposes.
331. Cf. Fan, supra note 314, at 924–28 (explaining the potential for body cameras to
positively alter the ways in which police departments engage with communities).
332. Patterns of crime may influence how officers see and react to particular neighborhoods
or patrol assignments. It might be the case that tactics in higher-crime areas differ from lower-
crime areas, and tracking those differences could alleviate community tension or improve officer
training.
333. STRATEGIES FOR CHANGE, supra note 284.
334. See id. at 15 (describing the author’s “analysis of Oakland Police Department (OPD)
stop data” in terms of linguistic exchanges).
335. Id.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
618 VANDERBILT LAW REVIEW [Vol. 72:2:561
and phrases like “excuse me,” “sorry,” and “apologies”), “gratitude”
(words like “thanks”), “formal titles” (words like “sir” and “ma’am”),
and police-relevant categories like “police equipment” (words and
phrases like “breathalyzer,” “radar,” “handcuffs,” and “badge”),336
researchers could not only see a racial difference but predict—just by
studying the language—whether the officer was speaking to an
African American or White suspect.337 The choice of words revealed
how police routinely provided more information and more procedural
details to nonminority suspects.338
The study’s results also open up the possibility to visualize
patterns of police-citizen interactions. As the researchers summarized:
One reason that law enforcement agencies do not systematically analyze BWC [body-
worn camera] footage is that they and the public tend to think of the footage as
evidence, rather than data. Evidence can prove liability or innocence in one specific case,
but data can show patterns across incidents and possibly be used to change those
patterns. Studying BWC footage in the aggregate could provide unparalleled insights
into how police officers typically interact with community members, as well as how to
improve those interactions.339
336. Id.:
To analyze officer language data on a large scale, we first created a set of categories of
officers’ language use . . . . These categories reflect both linguistics and social
psychology research, as well as new categories relevant to the particular
circumstances of police-community interactions. We then count how many officers’
utterances contain words or phrases that fit into each category. Finally, we use
statistical models to understand whether and how officers use these categories
differently depending on the race of the community member.
337. Id.:
We began with a preliminary question: Can we predict the race of a community
member simply from the words an officer uses with him or her? To answer this
question, we created a randomly selected, artificially balanced dataset of stops (N =
380) with 50% White and 50% African American community members. Then, for each
interaction, we measured a wide variety of linguistic indicators. These included:
counts of every word and pair of words, measurements for dozens of linguistic
categories, the total number of words spoken, the number of questions the officer
asked, and so on. Because we have the same number of White and African American
vehicle stops, a tool performing at chance would be 50% accurate at predicting the
race of the community member from the officer’s language. Yet our model is 68%
accurate—an improvement of 18% over chance. These results suggest that officers
speak differently to White versus African American community members.
338. Id. at 18:
After statistically controlling for whether there was a search, the result of the stop,
and the gender of the community member, we found that OPD officers more often
used these explanatory words with White community members than they did with
African American community members. These findings suggest that OPD officers
more often explain the reason for their stop to White community members . . . .
339. Id. at 14 (emphasis added).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 619
Such pattern analysis offers new abilities to improve training and
monitor different implicit or explicit biases that might undermine
trust in a community.340
Video footage can expose similar types of patterns. One
regularly recurring constitutional issue involves whether police
officers detained a suspect before requesting identification.341 As in
the Strieff case, the facts can be contested, with different and perhaps
contradictory understandings of consent, detention, and seizure at
issue. But with video surveillance, the pattern of such stops could be
studied and clarifications in trainings and protocols provided.
Similarly, the question of “furtive movements”—always difficult to
articulate—could be clarified with video evidence.
Digital surveillance technologies allow new visibility for
policing practices that usually operate without much transparency.
Systems of policing practices can be watched, analyzed, and improved
in trainings or protocols. And at some point in the future, this
surveillance will go beyond video into a whole world of wireless and
biometric data that can be collected and studied to optimize policing
practices and study policing patterns.
3. Monitoring Exclusion
Inverting the surveillance architecture to focus on police
accountability may or may not have a positive impact on improving
policing as a profession.342 But these new information sources do
provide a game-changing innovation to document systemic or
recurring negligence and thus rework the Supreme Court’s new
exclusionary rule.
At both an individual and a programmatic level, systemic and
recurring issues could be proven in court using available digital
340. See id.:
We plan to use these tools to quickly and accurately analyze the words officers use,
their tone of voice, how many turns they take in their conversations with community
members, and other indicators of the content and quality of the interaction. In
combination with other stop data (e.g., the race of the person stopped, the location of
the stop, the outcome of the stop), these tools will allow law enforcement agencies and
researchers to examine whether and how police-community interactions unfold
differently as a function of race.
341. One recurring dispute is whether the individual was “seized” for Fourth Amendment
purposes before the police officer asked for identification or whether it was a consensual
encounter.
342. See Capers, supra note 319, at 978 (“For many communities, public surveillance has the
potential to do more than simply deter crime and aid in the apprehension of law-breakers. Public
surveillance can also function to monitor the police, reduce racial profiling, curb police brutality,
and ultimately increase perceptions of legitimacy.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
620 VANDERBILT LAW REVIEW [Vol. 72:2:561
footage. Take again the example of detective Fackrell’s stop of Edward
Joseph Strieff. As the case came before the Supreme Court, there was
proof of isolated contact but no proof of systemic misconduct or any
recurring pattern of misconduct.343
But with the surveillance state watching detective Fackrell,
litigants could have an answer to whether the stop was part of a
larger pattern of unconstitutional stops. Litigants could review video
of Fackrell’s prior stops. Litigants could review audio of all the times
he asked for identification. In fact, they could search for all the times
any officer asked for identification. They could divine patterns out of
individual, fragmented practices. Studying detective Fackrell’s
movements could show that this incident was, in fact, just an isolated
mistake, and studying his interactions could demonstrate that his
misconduct deserves to be viewed with good faith deference. Or the
review could very well expose a pattern of negligence—again, a low
legal threshold signifying a failure to abide by a duty of care.344
More broadly, the same surveillance capabilities might show
that the general practices of the police department reveal a systemic
problem. The Supreme Court’s language redefining exclusion appears
to envision a structural problem akin to the NYPD’s systemic violation
of rights in their stop and frisk practices or to the type of excessive
force or unconstitutional stop practices revealed by DOJ Civil Rights
investigations.345 The issue was not the individual officer’s action but
the system that encouraged racially discriminatory stops.346 The
ability to track multiple officers over time using data analytics,
automated video searches, and audio searches could allow a more
systemic examination of police practice. The granular ability—offered
by companies like Dextro—to identify all foot chases, all interactions,
all frisks, or all physical contacts at particular places along a timeline
means that daily policing practices can be broken down into
quantifiable (and thus visible) segments.347 Patterns—for example, of
requesting identification—could be studied as a stand-alone issue.
This systemic proof would make any claim of exclusion much stronger.
Monitoring technologies could also provide capabilities for resolving
343. Utah v. Strieff, 136 S. Ct. 2056, 2063–64 (2016); see supra Section I.A.2.
344. See Strieff, 136 S. Ct. at 2063 (stating that officer Fackrell was at most negligent).
345. See id. at 2063–64; Herring v. U.S., 555 U.S. 135, 144 (2009); Floyd v. City of New York,
959 F. Supp. 2d 540, 660 (S.D.N.Y. 2013) (discussing NYPD’s systematic violation of rights); see
supra Section I.B.2.
346. See Meares, supra note 126, at 162 (describing the “organizationally determined
practice of stopping certain ‘sorts’ of people” as “imposed from the top down,” rather than
“individual incidents”).
347. See Fussell, supra note 316 (discussing the novel video-analysis technology, Dextro,
which uses object and movement identification to create a timeline of body camera footage).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 621
whether the particular officer’s actions were deliberate, reckless, or
grossly negligent and whether the officer told the truth about the
incident.348
As discussed in Part III, significant logistical and practical
challenges exist with respect to this practice of using surveillance
technologies, but in terms of technical capacity, the monitoring
technologies of the future will be quite capable of recording and
revealing systemic and recurring patterns and actions of all kinds.
C. Predictive Technologies
Police have long known that particular people drive up crime
rates.349 Stopping those suspects before they commit the next crime
has always been a challenge. New predictive technologies offer the
potential to narrow the list of suspects, using algorithmic forecasts to
target the highest-risk individuals.350 This Section looks at the
promise of predictive targeting technologies as a mechanism to
identify both at-risk suspects as well as at-risk police officers. Officers
with histories of recurring misconduct can be tracked and targeted. In
an exclusionary rule regime where failure to act on an identifiable risk
may be considered negligent, these predictive systems offer another
tool for proving systemic and recurring negligence within
departments.
1. Predicting Criminal Risk
In a handful of cities across the United States, police have
begun using algorithmic formulas to rank the most at-risk individuals
in a community.351 Most famously, the Chicago Police Department
348. See Melanie D. Wilson, An Exclusionary Rule for Police Lies, 47 AM. CRIM. L. REV. 1, 6
(2010) (“Technology and its widespread public availability provide increasing opportunities to
accurately capture police-citizen encounters and to expose police lies.”).
349. See, e.g., MEARES ET AL., supra note 242, at 1:
Data analysis immediately revealed that a very small number of neighborhoods in
Chicago are responsible for most of the city’s violence trends. The “city’s” crime
problem is in fact geographically and socially concentrated in a few highly
impoverished and socially isolated neighborhoods. Data also revealed that most
victims (and offenders) of gun violence in Chicago tend to be young African American
men who live in neighborhoods on the West or South sides of the city.
350. Ferguson, supra note 30, at 705; see id. at 736 (describing the process of targeting
individuals with predictive prosecution technologies).
351. See ANTHONY A. BRAGA ET AL., SMART APPROACHES TO REDUCING GUN VIOLENCE 12–
13, 19 (2014), https://www.nationalpublicsafetypartnership.org/clearinghouse/Content/Resource
Documents/SMART Approaches to Reducing Gun Violence.pdf [https://perma.cc/96XT-2UTP];
John Eligon & Timothy Williams, Police Program Aims to Pinpoint Those Most Likely to Commit
Crimes, N.Y. TIMES (Sept. 24, 2015), https://www.nytimes.com/2015/09/25/us/police-program-
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
622 VANDERBILT LAW REVIEW [Vol. 72:2:561
developed the Strategic Subjects List, colloquially known as the “Heat
List,” to rank the individuals most at risk of violence in the city.352 The
identified subjects on the Heat List are the people most likely to be
shot in an act of violence, as well as those most likely to do the
shooting.353 To be clear, this was not a “most wanted” list based on
past acts of criminality but rather a predictive judgment that these
individuals would be at risk in the future.
The original formula was created by Professor Miles Wernick
at the Illinois Institute of Technology and consisted largely of “co-
arrestees”—meaning the individuals arrested with suspects arrested
for violence.354 The theory behind the coarrestee connection was that
those individuals arrested with violent actors were more at risk for
being involved in reciprocal acts of gang violence.355 Because many of
the shootings in Chicago were gang related, this theory of looking at
the networks of gang members made a great deal of sense.356 In fact,
aims-to-pinpoint-those-most-likely-to-commit-crimes.html [https://perma.cc/Y4DE-B4W6]; Tony
Rizzo, Amid a Crackdown on Violent Criminals, Kansas City Homicides Sharply Decline, KAN.
CITY STAR (Jan. 1. 2015), https://www.kansascity.com/news/local/crime/article5304384.html
[https://perma.cc/GG66-E9R6].
352. Jeremy Gorner, Chicago Police Use ‘Heat List’ as Strategy to Prevent Violence, CHI. TRIB.
(Aug. 21, 2013), http://www.chicagotribune.com/news/ct-xpm-2013-08-21-ct-met-heat-list-
20130821-story.html [http://perma.cc/TTJ9-PZTW]; Mark Guarino, Can Math Stop Murder?,
CHRISTIAN SCI. MONITOR (July 20, 2014), http://www.csmonitor.com/USA/2014/0720/Can-math-
stop-murder-video [https://perma.cc/3H3N-YYMX] (“Armed with a plethora of statistics on
everything from gun violations to individual parole and arrest histories, police here are trying to
create a national model that will help them predict where shootings might occur and who might
be involved – both victims and offenders.”); Strategic Subject List, CHI. DATA PORTAL (last
updated Dec. 7, 2017), https://data.cityofchicago.org/Public-Safety/Strategic-Subject-List/4aki-
r3np [https://perma.cc/EJV4-HUYX].
353. See CHI. POLICE DEP’T, CUSTOM NOTIFICATIONS IN CHICAGO, SPECIAL ORDER S10-05, at
IV.A (Oct. 6, 2015); Editorial Board, Who Will Kill or Be Killed in Violence-Plagued Chicago? The
Algorithm Knows., CHI. TRIB. (May 10, 2016), http://www.chicagotribune.com/news/opinion/
editorials/ct-gangs-police-loury-algorithm-edit-md-20160510-story.html [https://perma.cc/EX5W-
VNCQ]; Nissa Rhee, Can Police Big Data Stop Chicago’s Spike in Crime?, CHRISTIAN SCI.
MONITOR (June 2, 2016), https://www.csmonitor.com/USA/Justice/2016/0602/Can-police-big-data-
stop-Chicago-s-spike-in-crime [https://perma.cc/T4AU-U5A3].
354. Davey, supra note 31 (discussing Professor Wernick’s original algorithm); see Guarino,
supra note 352 (describing researchers’ analysis of arrest and homicide records); Jessica
Saunders et al., Predictions Put into Practice: A Quasi-Experimental Evaluation of Chicago’s
Predictive Policing Pilot, 12 J. EXPERIMENTAL CRIMINOLOGY 347, 357 (2016) (noting the term “co-
arrestees”).
355. See Andrew V. Papachristos & David S. Kirk, Changing the Street Dynamic: Evaluating
Chicago’s Group Violence Reduction Strategy, 14 CRIMINOLOGY & PUB. POL’Y 525, 533 (2015)
[hereinafter Papachristos & Kirk, Changing the Street Dynamic] (discussing the relationship
between proximity to violence and risk of becoming a victim or perpetrator); see also Andrew V.
Papachristos et al., Why Do Criminals Obey the Law? The Influence of Legitimacy and Social
Networks on Active Gun Offenders, 102 J. CRIM. L. & CRIMINOLOGY 397, 436 (2012) (reporting
that individuals in social networks “saturated” with criminals tend to hold “negative opinions of
the law”).
356. Guarino, supra note 352.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 623
the study of networked violence had been demonstrated by social
scientists in a range of experiments, including a few studies in
Chicago.357 Essentially, the social science showed that small networks
of individuals respond to violence with cascading and escalating
additional violence.358 Put bluntly, the theory was based on the rough
logic that “if you shoot my friend, I will shoot you and your friend.”
And since those arrested together were assumed to be involved in
violent networks together, this linkage served as the proxy for
predictive risk.
An updated formula for the Heat List incorporated coarrestees,
but also included factors such as whether an individual was in a gang,
had dropped out of school, or was on probation, as well as the
individual’s connection to victims of shootings.359 The inputs have
continued to change. In May of 2017, the Chicago Police Department
explained that the list involved eight variables, “including arrests for
gun crimes, violent crimes or drugs, the number of times the person
had been assaulted or shot, age at the time of the last arrest, gang
membership and a formula that rated whether the person was
becoming more actively involved in crime.”360 The actual algorithmic
formula remains a secret, but the idea of looking for risk factors,
weighting the variables, and using the resulting list as a mechanism
to target at-risk individuals is generally well understood.361 Once
identified as being on the Heat List, police were expected to contact
the identified targets and provide them with custom notification
letters.362 These custom notification letters, and related in-person
357. See Papachristos & Kirk, Changing the Street Dynamic, supra note 355, at 533–34
(surveying studies evaluating the impact of “the focused deterrence approach” on violent crime
rates).
358. See Papachristos et al., Social Networks, supra note 242, at 1000–01 (analogizing a
Boston case study that explains the relevance of small social networks to the risk of gunshot
victimization in Chicago).
359. Saunders et al., supra note 354, at 357–58 (listing factors used for individual-level
analysis by Chicago police as “(1) demographics (gender, age at most proximate arrest, race), (2)
arrest history (number and type), (3) social network variables (number of first- and second-
degree co-arrestees who were victims of homicide), and (4) the risk score generated by IIT,” in
addition to a “second dataset contain[ing] all recorded police contact with the 17,754 arrestees
with at least one first- or second-degree association with a homicide victim and law enforcement
from 1980 through the end of the observation window”).
360. Mick Dumke & Frank Main, A Look Inside the Watch List Chicago Police Fought to
Keep Secret, CHI. SUN-TIMES (May 18, 2017), https://chicago.suntimes.com/news/what-gets-people
-on-watch-list-chicago-police-fought-to-keep-secret-watchdogs [https://perma.cc/Y64N-Z8VL].
361. See id.
362. See CHI. POLICE DEP’T, supra note 353, at IV.D:
The Custom Notification Letter will be used to inform individuals of the arrest,
prosecution, and sentencing consequences they may face if they choose to or continue to
engage in public violence. The letter will be specific to the identified individual and
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
624 VANDERBILT LAW REVIEW [Vol. 72:2:561
“call-in meetings,” were used to educate and warn the target that the
police were aware of their connection to violence and that they needed
to stop their violent ways.363 Similar predictive targeting systems—or
focused-deterrence programs—have been adopted in New Orleans,
Chicago, and Kansas City, among other cities.364
Augmenting the environmental risk factors (including, for
example, gang membership, unemployment, and neighborhood), more
sophisticated predictive systems incorporate social media usage to
predict violence.365 Antigang police units patrol YouTube and Twitter,
monitoring and interrupting gang feuds that may start on social
media but end with real bloodshed.366 In addition, social network
analysis that reveals linkages to various gangs or clues to various
disputes can be mapped through social media contacts.367 If police
incorporate those factors known about the individual inclusive of prior arrests, impact
of known associates, and potential sentencing outcomes for future criminal acts.
(emphasis added).
363. Editorial Board, ‘Moneyball’ Crime-Fighting Comes to St. Louis, ST. LOUIS POST-
DISPATCH (June 26, 2015), https://www.stltoday.com/opinion/editorial/editorial-moneyball-crime-
fighting-comes-to-st-louis/article_e61fbafa-e93c-5062-8d63-cb8ebb71ed53.html [https://perma.cc/
M9LT-RJJP] (quoting attorney Jennifer Joyce’s description of “call-in meeting” instructions:
“Here are the rules. The first group that commits a homicide, the first body that drops, we’re
coming after you and your friends. The group that does the most violence, we’re coming after
you.”); Eligon & Williams, supra note 351 (“Call-ins are central to the program. The authorities
invite about 120 of the group leaders they have identified (25 to 40 usually show up) to hear from
a range of officials, including the local and federal prosecutors, the police chief and the mayor.”);
see Editorial Board, supra (“Probation may be revoked, major and minor crimes will be
prosecuted and so will minor ordinance violations, building code violations and civil issues like
failure to pay child support.”).
364. See NOLA MURDER REDUCTION: TECHNOLOGY TO POWER DATA-DRIVEN PUBLIC HEALTH
STRATEGIES, PALANTIR 5 (2014) (on file with author); KENNETH J. NOVAK ET AL., KANSAS CITY,
MISSOURI SMART POLICING INITIATIVE: FROM FOOT PATROL TO FOCUSED DETERRENCE 7–12
(2015), http://www.strategiesforpolicinginnovation.com/sites/default/files/spotlights/Kansas City
SPI Spotlight FINAL 2015.pdf [https://perma.cc/P5E4-53A9]; Davey, supra note 31; Jason Shueh,
New Orleans Cuts Murder Rate Using Data Analytics, GOV’T TECH. (Oct. 22, 2014),
http://www.govtech.com/data/New-Orleans-Cuts-Murder-Rate-Using-Data-Analytics.html
[https://perma.cc/AT9P-WFJ7].
365. See Cheryl Corley, When Social Media Fuels Gang Violence, NPR: ALL TECH
CONSIDERED (Oct. 7, 2015), https://www.npr.org/sections/alltechconsidered/2015/10/07/
446300514/when-social-media-fuels-gang-violence [https://perma.cc/JWQ2-GPWX] (emphasizing
the importance of using social media to curb gang violence).
366. See Ben Austen, Public Enemies: Social Media is Fueling Gang Wars in Chicago, WIRED
(Sept. 17, 2013), https://www.wired.com/2013/09/gangs-of-social-media [https://perma.cc/VN8H-
Z447] (“Gang enforcement officers in Chicago started looking closely at social media sites about
three years ago . . . .”); Stroud, supra note 28 (describing various applications that allow police
officers to use keywords to search for violent social media posts).
367. See Cantú, supra note 28 (surveying different police departments that use social media
software to monitor protests); Elizabeth Dwoskin, Police Are Spending Millions of Dollars to
Monitor the Social Media of Protesters and Suspects, WASH. POST (Nov. 18, 2016),
https://www.washingtonpost.com/news/the-switch/wp/2016/11/18/police-are-spending-millions-to-
monitor-the-social-media-of-protesters-and-suspects [https://perma.cc/MHU8-PMVV] (describing
local police departments’ use of software tools to find individuals who brag about committing
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 625
want to determine gang involvement or the probable location of
potential violence, social media threats, boasts, and posturing provide
a good forecast for brewing trouble.368
Particular crimes have also been examined through the lens of
predictive analytics. The NYPD used an algorithmic process to
identify those homes most likely to be sites of domestic violence
incidents.369 Many domestic violence incidents escalate in severity, but
with over 263,207 domestic violence calls a year, New York police did
not know which homes to prioritize for additional attention.370 Using a
computer system that automatically scanned police reports for
keywords like “kill,” “alcohol,” or “suicide,” police were able to
prioritize which homes to visit and proactively respond to potentially
violent situations.371 Other predictive technologies that target
particular places or patterns of activity have been developed. Robbery,
fraud, and human trafficking all leave data trails that can be
monitored to track and predict future crime.372 The thread connecting
crimes or who post about witnessing criminal activity on social media); John Knefel, Your Social
Media Posts Are Fueling the Future of Police Surveillance: Activists Use Tech to Fuel Their
Movements, and Cops Turn to Geofeedia to Aggregate the Data, INVERSE (Nov. 20, 2015),
https://www.inverse.com/article/8358-your-social-media-posts-are-fueling-the-future-of-police-
surveillance [https://perma.cc/FD3J-FA62] (discussing police monitoring of social media and the
technology of “geofencing”).
368. See Chris J. Chasin, The Revolution Will be Tweeted, but the Tweets Will be
Subpoenaed: Reimagining Fourth Amendment Privacy to Protect Associational Anonymity, 2014
U. ILL. J.L. TECH. & POL’Y 1, 27 (2014):
Social media monitoring has also provided preemptive warnings of illegal activity,
allowing police to prevent crimes before they begin or to coordinate surveillance to
catch the criminals in the act. This preventative use is surprisingly common, with
forty-one percent of surveyed law enforcement officers reporting that they use social
media to monitor for potential criminal activity.;
Megan Behrman, Note, When Gangs Go Viral: Using Social Media and Surveillance Cameras to
Enhance Gang Databases, 29 HARV. J.L. & TECH. 315, 316–17 (2015) (describing the use of social
media, surveillance tools, and electronic databases to combat gang violence); Joseph Goldstein &
J. David Goodman, Seeking Clues to Gangs and Crime, Detectives Monitor Internet Rap Videos,
N.Y. TIMES (Jan. 7, 2014), https://www.nytimes.com/2014/01/08/nyregion/seeking-clues-to-gangs-
and-crime-detectives-monitor-internet-rap-videos.html [https://perma.cc/D3XA-PM27] (detailing
police use of music videos to target suspects).
369. See Joseph Goldstein, Police Take on Family Violence to Avert Deaths, N.Y. TIMES (July
24, 2013), https://www.nytimes.com/2013/07/25/nyregion/police-take-on-family-violence-to-avert-
deaths.html [https://perma.cc/W38J-CL8K].
370. Id.; see also Amanda Hitt & Lynn McLain, Stop the Killing: Potential Courtroom Use of
a Questionnaire That Predicts the Likelihood That a Victim of Intimate Partner Violence Will be
Murdered by Her Partner, 24 WIS. J.L. GENDER & SOC’Y 277, 283 (2009) (“Since the late 1970’s,
as researchers clamored to create instruments that could accurately predict the threat of
physical violence, over thirty-three IPV screening tools have been created.”).
371. Goldstein, supra note 369.
372. See Sneed, supra note 219 (detailing how anti–human trafficking groups can harness
data analysis software to pinpoint location information and victim demographic information);
Bernhard Warner, Google Turns to Big Data to Unmask Human Traffickers, BLOOMBERG (Apr.
10, 2013), https://www.bloomberg.com/news/articles/2013-04-10/google-turns-to-big-data-to-
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
626 VANDERBILT LAW REVIEW [Vol. 72:2:561
these strategies is that predictive variables showing a potential for
risk can be identified and, further, that intervention with the person
or location of that risk can reduce the chance of future crime.
2. Predicting Police Risk
Predictive policing technologies generally look outward toward
the criminal world. But those same risk-identification technologies can
also be turned inward toward police. In fact, the very same predictive
analytic techniques can be used to identify at-risk officers most likely
to be involved in recurring acts of excessive force or professional
misconduct.
As a technological matter, there is little difference between
isolating predictive variables that lead to high-risk behaviors in
criminals and officers. The technologies measure environmental or
personal factors that correlate with elevated risk. The variables are
different, but the underlying theory that certain environmental or
personal factors result in more risky behaviors remains the same. This
insight finds support in the long but ineffective history of Early
Intervention (“EI”) systems designed to identify and correct recurring
police misconduct.373 For decades, remedial systems to identify at-risk
officers have been implemented.374 These systems remained largely
retrospective, looking to past acts (usually limited to complaints,
accidents, or uses of force) in an effort to correct past bad behavior.
They also rarely reduced police misconduct, although in many
instances the EI systems were accompanied by other systemic changes
to improve police accountability.375
unmask-human-traffickers [https://perma.cc/S9K9-Y2MT] (reporting on the consolidation and
analysis of data from emergency calls to locate the sources of human trafficking).
373. See John A. Shjarback, Emerging Early Intervention Systems: An Agency-Specific Pre-
Post Comparison of Formal Citizen Complaints of Use of Force, POLICING, Mar. 2015, at 1 (“An
EI system is a non-punitive, data-driven management tool intended to spot officers who exhibit
performance problems such as frequent use of force incidents and high numbers of citizen
complaints.”); id. at 9 (“EI systems might have less of an influence on departments than
previously believed.”); see also id. at 2 (“An EI system is a data-driven management tool used by
departments as a mechanism for increasing police accountability.”).
374. See Harris, supra note 326, at 166:
Early intervention systems help police departments track the behavior of their
officers, something difficult to do in the absence of a data-driven, systematic effort.
The idea originated at least as long ago as 1981, in the seminal report on police by the
U.S. Civil Rights Commission, Who Is Guarding the Guardians?
(citing U.S. COMM’N ON CIVIL RIGHTS, WHO IS GUARDING THE GUARDIANS? A REPORT ON POLICE
PRACTICES (1981)).
375. See Shjarback, supra note 373, at 10 (“Overall, departments with emerging EI systems
did not appear to experience any positive outcomes (e.g. reduced complaint rates of use of force)
associated with the development and implementation of such systems . . . .”); see also id. at 8
(showing no improvement for ninety-four departments after implementation of EI systems); id.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 627
Predictive models promise something different. Predictive risk
assessments focus on police misconduct using a host of more
complicated variables beyond the traditional red flags used for
problem officers. Building off of some of the same insights used to
identify criminal actors most at risk for negative outcomes, these
sophisticated computer models look at systemic environmental risk
factors that contribute to stress, violence, and poor decisionmaking.
Professor Rayid Ghani, a data scientist at the University of
Chicago, decided to test whether big data models could predict
incidents of avoidable police-citizen conflict.376 With the full
cooperation of the Charlotte-Mecklenburg Police Department, Ghani
sought to predict the variables that might increase the risk of
potential conflicts between officers and citizens.377
The predictive model began by collecting the different types of
official police data corresponding to activities officers engage in on a
daily basis. For example, data on all police dispatches were recorded,
including time, location, and type of event.378 Similarly, all formal
(“In the aggregate, departments that have developed and implemented EI systems are generally
not experiencing lower levels of formal citizen complaints of use of force relative to before the
systems were employed.”); Robert E. Worden et al., Intervention with Problem Officers: An
Outcome Evaluation of an EIS Intervention, 40 CRIM. JUST. & BEHAV. 409, 415 (2013) (“The
evidence that supports the use of EI systems is not strong, and certainly not
commensurate . . . .”).
376. See Michael Gordon, CMPD’s Goal: To Predict Misconduct Before It Can Happen,
CHARLOTTE OBSERVER (Feb. 26, 2016), https://www.charlotteobserver.com/news/local/
crime/inside-courts-blog/article62772592.html [https://perma.cc/7PXE-F7JA] (describing the
Charlotte-Mecklenberg Police Department’s collaboration with the University of Chicago
research team in devising a way to better predict police behavior); Ted Gregory, U. of C.
Researchers Use Data to Predict Police Misconduct, CHI. TRIB. (Aug. 18, 2016),
http://www.chicagotribune.com/news/ct-big-data-police-misconduct-met-20160816-story.html
[https://perma.cc/22PW-5HH9] (describing the University of Chicago research team’s similar
work with the Chicago Police Department); Jaeah Lee, How Science Could Help Prevent Police
Shootings, MOTHER JONES (May/June 2016), https://www.motherjones.com/politics/2016/07/data-
prediction-police-misconduct-shootings [https://perma.cc/PYT5-P2RP] (recounting the
development of Professor Ghani’s unique big data approach to police-violence prevention).
377. Rayid Ghani et al., Identifying Police Officers at Risk of Adverse Events, DATA SCI. FOR
SOC. GOOD 1 (2016), https://dssg.uchicago.edu/wp-content/uploads/2016/04/identifying-police-
officers-3.pdf [https://perma.cc/7L7L-SLGS] (“Certain officers, at certain periods of time, can be
identified as being more at risk of involvement in an adverse event than others.”); see also id. at 2
(“To improve the current system, we focus on the following prediction task: Given the set of all
active officers at time t and all data from time periods prior to t, predict which officers will have
an adverse interaction in the next year.”).
378. Id. at 4:
[The Charlotte-Mecklenburg Police Department’s (“CMPD”)] system creates a
dispatch event every time an officer is dispatched to a scene—for example, in response
to a 911 call—and every time an officer reports an action to the department. . . .
Dispatch records include the time and location of all events, as well as the type of
event (e.g. robbery) and its priority. Dispatches are often linked in CMPD’s system to
other types of events, such as arrests or IA cases, that occurred during that dispatch.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
628 VANDERBILT LAW REVIEW [Vol. 72:2:561
citations,379 traffic stops,380 and arrests381 were inputted into the
system with corresponding event data and the suspect’s socioeconomic
information. Less formal “field interviews” recorded any time a person
was stopped or frisked and also included data about the event, the
officer, and the suspect.382
This event information was combined with more officer-specific
information. Internal affairs records involving prior complaints, prior
use of force allegations, vehicle pursuits and accidents, conduct
violations, injuries, and internal affairs investigations were
included.383 Actual criminal complaints against officers with all of the
accompanying location and force details were added.384 Because the
Charlotte-Mecklenburg Police Department had kept information from
an EI system, a decade’s worth of red flags for particular officers were
available for review. The information showed all officers who had been
flagged for having two or more incidents occur in the preceding 180
days.385 Finally, demographic information from employee records—
379. Id. (“The citations data provides details of each citation written by officers. Each record
contains the date and type of citation, a code corresponding to the division, and additional meta-
data such as whether the citation was written on paper or electronically.”).
380. Id. (“CMPD officers are required to record information about all traffic stops they
conduct. Records include time, location, the reason for and the outcome of the stop, if the traffic
stop resulted in the use of force, and the stopped driver’s socio-demographic profile.”).
381. Id. (“CMPD records every arrest made by its officers, including when and where the
arrest took place, what charges were associated, whether a judge deemed the officer to have had
probable cause, and the suspect’s demographic information.”).
382. Id. at 5:
A “field interview” is the broad name given by CMPD for any event in which a
pedestrian is stopped and/or frisked, or any time an officer enters or attempts to enter
the property of an individual. . . . Records contain temporal and spatial information as
well as information about the demographics about the interviewed person.
383. Id. at 34 (describing the department’s internal affairs records to include filed
complaints, as well as when “an officer uses force, engages in a vehicle pursuit, gets into a
vehicle accident, commits a rule-of-conduct violation, is injured, or conducts a raid and search,
CMPD creates an IA record”).
384. Id. at 4:
The criminal complaints data provided by CMPD contains records of criminal
complaints made by citizens. Each record includes a code for the incident, the location
of the incident, the type of weapons involved if weapons were involved, and details
about victims and responding officers. It also contains flags that include information
such as whether the event was associated with gang violence, domestic violence,
narcotics activity or hate crimes.
385. Id. at 5:
We were also given the history of EIS flags going back over 10 years to 2005. Each
record identifies the relevant officer and supervisor, the threshold triggered (e.g. more
than two accidents in a 180 day period or more than three uses of force in an 90 day
period) and the selected intervention for each flag, which can include training and
counseling.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 629
including education levels, years of experience, race, height, weight,
and gender386—was inputted along with training records.387
This event-specific and officer-specific data was then combined
with neighborhood data. Census data and city data on neighborhood
characteristics, crime, and economic health were layered in so that
responses to certain dispatches could be tracked by neighborhood.388
The final result was a computer model with 423 features that could
isolate when negative police-citizen incidents would be most likely to
occur.389
The predictive model proved quite accurate. As the researchers
summarized, “Our best performing model is able to flag 12% more
high-risk officers (true positives), while flagging 32% fewer low-risk
officers (false positives) compared to the current system.”390 Obvious
variables—like higher rates of prior adverse incidents—correlated
with higher risk, but so did unknown variables like the amount of
vacant land area in a neighborhood391 or whether the dispatch call
came from a civilian or a fellow officer (the latter corresponding with a
higher rate of violence).392 More intriguingly, the model showed that
386. Id. (“The department’s employee information includes demographic information on
every individual employed by the department, including those that have retired or been fired.
The data includes officer education levels, years of service, race, height, weight, and other
persistent qualities of officers.”).
387. Id. (“CMPD requires officers to receive rigorous training on a variety of topics, from
physical fitness to how to interact with members of the public. The department records each
officer’s training events.”).
388. Id.:
In addition to the data provided by CMPD, we also use publicly available data from
2010 and 2012 neighborhood quality-of-life studies to understand the geospatial
context of CMPD events. These studies collect data on many neighborhood features
including Census/ACS data on neighborhood demographics and data on physical
characteristics, crime, and economic vitality.
389. Id.:
The goal of the EIS is to predict which officers are likely to have an adverse event in
the near future. We formulate it as a binary classification problem where the class of
interest is whether a given officer will have an adverse event in a given period of time
into the future. . . . Efforts were chiefly geared towards the extraction of these
features - in total 432 features were used.
390. Id. at 6.
391. Id.:
First, significant controls at the neighborhood level exist within the model. Such
controls have an impact on prediction - for example, vacant land area rates are a
significant predictor of officer risk. Second, indicators such as the rates of prior
adverse incidents and sustained complaints indicate cases where IA officials
previously found officers to be at fault over and above these increased risk rates.
Combined, these observations provide support for the idea that a subset of officers are
at particular risk for adverse events, and that an EIS which controls for non-officer
level factors may be able to find such officers so that interventions can be applied.
392. Id. at 8:
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
630 VANDERBILT LAW REVIEW [Vol. 72:2:561
exposure to high stress incidents like suicides, domestic violence, or
cases involving young children resulted in a higher risk for a future
adverse outcome.393 This exposure to trauma figured prominently in
predicting future incidents.
The model offered improvements over the old EI systems. The
prior Charlotte-Mecklenburg EI system had proven to be
overinclusive,394 flagging almost half of all officers in the prior year.395
In addition, the old system did not differentiate between the types of
patrols that officers engaged in, so that midnight shifts in high-crime
areas were treated equally to more relaxed daytime patrols.396 The
result was a system both unhelpful to supervisors and easily gamed by
officers.397
Finally, the big data insights provide opportunities to change
police practices to avoid these repeating high-risk incidents and
improve training.398 The finding about trauma led police to reconsider
dispatch protocols. Now, a more targeted dispatch system avoids
sending officers who have recently been exposed to high-stress
situations to the next triggering crime scene.399 This insight could
encourage more officer-centric training and counseling services about
“Hot” dispatches initiated by officers themselves (as opposed to citizens by way of 911
calls), seem more likely to end in adverse outcomes. Indicators of heightened officer
stress (hours on duty) and aggressive policing style (discretionary arrest rate), seem to
also have a positive impact on the risk of adverse outcomes.
393. Id. at 5 (“Notably among incident sub-types, we track incidents we believe are likely to
contribute to officer stress, such as events involving suicides, domestic violence, young children,
gang violence, or narcotics.”).
394. Id. at 2:
Current EISs detect officers at risk of adverse events by observing a number of
performance indicators and raising a flag when certain selection criteria are met.
These criteria are usually thresholds on counts of certain kinds of incidents over a
specified time frame, such as two accidents within 180 days or three uses of force
within 90 days.
395. Id. (indicating that current EI system thresholds fail to consider important factors,
potentially rendering them overinclusive).
396. Id. (“For example, CMPD’s system uses the same thresholds for officers working the
midnight shift in a high-crime area as an officer working in the business district in the
morning.”).
397. Id. at 3.
398. Id. at 2:
The system described here is the beginning of an effort that has the potential to allow
police chiefs across the nation to see which of their officers are in need of training,
counseling, or additional assistance to make them better prepared to deal safely and
positively with individuals and groups in their communities.
399. Id. at 9 (“Our dispatch-level models take the first steps toward predictive risk-based
dispatch decisions, where an officer who is at higher risk of an adverse incident for that dispatch
can potentially be held back and a different officer, at a lower risk score, can be dispatched.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 631
trauma and how to address posttraumatic stress.400 It also could flag
circumstances that repeatedly create the potential for risk.
For privacy reasons, Professor Ghani’s team purposely removed
identifying material from the data and looked for common
environmental factors as predictors of conflict rather than looking at
individual “problem officers.” The idea was to identify patterns of
conflict, as opposed to predicting individuals within those patterns.
But, while necessary for political acceptance in Charlotte-
Mecklenburg, this limitation does not need to be implemented in the
future. In fact, the same targeted predictive assessments akin to the
Heat List could be used to identify particular at-risk officers.
For example, variables that have made it into various Early
Warning or EI systems demonstrate data points that could be used to
target at-risk officers. Professional factors such as prior complaints,
prior uses of force, unprofessional conduct, or accidents are known red
flags for behavioral problems (and were confirmed in Professor
Ghani’s data).401 Personal stressors such as financial difficulties,
divorce, injury, death in the family, or other losses could all signal a
higher risk of professional stress. Psychological or medical factors
resulting from posttraumatic stress, depression, or medical problems
could also factor into officer reactions. Finally, personal activities—
lifestyle choices and even hobbies402—can influence risk. While none of
these factors predict police misconduct outright, they might predict
when a higher risk of police misconduct exists.403 In combination with
the environmental assessments of Professor Ghani, these predictive
models could become very accurate.
400. Id.:
Our model significantly outperforms the existing system at the Charlotte-
Mecklenburg Police Department (CMPD). Our model also provides risk scores to the
department, allowing them to more accurately target training, counseling, and other
interventions toward officers who are at highest risk of having an adverse incident.
This will allow the department to better allocate resources, reduce the burden on
supervisors, and reduce unnecessary administrative work of officers who were not at
risk.
401. See Chani et al., supra note 383 (describing the situations in which CMPD creates an
internal affairs record).
402. David J. Krajicek, What’s the Best Way to Weed Out Potential Killer Cops, ALTERNET
(May 15, 2016), https://www.alternet.org/civil-liberties/whats-best-way-weed-out-potential-killer-
cops [https://perma.cc/23CA-7B58] (discussing a correlation between Muay Thai (a combat
martial art) and officer violence). But see Cynthia Lee, Race, Policing, and Lethal Force:
Remedying Shooter Bias with Martial Arts Training, 79 LAW & CONTEMP. PROBS., no. 4, 2016, at
145, 16070 (discussing the positive impact of certain martial arts on police training).
403. Just as the Heat List does not predict violence but instead merely predicts a higher risk
of potential violence, any algorithm using these variables will also only predict a “risk” of
misconduct.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
632 VANDERBILT LAW REVIEW [Vol. 72:2:561
In addition, if the predictive models incorporate social media
postings and other forms of communicative activity, other risk factors
might emerge. In the DOJ report on the Chicago Police Department,
the investigators discovered troubling examples of racial and ethnic
bias in police social media postings.404 As with law enforcement
monitoring for criminal risk, social media reflects current thinking,
emotions, and influences of real people in real time. Monitoring bias or
hate articulated in social media environments might provide red flags
of future behavioral problems. Obviously, this sort of employee
surveillance will elicit resistance from rank-and-file officers.405 No
employee enjoys at-work surveillance, and most would balk at
supervisors reviewing off-duty, even if publicly accessible, social media
posts. Ironically, the major complaint by officers was a feeling of
preemptive punishment for actions they had yet to take—the same
complaint of communities targeted by predictive policing technologies.
This tension is discussed further in Part III.
3. Predicting Exclusion
A predictive warning system that tracks past officer
misconduct would be relevant for proving recurring patterns of
misconduct.406 Evidence about a particular officer flagged for
repetitive unconstitutional stops would be relevant in a suppression
hearing to show that the stop at issue was not an act of isolated
negligence. Had detective Fackrell been flagged as an officer who
routinely had complaints of unconstitutional stops brought against
him, this information would fit the definition of “recurring” negligence.
Several jurisdictions have begun creating such Police Accountability
404. DOJ CHICAGO REPORT, supra note 158, at 15 (“Moreover, we found that some Chicago
police officers expressed discriminatory views and intolerance with regard to race, religion,
gender, and national origin in public social media forums . . . .”); see also id. at 147 (“One officer
posted a status stating, ‘Hopefully one of these pictures will make the black lives matter activist
organization feel a whole lot better!’ with two photos attached, including one of two slain black
men, in the front seats of a car, bloodied, covered in glass.”); id. (“Supervisors posted many of the
discriminatory posts we found, including one sergeant who posted at least 25 anti-Muslim
statements and at least 43 other discriminatory posts, and a lieutenant who posted at least five
anti-immigrant and anti-Latino statements.”); id. (describing an officer “who had posted racist
comments and had called for a race war on social media forums”).
405. See Ifeoma Ajunwa, Kate Crawford & Jason Schultz, Limitless Worker Surveillance, 105
CALIF. L. REV. 735 (2017) (discussing the potential privacy violations stemming from modern-day
worker surveillance technology and describing these innovations as a “decimat[ion] [to] worker
privacy”).
406. See Harris, supra note 326, at 165–66 (“[W]e should use early intervention systems:
data-driven accountability structures designed to detect, track, and highlight various aspects of
police officer conduct.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 633
Databases—or digital “bad cop” lists—with precisely this type of
information.407
The identified risk may not only be with individual officers but
with patterns of departmental misconduct as well. The same
predictive risk assessment could identify whether particular police
units possess a heightened risk of violence or exhibit patterns of
unconstitutional stops. The percentages of constitutional stops
memorialized in the DOJ reports could be broken down to particular
units or officers and used as evidence in suppression hearings.
Obviously, variables such as the type of patrol, neighborhood, and
time of day would need to be factored in, but this is exactly the type of
nuance that Professor Ghani’s researchers focused on in their study.408
Recurring patterns in particular places might give reason to see a
problem that cannot be excused as isolated negligence.
In addition, recurring incidents of misconduct could be
identified and, if not addressed, could lead to liability for negligence.
For example, if a predictive warning system flagged an officer as likely
to be involved in unconstitutional misconduct and police
administrators did not adequately respond to the warning despite a
duty to train and to develop policies and practices, this failure to act
could give rise to a negligence claim. Or perhaps if it were shown (as it
was in Charlotte-Mecklenburg) that responding to a traumatic event
such as suicide leads to a higher likelihood that the officer’s next
interaction will be violent and police administrators still assign the
traumatized officer to the next high-risk situation, it could be argued
that the police department acted negligently (if not recklessly) in
ignoring a clear risk. In such a case, there exists a foreseeable risk, an
alternative option, and a decision to ignore the risk. Such patterns,
once revealed through data, put the administrators on notice of a
systemic problem. And if that systemic problem arises in a case before
a court, then the pattern could be relevant to the exclusionary rule
decision.
407. See Cynthia H. Conti-Cook, Defending the Public: Police Accountability in the
Courtroom, 46 SETON HALL L. REV. 1063, 1084 (2016) (“In 2014, The Legal Aid Society
announced the Cop Accountability Project—anchored by a database for police misconduct—
intended to serve its clients, its attorneys, and the community.”); Jason Tashea, Clicking for
Complaints: Databases Create Access to Police Misconduct Cases and Offer a Handy Tool for
Defense Lawyers, 102 A.B.A. J., Feb. 2016, at 17, 18:
The New York City database houses information on more than 7,000 NYPD officers
with a paper trail of alleged or proven misconduct. The files come from a number of
sources, including the news, state and federal lawsuits, criminal decisions, the federal
court’s PACER database, New York City Civilian Complaint Review Board hearings,
NYPD Internal Affairs complaints, social media, and Legal Aid Society attorneys’ own
experiences in court and with clients.
408. See supra notes 377–400 and accompanying text (describing Professor Ghani’s study).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
634 VANDERBILT LAW REVIEW [Vol. 72:2:561
D. Programmatic Benefit of Blue Data
In addition to the instrumental benefits discussed above, the
move toward blue data offers one final, broader benefit to how courts
think about the Fourth Amendment. Blue data encourages courts to
think programmatically about the Fourth Amendment.409
In recent years, scholars have begun to rethink the Fourth
Amendment as a system of rules to be analyzed separate from the
individual cases coming before courts for suppression. In Professor
Daphna Renan’s words, the Fourth Amendment should be understood
in terms of “programs of surveillance” not in terms of transactional
acts.410 As such, policing can borrow from administrative law
principles and be regulated accordingly.411 Professor Tracy Meares
demonstrated that the NYPD stop and frisk program should be better
understood as an unconstitutional “program” and not as a series of
individualized unconstitutional incidents.412 This was also Justice
Sotomayor’s insight in Strieff, where she wrote that the warrant check
was part of a system of unconstitutional searches for evidence.413
Obviously, blue data systems offer new ways to visualize the
programmatic or systemic nature of police misconduct. Blue data is a
visualization tool, and courts will thus have the ability to see beyond
individual actions to systemic conduct, whether through data mining
or video surveillance or some other technology. Blue data can thus be
a tool to bolster these new Fourth Amendment theories.
More practically, once Fourth Amendment “incidents” are
thought of as programmatic, it becomes easier to bring claims of
systemic negligence in court. It may not be possible for police
administrators to know about unconstitutional “transactions” of
individual officers, but they can know—and should know—about
409. See, e.g., Andrew Manuel Crespo, Systemic Facts: Toward Institutional Awareness in
Criminal Courts, 129 HARV. L. REV. 2049, 2052 (2016) (explaining that in order to accomplish
broader institutional awareness, criminal courts must consider “facts about the criminal justice
system itself, and . . . the institutional behavior of its key actors”).
410. Daphna Renan, The Fourth Amendment As Administrative Governance, 68 STAN. L.
REV. 1039, 1041 (2016); see id. at 1042 (“While our Fourth Amendment framework is
transactional, then, surveillance is increasingly programmatic.”).
411. See Christopher Slobogin, Policing As Administration, 165 U. PA. L. REV. 91, 97 (2016)
(“[A] reframing of panvasive searches and seizures as administrative actions gives significant
weight to legislative and executive decisionmaking, and it draws from the Court’s precedent.”).
412. See Meares, supra note 126, at 162 (arguing that a mass of stop and frisks is not simply
an aggregation of individual incidents but rather a program in which police “engage in an
organizationally determined practice of stopping certain ‘sorts’ of people for the stated purpose of
preventing or deterring crime”).
413. See Utah v. Strieff, 136 S. Ct. 2056, 2066 (2016) (Sotomayor, J., dissenting) (“[The
warrant check] was part and parcel of the officer’s illegal ‘expedition for evidence in the hope
that something might turn up.’ ” (quoting Brown v. Illinois, 422 U.S. 590, 605 (1975))).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 635
unconstitutional programs. If administrators become aware of these
problems, courts can find the requisite negligence. If administrators
do not know about these recurring problems but should be aware of
them, courts may find negligence. And if administrators do nothing to
find out about these problems, courts may still find negligence if the
administrators had a duty to be aware. Courts could even create an
affirmative duty to investigate if policing programs continue to
generate recurring constitutional problems.
The ability to visualize recurring problems creates legal
liability for failing to act. In so doing, blue data can provide another
negligence-related legal avenue to bring suppression claims under the
Supreme Court’s new application of the exclusionary rule, which
emphasizes that claims of systemic negligence warrant exclusion
while claims of isolated negligence do not. In addition, this type of
systemic misconduct can be used in civil rights actions under 42
U.S.C. § 1983 and the federal government’s ability to investigate
patterns and practices of police abuse under 42 U.S.C. § 14141.414
III. THE REVEAL OF RESISTANCE
The use of data mining, surveillance, and predictive analytics
to target police negligence will likely face resistance. Police officers,
administrators, and unions will probably protest the invasion of
personal and professional privacy it threatens. Legal battles will erupt
over whether (and how) to collect, sort, and introduce evidence from
these new blue data systems in ordinary suppression hearings.
Technological hurdles will divide jurisdictions between those
departments that can turn surveillance technology into methods of
police accountability and those without that capacity. Police will be
joined in this criticism of data-driven surveillance by an odd
consortium of civil liberties groups resistant to erecting the larger
surveillance architecture and defense lawyers unwilling to concede a
need for a secondary Herring analysis before suppression. The future
of the exclusionary rule is already clouded, and the rise of new
information streams may not make it any clearer.
Yet, this response of resistance is itself revealing and worth
studying. Arguments pushing back against surveilling police officers
also have application to surveilling citizens. The challenges of
technology and a growing reliance on big data systems suggest
universal cautions about the dangers and costs of any data-dependent
system. These issues of professional resistance, legal resistance, and
414. See supra note 10.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
636 VANDERBILT LAW REVIEW [Vol. 72:2:561
technological barriers offer no simple answers but do offer an
opportunity to rethink how the new exclusionary rule should interact
with the even newer technologies being developed to assist law
enforcement. This Part seeks to understand this resistance, unpacking
the practical realities and possible responses, as well as the insights to
be gained from examining why there will be—and probably should
be—serious resistance to blue data.
A. Police Resistance
Police accountability measures have faced resistance in the
past. In fact, one can easily find open resistance in response to many
prior police accountability proposals.415 Police officers have resisted
the implementation of early warning systems.416 Police unions have
resisted releasing officer personnel (i.e., misconduct) records and other
accountability reforms.417 Police administrations (and cities) have
pushed back on federal oversight.418 And while some jurisdictions have
embraced police accountability, many more have fought vocally to stop
proposed changes. Even in the era of Black Lives Matter, which raised
consciousness of racial bias and excessive force in policing, the rise of
the Blue Lives Matter countermovement shows the long-standing
protective reaction to any public criticism of police misconduct.419
415. See, e.g., David H. Bayley, Police Reform: Who Done It?, 18 POLICING & SOC’Y 7 (2008)
(describing how modern reform in policing has been met by resistance); Sklansky, Police and
Democracy, supra note 123, at 1773–74 (detailing historical aspects of police reform); Steve
Wilson & Kevin Buckler, The Debate over Police Reform: Examining Minority Support for Citizen
Oversight and Resistance by Police Unions, 35 AM. J. CRIM. JUST. 184, 188 (2010).
416. See Harris, supra note 326, at 168 (discussing the professional sanctions that officers
can face if flagged by an early warning system).
417. See Rushin, supra note 131, at 154 (“[C]ollective bargaining and civil service protections
inadvertently discourage police management from responding forcefully to misconduct.”);
Walker, supra note 122, at 72:
Collective bargaining agreements, for example, contain provisions related to the
investigation of alleged officer misconduct (whether on the basis of a citizen complaint
or an internally generated complaint) that impede a timely and thorough
investigation. Officer appeals of discipline, meanwhile, may involve procedures that
tend to increase the likelihood of disciplinary sanctions being mitigated or overturned.
418. See Barbara E. Armacost, Organizational Culture and Police Misconduct, 72 GEO.
WASH. L. REV. 453, 533 (2004) (“[E]fforts by outside agencies to collect and analyze information
in a potentially adversarial framework, such as a § 14141 lawsuit, may lead police officers to be
defensive and uncooperative.”). But see PRESIDENT’S TASK FORCE ON 21ST CENTURY POLICING,
FINAL REPORT 61 (2015), https://cops.usdoj.gov/pdf/taskforce/taskforce_finalreport.pdf
[https://perma.cc/E3YS-ULAS] (proposing an increased focus on data collection about policing
practices).
419. See Jim Salter & David A. Lieb, New ‘Blue Lives Matter’ Laws Raise Concern Among
Activists, ASSOCIATED PRESS (May 26, 2017), https://www.apnews.com/f550bca209d6467995530d
4d82b4fbb7 [https://perma.cc/3XJQ-YBTN] (discussing activists’ reactions to laws permitting
heightened sentences for people who assault or kill police officers).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 637
A proposal to redirect existing data-driven surveillance
systems toward police accountability will likely meet similar
resistance. The reasons are fairly obvious. Such a system would
invade personal privacy, restrict professional autonomy, constrain
actions and language, and lead to increased supervision, training, and
potentially negative professional outcomes. Police officers, like most
employees, would rather avoid the adverse effects of worker
surveillance, especially when such oversight is couched in
dehumanizing terms like “predictive analytics” or “data mining.”420
Such invasive personal investigation could also undermine
recruitment efforts and employee morale if potential officers did not
want to have their own lives policed.
Police resistance exerted in either formal/informal or
intentional/unintentional ways could undermine the ability to use blue
data for exclusionary rule purposes. At the front end, since police
agencies and officers would be responsible for setting up the
technologies, they could also thwart any application directed toward
police.421 This resistance could be intentional, inadvertent, or due to
cost and logistical concerns.
Similarly, as has been seen with other accountability
technologies like dashboard cameras or body cameras, police have
been known to intentionally frustrate the system by turning the
cameras off.422 Put simply, if police wished to not comply with a data
collection system or figured out ways to make recovering the data too
difficult, the information’s utility in suppression hearings would be
quite limited. If police simply stopped collecting the underlying data,
blue data would not exist. In both Oakland and New York City, the
data collection was mandated by a court order. Unquestionably, an
intentional effort to undermine data collection would undercut the
value of this Article’s proposal to use such data in suppression
hearings.
Inadvertent resistance also occurs when police make errors in
data collection. The problem of data bias is endemic to all data-driven
systems, and the difficulties of collecting police data are no different.
420. See Don Peck, They’re Watching You at Work, ATLANTIC (Dec. 2013),
http://www.theatlantic.com/magazine/archive/2013/12/theyre-watching-you-at-work/354681/
[https://perma.cc/J94M-6XZP] (explaining how the use of big data in human resources is
transforming how employers hire, fire, and promote employees).
421. Most of the technologies already exist but are currently directed at civilians and not
police, demonstrating that, given the choice, police may choose not to have surveillance directed
toward their professional work.
422. See Laurent Sacharoff & Sarah Lustbader, Who Should Own Police Body Camera
Videos?, 95 WASH. U. L. REV. 269, 290 (2017) (recognizing that the “power to stop recording has
led, in a great many cases, to abuse”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
638 VANDERBILT LAW REVIEW [Vol. 72:2:561
As I have written in other contexts, bad data can corrupt an otherwise
good data-driven system.423 Ensuring data integrity means adopting
systems of data collection that are automatic and create automatic
data trails (so as to see if the data is being manipulated).424 Data
errors can be compounded by the growing volume of information being
produced. Every day, every shift, there is more data collected without
the commensurate resources to manage the accumulating data.
Because of this volume of data, police may not be able to maintain the
data systems to a level of accuracy necessary for use in court.
More practically, police may not be able to afford the cost of
these new technologies. Big data technologies are expensive.
Additionally, the data needs to be integrated into existing systems and
must be updated and its accuracy maintained. Both in terms of having
the financial ability to invest in the technology and the human
capacity to use the available amount of data, cost constraints may
undermine any potential utility. Cost workarounds such as partnering
with private companies might make sense in terms of efficiency and
expertise, but the outsourcing of local police power creates real
dangers.425 Private companies could face ethical problems, conflicts, or
confidentiality issues, and a growing dependence on private companies
could undermine local public authority. Cost might thus create a real
if unintentional barrier to adoption of blue data systems.
Whatever the practical limitations to implementation, police
resistance to blue data does reveal a deeper truth about surveillance
and data-driven suspicion. The natural police resistance to technology
parallels community resistance to the same technology. Citizens also
reflexively resist any technology which threatens to invade personal
privacy, restrict personal autonomy, constrain actions or language, or
lead to increased surveillance or negative outcomes. Police fears of
blue data are the fears of big data surveillance more generally.
One insight from the police pushback to blue data is that this
resistance might inform how local communities should respond to
proposed new surveillance technologies. Resistance can be an
educational moment. The successful push of police unions to thwart
423. See Ferguson, supra note 2, at 398–400 (discussing issues of accuracy with big data
collection systems).
424. See Miriam H. Baer, Pricing the Fourth Amendment, 58 WM. & MARY L. REV. 1103, 1160
(2017) (proposing a regime that includes the use of body and dash cameras, periodic audits of
search data, and imposition of stiff penalties for providing false information for ensuring the
accurate accounting of the number and types of searches officers perform).
425. See Elizabeth E. Joh, The Undue Influence of Surveillance Technology Companies on
Policing, 92 N.Y.U. L. REV. 101, 126 (2017) (“The continuing influence of surveillance companies
even after police have purchased their services further removes policing from traditional
mechanisms of oversight.”).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 639
any data accountability project stems from organized efforts framed
around appeals to fairness, due process, and concerns about personal
privacy and free expression. Police unions have successfully
weaponized the fear that a hard-working civil servant may be
professionally penalized because of an algorithmic judgment of future
risk. Yet, those same basic fairness issues apply in civilian
surveillance of targeted communities and can also be used to frame a
message of resistance.
In general, however, citizens have lacked the political
organization and urgency that police advocates have developed. The
message may be felt but not always heard. This may be changing in a
few cities where this democratic voice against police surveillance has
been growing louder.426 In Oakland, a Privacy Advisory Board was
created to advise the city council on new police surveillance
technologies,427 and similar surveillance awareness bills have been
considered in eleven other jurisdictions.428 Seattle enacted one of the
most comprehensive local surveillance ordinances in the country,
mandating review of police surveillance technologies.429 On a local
stage, many groups are coalescing around the idea of ensuring
transparency and accountability for new data-driven policing
technologies.430
426. See Andrew Guthrie Ferguson, The Fragmented Surveillance State, SLATE (Nov. 10,
2017), http://www.slate.com/articles/technology/future_tense/2017/11/the_united_states_
fragmented_surveillance_system.html [https://perma.cc/54XX-98UM] (discussing how some cities
require civilian surveillance over new police technologies).
427. See Darwin BondGraham, Oakland Privacy Commission Approves Surveillance
Transparency and Oversight Law, E. BAY EXPRESS (Jan. 6, 2017),
https://www.eastbayexpress.com/SevenDays/archives/2017/01/06/oakland-privacy-commission-
approves-surveillance-transparency-and-oversight-law [https://perma.cc/G6FP-5T4U] (detailing
the proposal for a Surveillance and Community Safety Ordinance which would require “[c]ity
agencies . . . to seek city council approval before purchasing new technologies, and the law also
imposes reporting requirements so that the public can evaluate the costs and benefits of
technologies that monitor and track people”).
428. Jessica Anderson, 11 U.S. Cities to Consider Legislation to Require Greater
Transparency for Police Surveillance Programs, BALT. SUN (Sept. 21, 2016),
http://www.baltimoresun.com/news/maryland/baltimore-city/bs-md-aclu-police-surveillance-
20160921-story.html [https://perma.cc/JJ47-QWE3] (listing other cities that have considered
surveillance awareness bills, including Seattle, Richmond, and Milkwaukee).
429. About the Surveillance Ordinance, SEATTLE.GOV, https://www.seattle.gov/tech/
initiatives/privacy/surveillance-technologies/about-surveillance-ordinance (last visited Mar. 16,
2019) [https://perma.cc/72JL-2AF2].
430. See Jose Pagliery, ACLU Unveils Privacy Fight in 16 States, CNN MONEY (Jan. 21,
2016, 11:53 AM), http://money.cnn.com/2016/01/20/technology/aclu-state-privacy-laws
[https://perma.cc/EN2J-P9CY] (explaining how sixteen states are considering measures to
protect personal information).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
640 VANDERBILT LAW REVIEW [Vol. 72:2:561
To be clear, the symmetry of surveillance resistance431 need not
be exact. There may be reasons to increase surveillance on citizens but
not on police. But the fact that police raise reasonable concerns about
the intrusiveness of surveillance offers a lesson on how to evaluate
new privacy-invading technologies. The argument that the
surveillance targets good cops as well as bad also highlights the
overinclusive nature of public surveillance (targeting innocent citizens
along with the guilty). The argument about the unfairness of being
predictively flagged for conduct which has not yet occurred parallels
the community’s fear of predictive targeting. Arguments about the
danger of correlative suspicion, as opposed to observed suspicion, raise
wide-ranging issues of accuracy, transparency, and individualized
justice. Seen through the eyes of police officers wishing to avoid
negative professional discipline, the arguments against surveillance
are sympathetic and meritorious. But that feeling should also transfer
to communities wishing to avoid the same harms.
Whether practically feasible or not, as a thought experiment,
the push for blue data brings in stark relief the concerns of all citizens
wishing to avoid heightened surveillance. The pushback of police
resistance offers a powerful example for ordinary citizens also
concerned with invasive new technologies. If we take seriously the
resistance to blue data, we may also moderate the rush toward greater
surveillance. If police fear that predictive analytics are unfair to them,
then how can one dismiss citizens’ complaints about a similar
technology?
In the end, police resistance to blue data may also be
unavailing. The reason: once public-safety-oriented surveillance
technologies have been turned against the citizens, it will be difficult
to hide the data that also captures the police. The always-recording
cameras exist in parts of New York City. The technology to search for
each and every stop exists. The data will be available. Big data
surveillance technologies are largely undiscriminating in who gets
captured in the net and, once vacuumed up, the data exists for
enterprising litigants to find. As such, the choice may really be about
whether to adopt big data surveillance technologies in the first
instance, recognizing that once adopted these technologies will watch
everyone.
Further, in an exclusionary rule regime in which recurring and
systemic negligence legally matters, incentives exist to use this data
in suppression hearings. Whether or not police wish to give the data
431. See Joh, supra note 35, at 1000–02 (describing the rise of antisurveillance methods of
protest and privacy).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 641
up, it will be demanded and likely produced under court order.432 If
video of all of detective Fackrell’s past requests for identification
existed, it may be difficult for a police department not to comply with
a valid lawful request for the information. If a computer model
predicted detective Fackrell as someone likely to violate constitutional
rights and that risk assessment is requested by the defense, the data
will need to be turned over. Once built, the surveillance systems will
not be limited only to the police.
B. Legal Resistance
In addition to police resisting the creation of blue data systems,
criminal courts—including judges and litigants—may resist
developing a record of recurring or systemic negligence. For different
reasons, judges, prosecutors, and defense lawyers may choose a path
of resistance rather than acquiescence to the introduction of blue data.
To put the legal burden in context, most suppression hearings
in criminal cases (suppressing, for example, narcotics, weapons, or
stolen goods) occur quickly, without a significant amount of pretrial
litigation. Motions are filed and witnesses are called, but within a
limited scope. Tactical pressures to limit the amount of evidence
introduced before trial, as well as relevance considerations, further
reduce the amount of testimony. At most, each side might call a few
witnesses to testify to the relevant facts and might make a few
arguments about the relevant case law governing those facts before
the proceeding is over. Within this practice, which is fairly standard in
state courts, the idea of introducing evidence of systemic and
recurring misconduct becomes quite disruptive, requiring more
resources, time, and effort for the court system.
From a trial judge’s perspective, the additional burden of
applying the Herring test to an ordinary case will be both time-
consuming and confusing. In an earlier article, I examined the
definitional and practical problems with the Supreme Court’s use of
“deliberate,” “reckless,” and “grossly negligent” as those terms relates
to individual officers.433 Similarly, the burden to show systemic or
432. Or production of the data may be required under Brady v. Maryland, 373 U.S. 83
(1963), or other discovery rules. See, e.g., Conti-Cook, supra note 407, at 1074.
433. Ferguson, supra note 9, at 644–56; see also Herring v. United States, 555 U.S. 135, 151
(2009) (Ginsburg, J., dissenting); id. at 157–58 n.7 (Breyer, J., dissenting) (“It is not clear how
the Court squares its focus on deliberate conduct with its recognition that application of the
exclusionary rule does not require inquiry into the mental state of the police.”); Laurin, supra
note 60, at 727 (“On its face, the Court’s insistence that the standard it articulates be applied
objectively seems nonsensical: Even if the lowest grade of culpability to trigger exclusion, gross
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
642 VANDERBILT LAW REVIEW [Vol. 72:2:561
recurring patterns means opening up an ordinarily limited hearing to
significantly more information. Aggressive defense lawyers will
demand truncated Section 1983 hearings, developing the same record
of a custom, policy, or practice, but with blue data evidence. Additional
witnesses will be needed, including experts, to establish the duty of
care and a baseline number for “recurring” problems.434 Trial judges
who would prefer not to be reversed on appeal might be cautious in
limiting evidence legally necessary to demonstrate recurring or
systemic problems since the Supreme Court has suggested their
importance. Findings of fact will need to be longer and will be more
labor intensive. And all of these decisions will be made in an uncertain
legal atmosphere, with little clarity about the definitions of
“recurring,” “systemic,” or even “negligence” in the context of a
suppression hearing.
From the defense lawyer’s perspective, the burden of proving
non-case-related facts may be too taxing to undertake. Defense
counsel may find it hard enough to litigate the facts at hand, let alone
all other stops an officer conducted. In busy, urban courthouses, the
ability to litigate pretrial motions ahead of time may be nonexistent.
Within this crush of cases, litigating the equivalent of a massive
structural reform challenge borders on impossible. Even on a small
scale, the burden of blue data requires additional discovery motions,
additional time to review hours of footage, and the wherewithal to use
the available data in one’s case. While technically possible—and
perhaps even appealing—this change adds real practical difficulties
for defense attorneys.
Further, Herring’s change legally weakens the defense’s overall
constitutional claims. Many defense attorneys may resist the idea that
Herring imposes a second analytical step for suppression. The
automatic linkage of a constitutional wrong and the suppression
remedy has been ingrained in practice for decades, and the idea of
conceding that automatic linkage is not appealing to defense lawyers.
It is for that reason, perhaps, that Herring’s second step appears to be
ignored in many courthouses. While some judges certainly conduct the
second step of the analysis,435 many simply suppress evidence after
finding a constitutional violation.
negligence, could be assessed solely by reference to objective factors, proof of reckless or
deliberate conduct typically requires a subjective inquiry.” (footnotes omitted)).
434. An open question is how to define “recurring.” The threshold question will require both
a numerical answer as well as a temporal answer since the question of recurring within what
timeframe would also have to be answered.
435. See Utah v. Strieff—Leading Case, supra note 49, at 337–38 (explaining that the Utah
Supreme Court performed the second-step attenuation analysis).
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 643
From a prosecutor’s perspective, Herring offers a second bite at
the apple after a constitutional violation. While appealing in theory, if
recurring or systemic negligence becomes the centerpiece of a motions
hearing, the entire proceeding shifts from the actions taken by the
police officer in this particular case to all actions taken in other cases.
Having to defend, or more likely seek to limit, extraneous information
about bad policing practices across a city creates a real burden on
prosecutors. Prosecutors also have too many cases and not enough
time, so the burden of adding civil litigation-like responsibilities—
culling discovery requests, sorting through data and footage, and the
like—may be too much. Finally, the introduction of prior police
misconduct or evidence of systemic behavior raises a concern that this
information could spill over into trial.436
These arguments for resistance reveal an underappreciated
difficulty of the Supreme Court’s new exclusionary rule: as a practical
matter, this new rule may only serve to confuse trial practice. For
instance, who has the burden of proving systemic problems—the
defense or the prosecution? When would the defense get access to
discovery regarding patterns of misconduct? How specific must the
patterns of misconduct be (e.g., is a pattern of unconstitutional stops
relevant to a case involving unconstitutional frisks)? Who would hire
the experts to interpret the data? What if police are deliberately
indifferent to bad practices? Given that this data represents
impeachment evidence, would it be subject to the disclosure
requirements of Brady v. Maryland?437 And at what level of generality
(local, city, or state) would the pattern need to be proven?438
Seemingly, in an effort to restrict the scope of the exclusionary rule
remedy, the Supreme Court created a test that overburdens trial
practice. Did the Court intend to turn every suppression hearing into
a Section 1983 proxy or a pattern and practice investigation? Did the
Court really want litigants to explore the systems that cause citizen-
police tension across the nation? The dissenting Justices in Herring
and Strieff make clear that this is the logical conclusion of the
holdings,439 yet no one seems to know how it would work in practice.
436. For example, allegations of police misconduct could be used to challenge the credibility
or veracity of officer testimony as the officers’ felt need to defend the constitutionality of police
actions could create an incentive to shade their testimony and thus create a form of bias cross-
examination.
437. 373 U.S. 83 (1963) (requiring prosecutors to disclose all exculpatory evidence to criminal
defendants prior to trial).
438. Thank you to Cynthia Conti-Cook for these insights and many more.
439. See supra notes 81–86, 433 and accompanying text.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
644 VANDERBILT LAW REVIEW [Vol. 72:2:561
Some might argue that the conservative majority of the
Supreme Court likely believed the task of proving such recurring
negligence near impossible, so it did not wrestle with the practicalities
of implementing such a requirement. But now, enhanced technological
capacities may have upended that plan by providing proof of recurring
patterns or systemic problems with relative ease. In fact, the incentive
exists for the entire defense bar to bring such challenges, because if
systemic or recurring negligence is shown in certain practices, then
such a finding will mean suppression in all related cases. For example,
if Salt Lake City police officers routinely conducted unconstitutional
stops for identification, all such cases involving that particular
practice would result in suppression. What began as a narrowing of
the exclusionary remedy might, in fact, turn out to be a much broader
mandate to expand court-overseen police accountability practices.
Every police stop will be analyzed in the context of a larger police
practice, with particular attention paid to the policies, practice, and
trainings of the police department at issue. This data collection will
also be useful for future civil rights cases and federal investigations.
This in turn raises a different, although related, problem: not
all jurisdictions will have equally sophisticated technology, leading to
unequal or divergent applications of blue data accountability. For
courts tasked with applying constitutional law and the exclusionary
rule uniformly across the legal system, this reality presents more
practical problems. Technology costs money and many jurisdictions
will not or cannot invest the time and capital into developing big data
policing strategies. There will then exist two tiers of policing systems:
the technology haves and the technology have-nots. The open question
will be what to do when big data becomes the preferred mechanism to
demand police accountability but does not exist in a particular
jurisdiction.
Again, these seem to be practical issues that the Court did not
think through in offering its exclusionary rule pronouncement in
Herring. Courts and lawyers resisting this change may well show the
need to rethink Herring’s lessons and to develop new ways to evaluate
which types of misconduct warrant suppression. In taking Herring
seriously, courts may realize that it does not work in practice. When
faced with resistance to the practice on the ground, courts may be
prompted to reevaluate the future of the exclusionary rule.
Ferguson_ PAGE (Do Not Delete) 3/31/2019 10:39 AM
2019] EXCLUSIONARY RULE IN THE AGE OF BLUE DATA 645
CONCLUSION
The idea of data-driven accountability is not new. But the
convergence of new surveillance and data-driven technologies along
with the Supreme Court’s requirement to demonstrate recurring or
systemic problems of police misconduct suggest a new way to visualize
the exclusionary rule in the age of blue data. A data-driven vision is
needed now more than ever, as other traditional means of police
accountability diminish due to a shifting political landscape. This
vision also fills the “proof gap” that exists when litigating police
misconduct. In ordinary suppression hearings, civil cases, and civil
rights investigations, this blue data will assist litigants and courts in
proving and visualizing large-scale police accountability projects. Even
if blue data accountability systems are not put in place, the push to
develop them will offer powerful clues about how citizens and police
react to new forms of big data surveillance. The reveal of the
resistance is that all people—citizens and police—have reason to be
concerned about growing big data surveillance systems.
Copyright of Vanderbilt Law Review is the property of Vanderbilt Law Review and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
