Big data and predictive reasonable suspicion.

• Author: Ferguson, Andrew Guthrie
• Position: Introduction through II. The Rise of Big Data Policing C. Predictive Data, p. 327-373

INTRODUCTION I. REASONABLE SUSPICION: A SMALL DATA DOCTRINE A. The Reasonable Suspicion Standard B. Reasonable Suspicion in Application 1. Unknown Suspect Cases 2. Some Information on the Suspect 3. Known Suspects C. Concluding Thoughts II. THE RISE OF BIG DATA POLICING A. Big Data: An Introduction B. The Growth of Data Collection 1. Volume of Data 2. Networked Data 3. Identifiable Data C. Predictive Data D. Unprotected Data III. BIG DATA AND REASONABLE SUSPICION ON THE STREETS A. Observation of Ongoing or Imminent Crimes B. Investigation of Completed Crimes C. Predicting Crimes D. Big Data Suspicion IV. EVALUATING BIG DATA AND PREDICTIVE REASONABLE SUSPICION A. The Positives of Big Data Suspicion 1. Improved Accuracy 2. Exculpatory Facts 3. Accountability 4. Efficiency 5. Unexpected Insights B. The Negatives of Big Data Suspicion 1. Bad Data 2. False Positives 3. Shifting Power Balance V. TENTATIVE SOLUTIONS A. Doctrinal Changes B. Big Data Changes 1. Statistical Analysis 2. Precision of Place and Time 3. Limited Link Analysis CONCLUSION [F]rom a legal point of view there is nothing inherently unattainable about a prediction of future criminal conduct. (1)

Electronic databases form the nervous system of contemporary criminal justice operations. In recent years, their breadth and influence have dramatically expanded.... The risk of error stemming from these databases is not slim.... Inaccuracies in expansive, interconnected collections of electronic information raise grave concerns for individual liberty. (2)

INTRODUCTION

The Fourth Amendment requires "reasonable suspicion" to stop a suspect. (3) As a general matter, police officers develop this suspicion based on information they know or activities they observe. Suspicion is individualized to a particular person at a particular place. (4) Most reasonable suspicion cases involve police confronting unknown suspects engaged in observable suspicious activities. (5) Essentially, the reasonable suspicion doctrine is based on "small data"--discrete facts, limited information, and little knowledge about the suspect. (6)

But what happens if this small data suspicion is replaced by "big data" suspicion? (7) What if police can "know" personal information about the suspect by searching vast networked information sources? The rise of big data technologies offers a challenge to the traditional paradigm of Fourth Amendment law. With little effort, officers can now identify most unknown suspects, not through their observations, but by accessing a web of information containing extensive personal data about suspects. (8) New data sources, including law enforcement databases, third-party records, and predictive analytics, combined with biometric or facial recognition software, allow officers access to information with just a few search queries. (9) At some point, inferences from this personal data (independent of the observation) may become sufficiently individualized and predictive to justify the seizure of a suspect. The question this Article poses is whether a Fourth Amendment stop can be predicated on the aggregation of specific and individualized, but otherwise noncriminal, factors.

For example, suppose police are investigating a series of robberies in a particular neighborhood. Arrest photos from a computerized database are uploaded in patrol cars. Facial recognition software scans people on the street. (10) Suddenly there is a match--police recognize a known robber in the targeted neighborhood. The suspect's personal information scrolls across the patrol car's computer screen--prior robbery arrests, prior robbery convictions, and a list of criminal associates also involved in robberies. (11) The officer then searches additional sources of third-party data, including the suspect's GPS location information for the last six hours or license plate records which tie the suspect to pawn shop trades close in time to prior robberies. (12) The police now have particularized, individualized suspicion about a man who is not doing anything overtly criminal. Or perhaps predictive software has already identified the man as a potential reoffender for this particular type of crime. (13) Or perhaps software has flagged the suspect's social media comments or other Internet postings that suggest planned criminal or gang activity. (14) Can this aggregation of individualized information be sufficient to justify interfering with a person's constitutional liberty?

This Article traces the consequences of a shift from "small data" reasonable suspicion, focused on specific, observable actions of unknown suspects, to a "big data" reality of an interconnected, information rich world of known suspects. With more specific information, police officers on the streets may have a stronger predictive sense about the likelihood that they are observing criminal activity. (15) This evolution, however, only hints at the promise of big data policing. The next phase will use existing predictive analytics to target suspects without any firsthand observation of criminal activity, relying instead on the accumulation of various data points. (16) Unknown suspects will become known to police because of the data left behind. (17) Software will use pattern-matching techniques (18) to identify individuals by sorting through information about millions of people contained in networked databases. This new reality simultaneously undermines the protection that reasonable suspicion provides against police stops and potentially transforms reasonable suspicion into a means of justifying those same stops.

This Article seeks to offer three contributions to the development of Fourth Amendment theory. First, it demonstrates that reasonable suspicion--as a small data doctrine--may become practically irrelevant in an era of big data policing. Second, it examines the distortions of big data on police observation, investigation, and prediction, concluding that big data information will impact all major aspects of traditional policing. Third, it seeks to offer a solution to potential problems using the insights and value of big data itself to strengthen the existing reasonable suspicion standard.

Part I of this Article examines the development of Fourth Amendment law on reasonable suspicion. Much of this case law involves "unknown" suspects, such as when a police officer sees an individual on the street but does not know his or her identity. In these cases, reasonable suspicion necessarily derives from the suspect's observable actions. Most Fourth Amendment cases involving police-citizen encounters are of this "stranger" variety. (19) Thus, the reasonable suspicion test, as it evolved, required the police officer to articulate individualized, particularized suspicion to distinguish a stranger's suspicious actions from non-suspicious actions. (20) The resulting doctrine, created around actions, not individuals, makes sense within the context it arose (as presumably most officers would not know all of the potential criminals in their patrol areas). (21) The resulting reasonable suspicion test, however, becomes significantly distorted when officers have access to more individualized or predictive information about a suspect.

Part II of this Article addresses the rise of "big data" in criminal law enforcement. Law enforcement organizations are working to grow the scope, sophistication, and detail of their databases. (22) Agencies and their officers may now search national databases and gain instant access to the information. (23) Indeed, "data" is the new watchword in many smart-policing districts. (24) Crimes are recorded. (25) Criminals are cataloged. (26) Some jurisdictions record data about every police-citizen encounter, making both the person and justification for the stop (not necessarily even an arrest) instantly available to any officer. (27) Some jurisdictions have compiled "bad guy lists" identifying suspects in a neighborhood based on computer analysis of past actions and arrests. (28) In addition, law enforcement agencies increasingly rely on predictive algorithms to forecast individual recidivism and areas of likely criminal activity. (29)

Just as law enforcement agencies now collect and electronically analyze more personal data, so do private, third-party organizations. (30) These third-party entities are a familiar part of our daily lives. "Smartphones" record where we go. (31) Credit card companies record what we buy, and banks chronicle what we spend. (32) "OnStar" systems in cars catalog where and how fast we drive. (33) Phone records reflect our contacts and communications. (34) Internet searches reveal what we read and expose our interests. (35) Social media sites, such as Twitter and Facebook, even disclose what we think. (36) Currently, law enforcement officers may access many of these records without violating the Fourth Amendment, under the theory that there is no reasonable expectation of privacy in information knowingly revealed to third parties. (37) While certain statutory protections exist, most statutes include law enforcement exceptions, (38) and in any case, these private, commercial data aggregators have turned personal data into a commodity, available for purchase and analysis to anyone willing to pay. (39)

The rise of "big data" means that this information is potentially available for use by law enforcement. In the same way that a drug store can predict that you will need a coupon this month because you bought a similar product last month, (40) the police will be able to anticipate that you will be selling drugs this week because you purchased an unusual number of mini-plastic bags last week. (41) Neither prediction is necessarily accurate, but both are based on individualized and particularized data that makes the prediction more likely.

Part III analyzes the intersection of big data and the current Fourth Amendment framework. The wrinkle of big data is that now officers are no longer...