WHAT WE DO
The Public Safety Lab uses the tools of data science to promote cost-effective public safety, with an awareness of both resource and social costs. We work with communities and law enforcement agencies to design, implement, and test analytic solutions that meet jurisdictions’ needs.
Communities and agencies interested in working with the Public Safety Lab can contact us at email@example.com.
WHO WE ARE
The Public Safety Lab coordinates the work of researchers across multiple disciplines and universities.
Assistant Professor of Management Science and Engineering
Assistant Professor of Computer Science
Professor of Law
Southern Methodist University
Director, Deason Criminal Justice Reform Center
Associate Professor of Information Systems
Carnegie Mellon University
Visiting Professor of Urban Analytics
New York University
Public Safety Lab Data Science Team
Public Safety Lab Doctoral Students
Ph.D. and J.D. Student
Department of Economics
University of Pennsylvania
New York University School of Law
The application of data science to public safety may offer opportunities to achieve better outcomes. For example, many relatively minor offenses are easy to detect and interdict. Many more serious offenses are much harder to detect and interdict. As a consequence we may overpunish relatively minor offenses, and underpunish more serious offenses.
In our projects we work with jurisdictions to promote socially productive allocations of public safety resources.
911 Response Analytics
Calls to 911 in large urban jurisdictions are first assigned to a responding service (e.g., police, fire, EMT), and are then assigned priority codes determining the speed and nature of the service's response. These human-assigned call and priority codes may not be optimal. For example, many calls assigned to policing agencies might more appropriately be assigned to medical or social service professionals. Calls assigned to policing agencies might result in better outcomes if their human-assigned priority codes incorporated knowable information about the location or phone number, or if their codes were purged of any call taker biases. Working with a large urban policing agency, we are exploiting the random assignment of 911 call takers to calls to evaluate the causal effect of call and priority codes on call outcomes, and are developing and testing an analytic to assign more optimal call and priority codes to 911 calls.
Human Trafficking Analytics
Sex trafficking is a crime of extraordinary violence visited largely upon vulnerable girls and young women. Its frequency is likely severely underreported, given the isolation within which its victims are kept. We are developing and implementing a human trafficking analytic that predicts instances of trafficking from the corpus of online commercial sex ads and provider reviews, using an extensive set of verified instances of trafficking sourced from multiple law enforcement agencies. The analytic then pushes predicted risk scores to agencies for investigation through a searchable database of ads and reviews; validated outcomes are used to further refine the prediction model. We are also exploring the responsiveness of sex trafficking to local labor markets and interdiction efforts.
Traffic Fatality Analytics
Numerous studies have suggested that increased officer deployment reduces both speeds and serious traffic accidents, including fatal accidents. Yet the institutional incentives affecting the spatial and temporal patterns in deployment have received little attention; these incentives may result in unintended consequences for roadway safety. Working with a large regional policing agency, we are using a series of policy discontinuities to explore the consequences of institutionally-induced incentives on traffic fatalities. Preliminary findings indicate that relatively obscure budgeting and revenue institutions can significantly impact traffic fatality rates.
Prosecutorial Reform Analytics
In many jurisdictions, offenders who commit relatively minor violations are arrested and prosecuted to the full extent of the criminal law. In these jurisdictions, cases involving relatively minor offenses typically consume a large share of law enforcement, prosecutorial, and judicial resources. Yet is is unclear whether this practice reduces offender recidivism; the practice may actually result in increased recidivism, and/or the escalation of minor offending into more serious offending. We are investigating the impact on offender recidivism of recent exercises of prosecutorial discretion to reduce the incidence of arrest and prosecution for relatively minor offenses.
Criminal Defendant Analytics
Ensuring that criminal defendants are treated equally before the law is one of the most important imperatives of criminal justice institutions. It can be a challenging question to investigate empirically, however, due to data limitations and problems of causal inference. Working with extensive data crawled from public access websites, including the text of every criminal opinion issued by New York State's appellate courts between 2003 and 2017, and demographic data on all criminal defendants processed by New York State's Department of Corrections since the 1970s, we are building predictive models of criminal trial and appeal dispositions.
Civil Asset Forfeiture Analytics
Civil asset forfeiture allows law enforcement agencies to seize property upon the suspicion of criminal activity. While the practice may deter some criminal activity, it may also divert agency resources away from other productive pursuits, and may impose economic costs unevenly. We are collecting data on the amounts and types of assets being seized and forfeited by state and local law enforcement agencies, whether forfeitures are contested, demographic data on the subjects of asset forfeiture, incident-level data on law enforcement behavior, and cross-state and within-state variation in civil asset forfeiture statutes. Using these data we are building predictive models of the incidence and effects of civil asset forfeiture.
911 Response Feedback Survey
Patrol officers spend most of their time responding to calls for service. Yet callers to 911 often don't observe the law enforcement response to or resolution of their calls, possibly contributing to perceptions that their calls did not receive the attention or effort they deserved. At the same time, law enforcement agencies generally don't observe community perceptions of their 911 call response. Working with a large urban policing agency, we are conducting a randomized controlled trial of a platform that pushes an SMS-based survey to recent 911 callers, asking them to provide feedback on their experience of the police response.
Social Networks and Violent Crime
Recent work has suggested that violent criminal behavior is often perpetrated through closely knit social networks, networks that facilitate overlapping forms of crime often associated with high rates of gun violence (e.g. drugs, guns, sex trafficking). This work suggests that identifying social networks with the potential to either encourage or discourage violent criminal behavior may be a promising policing strategy. Working with two large urban police forces in South America, we are analyzing whether randomized police interventions delivered through socially central individuals can reduce participation in violent criminal activity.
Although "community policing" has become a popular catchphrase, little is known systematically about whether community policing can improve communities' trust in their local law enforcement agencies. Working with a large urban police force in South America, we are analyzing whether randomly assigned town hall meetings bringing together residents and officers to discuss policing strategies can improve trust in law enforcement.
Domestic Violence Policing
Law enforcement agencies in developing countries often fail to appropriately respond to domestic violence complaints. Working with a state policing agency in India, we are conducting a randomized controlled trial of several interventions designed to increase the efficacy of the police response to domestic violence complaints, including randomized officer training in alternative dispute resolution and random assignment of paralegals to assist in case development.
Crowdsourcing Public Safety
Community residents may often have valuable information about recently committed crimes, information that is never shared with their local policing agency. Working with a large urban police department, we are developing and piloting a mobile application that will allow the community to partner with the department to co-produce public safety. Through the app, criminal investigators will be able to communicate with civilians near the location and time of day of recently committed crimes, with the capacity to solicit and receive video, photographic, or textual information. We will then test whether this increased information flow increases case clearances, deters criminal activity, and/or increases neighborhood trust in law enforcement.
Workshop on Data-Driven Criminal Justice Reform
Friday October 19th 2018 9AM-5PM
Saturday October 20th 2018 9AM-1PM
Lester Pollack Colloquium Room
Furman Hall, New York University School of Law
The application of data science to criminal justice questions may offer opportunities to acheive better public safety outcomes. For example, many relatively minor offenses are easy to detect and interdict. Many more serious offenses are much harder to detect and interdict. As a consequence we may overpunish relatively minor offenses, and underpunish more serious offenses. Data science strategies may allow us to achieve more socially productive allocations of criminal justice resources.
This workshop will convene university researchers along with representatives from public agencies, nonprofit organizations, and private firms to engage in a broad-ranging conversation on data-driven criminal justice reform. Workshop details and logistics will be provided at a later date.
Space is limited and registration is required; please contact us at firstname.lastname@example.org to receive an RSVP link.