Editor’s note: This is the first of a two-part blog post examining the potential impact of data visibility on law enforcement.
The Counted, Fatal Force, and Mapping Police Violence websites each collect, store, and display data about people killed by police in the United States. These websites are just a few of the emerging platforms designed to address the significant gap in information left by US police organisations’ failure to create, maintain, and publically disclose data about “fatal force” incidents. When visiting any of the three websites mentioned above, visitors can access in-depth statistics, charts, graphs, and maps, which provide details about the number of fatal force incidents that have occurred, their locations, the identity of officers involved, and the demographics of victims. The availability of this information has solicited questions about if and how digital data can address persistent problems related to a lack of transparency and accountability in policing, and the lack of information about fatal force incidents:
- Can data enable new opportunities to scrutinize fatal force incidents?
- Can data provide an opportunity to discover trends associate with fatal force incidents?
- Can data analysis provide the police with the knowledge required to reduce fatal force incidents?
This two-part blog focuses on the last question by considering the opportunities and limitations of using digital data to monitor police work, document fatal force incidents, and create intervention programs designed to reduce fatal force incidents.
Police Visibility and Dataveillance
Recent literature suggests that police officers are among the most extensively monitored subjects in today’s surveillance society. To understand the nature of the surveillance of police officers, scholars have built a taxonomy which splits “police visibility” into three subcategories entitled primary, secondary, and new visibility. Unfortunately, this taxonomy focuses exclusively on the visual contributions to police visibility made by cameras. To understand the contributions which digital data make to police visibility, I propose expanding the taxonomy of police visibility by exploring how non-visual forms of surveillance are used to monitor police. Examples of non-visual surveillance targeting police include “dataveillance” mechanisms which collect data about police officers’ emails, radio communications, internet browsing history, movements, as well as information from police databases documenting traffic stops, fines, arrests, use of force incidents, and more. Because of the variety of ways that dataveillance can be used to monitor police officer, officers can be described as having a high “data visibility,” referring to the perceptibility of police officers and police work because of the production and brokering of associated data.
To nuance this blog’s conversation about the police’s data visibility, I focus on the impact of dataveillance mechanisms which document fatal force. More specifically, I focus on if and how dataveillance can be used to analyse and reduce fatal force incidents. Fatal force is given particular attention due to high rates of fatal force incidents in the US (documented in this Amnesty International report), as well as my interest in human rights concerns relating the balance between police powers and the right to life, liberty & security of person, privacy, and the prohibition of discrimination.
How Can Digital Data Reduce Fatal Force Incidents?
Digital data can be used to reduce fatal force incidents in at least two ways. The first method involves using digital data to study fatal force incidents and make related changes to policy and practice. Studying fatal force is made possible once brokers (including police organisations, journalists and “citizen journalists”) collect, share, and display data about fatal force incidents, creating opportunities for long-form analysis, including the search for patterns in the data. For instance, by enabling the filtering of data, websites like Fatal Force, the Counted, and MappingPoliceViolence allow users to discover patterns such as the consistency with which incidents of police shootings involve males, the high number of incidents in which mental illness plays a role, or the consistency with which the deceased were carrying firearms. Knowledge about patterns in fatal force incidents can then be used to inform changes to use of force policies, and/or to improve training concerning interactions with persons who are mentally ill or interactions with individuals who are armed.
The second and more proactive method of using digital data to reduce fatal force incidents involves predictive analysis and the pre-emption of fatal force. For example, based on the analysis of fatal force data, data scientists may discover early warning signs (early warning signs might include poor performance during training or high numbers of complaints related to use of force) which consistently seem to precede fatal force incidents. If data scientists’ findings are reported to police organisations, those organisations can search for officers who display early warning signs and flag them as high-risk. Police organisations can then initiate early intervention programs which may include re-training or counselling high-risk officers, or perhaps temporarily reassigning high-risk officers to positions in which interactions with citizens is unlikely. Unlike the primary, secondary, and new visibility, the police’s data visibility may, therefore, provide the opportunity to not only document fatal force, but predict and pre-empt it. Note that early intervention programs would not imply that fatal force is always egregious. Rather, intervention program would be an effort to reduce fatal force incidents as much as possible while also protecting fatal force interventions that are deemed necessary and unavoidable.
While the predictive potential of data about police work may sound fantastical to some, recalling images of Minority Report, data scientists are already testing algorithms designed to predict use of force incidents in the US. For example, data scientists at the University of Chicago’s Centre for Data Science and Public Policy have developed a prototype algorithm, based on data provided by the Charlotte-Mecklenburg Police Department (CMPD), to identify early warning signs that may predict adverse interactions between police and the public. Early tests show that the Centre’s algorithm has successfully identified officers who were later involved in adverse interactions with the public. Based on successful tests, data scientists have already translated their research into Flag Analytics, a company designed to commercialise the Centre’s algorithm and make it available to police organisations across the United States. Similar studies are exploring how data can be used to document and predict the effective use of stop-and-search powers to limit discriminatory decision making.
The above discussion demonstrates that there may be a reason to be optimistic about the implications of the police’s growing data visibility for analysing and reducing fatal force incidents. However, as part 2 of this blog will suggest, there are also several limitations to be aware of before declaring data visibility a panacea to problems related to high rates of fatal force incidents.
Disclaimer: The views expressed herein are the author(s) alone.