Page 15 - Case study: Crime prediction for more agile policing in cities – Rio de Janeiro, Brazil
P. 15

terrorist acts. The rapid roll-out of these tools invariably raise complex ethical questions in relation
            to police action and civil rights.

            Establishing standards and regulations: There are several considerations that law enforcement
            agencies would do well to consider before implementing predictive policing systems. First and
            foremost, departments must evaluate the quality of their crime data and the capabilities of officers
            and officials to perform such an evaluation. Specifically, it is imperative that crime underreporting
            and blind-spots are dealt with, and that departments can ensure that citizens from all
            neighborhoods and social groups have confidence - and make use of - the police emergency hotline
            to report crime when in need. Likewise, police reports must indicate precise addresses including
            their geographic coordinates. Also, to help mitigate the risk of runaway feedback loops, incidents
            must be labeled to indicate if they were reported by citizens, or if they were initiated by an on-duty
            officer while on routine patrol.

            Capital and operational expenditures associated with crime prediction tools must also be carefully
            assessed. The most sophisticated forecasting packages are expensive and may not be suitable for
            all police departments, especially smaller and mid-sized units in low- and medium-income settings.
            Instead of purchasing expensive software, some police departments may benefit more from hiring
            and training analysts  to use standard (and often open source) software to plot crime events on a
                                20
            map and run simple (yet useful) time series analysis.

            Policing innovations for agile security also should make use of the interconnection of urban
            infrastructure  including sensors and unstructured data. Even so, privacy concerns should be
                          21
            paramount in the decision to process such information. Where possible, predictive tools should
            allow citizens to understand what is inside the “black box”. While private vendors understandably
            seek to protect their source code, this lack of transparency (coupled with their underlying
            mathematical complexity) makes it difficult for law enforcement agencies and civil society to
            understand how the predictions are generated. This can undermine confidence in the tool.
            Complicating matters, the secrecy associated with predictive tools may subject departments to
            increasing legal liabilities as cybersecurity and privacy regulations continue to evolve.

            Above all, people must remain the most important element in the crime forecasting process,
            even when the most advanced software packages are used. Predictive tools need not replace the
            intuition and experience of law enforcement officers, but rather complement them in a transparent
            and auditable manner. When responsibly implemented, predictive policing tools can improve
            law enforcement's capabilities to solve problems, make decisions, and more effectively plan their
            operations.


















                                                          Crime prediction for more agile policing in cities – Rio de Janeiro, Brazil  9
   10   11   12   13   14   15   16   17   18   19   20