Predictive Policing: How Data Can Create Its Own Crime Map

In 2015, the Philadelphia Police Department began testing HunchLab, a predictive-policing system. The city was divided into squares measuring 500 feet on each side. Every day, the software used old crime data, calculated the expected number of incidents in each cell, and highlighted the cells with the highest scores so officers could patrol them. The goal sounded reasonable: if the system could identify high-risk places, police could arrive before a crime occurred.

In 2015, the Philadelphia Police Department began testing HunchLab, a predictive-policing system. The city was divided into squares measuring 500 feet on each side. Every day, the software used old crime data, calculated the expected number of incidents in each cell, and highlighted the cells with the highest scores so officers could patrol them. The goal sounded reasonable: if the system could identify high-risk places, police could arrive before a crime occurred.

In 2015, the Philadelphia Police Department began testing HunchLab, a predictive-policing system. The city was divided into squares measuring 500 feet on each side. Every day, the software used old crime data, calculated the expected number of incidents in each cell, and highlighted the cells with the highest scores so officers could patrol them. The goal sounded reasonable: if the system could identify high-risk places, police could arrive before a crime occurred.

In 2015, the Philadelphia Police Department began testing HunchLab, a predictive-policing system. The city was divided into squares measuring 500 feet on each side. Every day, the software used old crime data, calculated the expected number of incidents in each cell, and highlighted the cells with the highest scores so officers could patrol them. The goal sounded reasonable: if the system could identify high-risk places, police could arrive before a crime occurred.