A system of predicting future criminality by Philadelphia probationers was correct nearly two thirds of the time, the National Institute of Justice’s annual crime-research conference near Washington, D.C., was told yesterday. The model produced an accurate forecast for 79,299 of the 119,935 probation case starts in the sample. The model now being used, designed by criminologists at the University of Pennsylvania’s Jerry Lee Center of Criminology, predicts whether probationers will commit more crimes by using a dozen variables, including their age, residential zip code, criminal record, and history of sanctions imposed on them.
The system was conceived in 2005, and various versions have been tested since then. The aim is to have the city’s probation officers tailor their caseloads to the risk level of probationers, with those handling high-risk people having smaller caseloads and those supervising low-risk criminals having larger caseloads. Researchers concede that no prediction method is perfect. One problem with Philadelphia’s is that there are thousands of false positives–people placed into the high-risk supervision category even though they do not commit repeat crimes. Still, in a paper issued by NIJ, the researchers conclude that although their “techniques may give rise to important questions of ethics and justice, they also represent an opportunity to advance the capabilities of the criminal justice system to protect communities.”