It seems like a Philip K. Dick novel brought to life. New software developed by the University of Pennsylvania uses algorithms to predict criminal behavior. According to MSNBC, researchers aim to reduce the murder rate by predicting whether "individuals on probation or parole are ... likely to murder and to be murdered."

According to the article, a further application of the software being rolled out in Washington, D.C., will look at suspects' probability of committing other crimes and "could influence sentencing recommendations and bail amounts." The article discusses how the software attempts to objectively evaluate what used to be a judgment-based process.

According to the MSNBC article, murder is a very rare crime, occurring in only one of 100 people even in high-risk populations. Theoretically, the UPenn software could help to find the person most likely to commit the crime.

Researchers analyzed data from 60,000 crimes and identified traits in people likely to commit murders. Factors like age, criminal history, and location reveal information like how much supervision someone should have on parole, for instance. According to the article, "the UPenn researchers could identify eight future murderers out of 100."

The article quotes SUNY Albany criminal justice professor Shawn Bushway, who calls the software scientifically "very impressive," but the article also says policy makers are nervous about labeling groups of people as high-risk.

Can software predict criminal behavior?
Algorithm software to be rolled out in Washington, D.C., could assist parole boards and help set sentencing recommendations.