Get out of jail or do more time? Risk tools help judges decide.
Connecting state and local government leaders
Former New Jersey attorney general leads an effort to develop risk-assessment tool to help judges make data-driven sentencing decisions.
How does a criminal court judge decide whether a given defendant is likely to commit another crime or become violent if released? Mostly by a combination of guess work and criminal histories, according to former New Jersey attorney general Anne Milgram, who was shocked to find out no tool was available to make an objective risk assessment of such cases when she joined the office in 2007.
“We were essentially trying to fight crime with yellow Post-It Notes,” said Milgram, in a January TED Talks presentation. So she helped lead an effort to “let judges understand in a scientific and objective way what the risk was posed by someone in front of them.”
To do so, Milgram formed a team of data scientists, researchers and statisticians who collected 1.5 million court cases from around the country — “the largest pretrial data set in the U.S.,” she said. From that database, the researchers extracted 900 risk factors from which they identified a handful of data points that were most predictive of risk.
Those data factors, including whether a defendant had committed an act of violence when released before, formed the basis of a “universal risk-assessment tool,” a dashboard that, “every single judge can use because it’s been created on a universal data set,” according to Milgram, who subsequently joined the Laura and John Arnold Foundation, which developed the tool that became called the Public Safety Assessment-Court (PSA-Court).
“Giving judges data-driven, analytic tools to assist them in making pretrial decisions can create safer communities, save taxpayer dollars and make our criminal justice system fairer,” said Milgram in announcing the tool.
In a research brief, Developing a National Model for Pretrial Risk Assessment, the foundation reported that jurisdictions that use data-driven pretrial risk assessments (less than 10 percent of jurisdictions) spend less on pretrial incarceration and enhance public safety.
Using data analytics, New Jersey has been able to take Camden off the list as the most dangerous city in America, reducing the homicide rate by 41 percent and crime in the city by 26 percent, said Milgram. The city also changed the way it conducted prosecutions, shifting the focus from low-level drug crimes to a greater focus on crimes of statewide importance, such as prosecuting street gangs, gun and drug trafficking and political corruption, she said.
Elsewhere, prisons are using data analytics in parole decisions to help determine if an inmate is likely to commit a crime when released. At least 15 states now require prison systems to use some sort of risk-assessment tool to evaluate inmates for parole.
Police officers are starting to use data analytics to both solve and prevent crimes. But the ever-increasing amount of data available, including that from social media networks, video footage and geographic profiles, increases the complexity of gathering and integrating data so it can be analyzed.
That’s why training the legal and law enforcement communities to use the tools remains essential. “New analytic technologies will be most effective if used within a force that has structured itself to exploit technological benefits, trained its officers in how to use the technology and provided appropriate tools for them to access the technology when on the move,” noted InformationWeek’s columnist Wai-Ming Yu.