The Florida State Department of Juvenile Justice is implementing predictive analytics to help identify at-risk youth in an attempt to reduce recidivism in state’s penal system.
Using a technology made familiar by Amazon and NetFlix’s personalized recommendations, browsers have become accustom, even demand, to see similar books and movie titles. Through association algorithms or by looking at a user’s click stream – the sequence of pages viewed from browsing to buying – the movie and e-commerce giants hope to predict what the user might buy next.
Traditionally these tools have been adopted by retailers or merchants in order to make more sales. But more recently, predictive analytics is being leveraged by the public sector to provide better services and produce more efficient outcomes. By looking at historical data and using statistical algorithms, state and local governments can transform their seemingly infinite amount of latent data into meaningful, actionable information.
At the Florida State Department of Juvenile Justice, and in other segments of the detention system across the country, they are hoping to better understand and treat causes of recidivism, especially in young adults. The Florida Department of Juvenile Justice (DJJ) announced Wednesday a partnership with SPSS, an IBM company, to identify high-risk youth and prescribe individualized rehabilitation programs to keep them from re-entering the system as adults.
CivSource spoke to Bill Haffey, predictive analytics strategist for the public sector at SPSS, about how DJJ is leveraging the technology and what kinds of predictors could indicate high and low risk offenders.
“At the Department of Juvenile Justice, their interest has always been the same – to identify offenders who are likely to accelerate, or graduate, into the adult court system,” Mr. Haffey said. Generally, the juvenile justice system in the US has takes a uni-dimensional analysis. “If a juvenile was in the system two or three times,” Mr. Haffey said, “there was a knee-jerk reaction to ‘watch out’ for that person. But there are other pieces of information that haven’t been fully exploited.”
With the new IBM SPSS software, DJJ will move beyond Excel spreadsheets to analyze key predictors such as past offense history, home life environment, gang affiliation and peer associations, predicting which youths have a higher likelihood to reoffend. But Mr. Haffey said it is often hard to look at any one of those factors because individually they mask the larger effects of their combined influence.
To paint a complete picture, and to understand who is more or less likely to reenter the system, Haffey said the technology could be used prescriptively and be built in as an operational process tool. Once you look at the juvenile offenses, and the various factors that might impact those offenses, you can begin to make sense of what is causing them, Haffey said. Over time, trends and patterns can be incorporated during the intake process, and indicators – such as education levels, parental status, working status, and other types of non-crime related information – can be combined.
“[This information] can then predict the level of risk for an individual, and based on that assessment, a program can be designed.”
DJJ officials plan to do exactly this. In a blog posting, Mark Greenwald, chief of research and planning at DJJ, said he hopes to improve existing screening and placement processes with evidence-based interventions. “We can [then] direct youth toward treatment that will address their specific criminogenic needs,” Greenwald wrote. “This gives us the opportunity to place individuals in specific programs, such as combating substance abuse or addressing mental health issues, creating personalized – versus generic – rehabilitation programs.”
This approach is not unique to DJJ, though. Mr Haffey suggested there are a number of county and state parole boards that are also interested in using predictive analytics to assess risk of their parolees. Similar to the DJJ example, parole officers ask parolees specific questions, which are fed into a predictive model to decide how often a parolee should be required to check-in.
More than 85,000 youth enter the juvenile justice system in Florida. Moving forward with predictive analytics, officials said they hope the first time is the last time.