Parole Boards Using Software to Predict Recidivism!!

Here in New Orleans we have the fabulous New York based Vera Institute of Justice. They came up with the inmate risk assessment ranking system that our city council funds. It’s a controversial program because it takes money out of the bail bond system. I think it’s a good program but needs time to work. I’m not really concerned about the bail bondsmen making a few less bucks.

Now, there’s software that predicts if a prisoner should be paroled. The program is called Compas. From the Reentry Policy Council-

The Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) system is a statistically based risk and needs assessment specifically designed to assess key risk and needs factors in adult and youth correctional populations and to provide decision-support for justice professionals who must make decisions regarding the placement, supervision, and case-management of individuals in community supervision and correctional institution settings. It achieves this by providing valid measurement and succinct organization of research supported risk/need dimensions. COMPAS scores each individual based on three different types of risk (violence, recidivism, and failure to appear in court) and 19 different criminogenic needs. The software also includes case planning, outcomes measurement, and reports generation modules. The internal Research Division (staffed by five PhDs) and IT Division provide the research and technical support to norm the assessment for the local population and configure the software to local policy and procedure. The time required to administer each battery of tests varies, and can be adapted to the needs of the jurisdiction. A peer reviewed validation study of the COMPAS has been accepted by Criminal Justice and Behavior for publication in the June 2009 edition. An additional independent validation of the COMPAS in a California study by Zhang and Farabee (2007) indicated predictive accuracies comparable to other major instruments.

Recidivism Revolving Door 2

Recidivism Revolving Door

Some of the software conclusions defy conventional correction officers’ logic- some violent offenders are categorized as lower recidivism risks than non violent inmates. For example, inmates convicted of murder serve long sentences and are paroled at older ages, making them less likely to re-offend. Non violent criminals serve shorter sentences and are released at younger ages, making them more likely to commit crimes when freed. I guess this makes sense.

Some states are claiming that this software works for them by lowering the rate of readmission for released prisoners. Texas, Michigan and Ohio are releasing¬† more and more parolees each year and less are going back to jail. This is all a good thing- the public is safe as the parolees are committing less crimes, and the states’ save money by having less folks in the pokey.

Ohio Recidivism Chart

Ohio Recidivism Chart

~ by neworleansmusicman on October 22, 2013.

2 Responses to “Parole Boards Using Software to Predict Recidivism!!”

  1. I would not be in favor of using this software to predict outcomes for potential parolees in Louisiana because our state is so different from so many other jurisdictions. For example, depending on WHEN one commits a crime in Louisiana, the percentage of the sentence that the offender must spend incarcerated before being released under Louisiana’s good-time laws may vary. If you compare all burglary offenders, for example, you may find that some of them have received 15 days per month of good-time, while others have received 30 days per month in good-time credits, and others may receive 45 days of good-time credit on their sentences. If they attend and participate in educational or vocational programs – that is, if they are lucky enough to get into such programs -they can receive additional good-time credits. All prisoners who are eligible to receive good-time credits are not treated the same. Everyone does not enjoy the same opportunities. Essentially, there can be several rates of good-time credits applied to convicted and sentenced burglars who may all be incarcerated at the same time. This inequity creates problems, increases recidivism, and destroys the incentive to turn one’s life around and remain crime-free once the offenders are released. An article that I wrote about this appeared in the Chainlink Chronicle in 1993. It was entitled “Good-Time and Parole in Louisiana.” Twenty years later, not much has changed, except perhaps, that some of the inmates who were incarcerated in 1993 may have been released since then. The disparity is not limited to burglary offenses. That crime was chosen randomly to make a simple point. Relying on computer generated factors, including recidivism risk factors, is a bad idea. It may be nothing more than just another way to enrich the folks who sell the idea and the software to the states. It is a mistake to use the work of the Vera Institute in New Orleans to promote computer generated parole decisions. Vera Institute’s work is flawed beyond fixing. Nothing can replace a Judge’s own assessment of the bailbond risks of criminal offenders. Prosecutors and defense attorneys can inform the judge of everything (and more) that he/she needs in order to set a fair amount of bail. Paying the Vera Institute a ton of money to make their often terrible and irrational recommendations is a mistake. Nothing beats hands-on interaction with the offender, looking at one’s record, his/her residency, family support, employment history, and other factors in order to make rational decisions. We have to trust the people we put in place to do the right thing. I’m sure that everyone in New Orleans have heard about the low risk assessment that Vera provided for a young man who ultimately shot nearly two dozen people shortly thereafter.

    • Thanks for the detailed comment, Lynell. You bring up a lot of good points. I would like to see a combination of new and old techniques for the best results. Vera grades many of the factors you attribute to hands-on interaction.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: