Predicting Future Crime

The Philadelphia Police Department is experimenting with an advanced computing system that statisticians and criminologists believe can predict future violent crime with some accuracy. Using a learning computer, the city is working to predict which of the city's 49,000 parolees are most likely to commit violent crimes in the future. While this system raises some serious red flags, it also has potential to improve the efficiency and fairness of our parole system if it is used with restraint.

On one hand, people are not their rap sheets and they deserve individual attention after release from prison. They change and grow and learn and that's why we have thousands of parole officers in this country - to meet with parolees and gauge their progress and potential for successful reintegration. On the other hand, this isn't always the case in practice. Parolees are sometimes treated as numbers. A failed drug test or a driving violation could stop a non-violent parolee's progress in its tracks and send them back. Perhaps we should focus this limited attention on those more likely to commit future crime?

We have a human system and it can be susceptible to the predjudices and mistakes of all of us. Adding an objective computer prediction to the mix - as long as it is kept confidential and not given undue weight - could add fairness to the parole system. The experimental results from Philly seem promising:

To "train" the system, (University of California statistician Richard) Berk fed in data on 30,000 past cases; about 1 percent had committed homicide or attempted homicide within two years of beginning probation or parole.

The data included the number and types of past crimes, sex, race, income, and other factors.

To test its power, he fed in a different set of data on 30,000 other parolees. This time he didn't tell the computer who would go on to kill.

Applying what it had previously learned, the system identified a group of several hundred who were considered especially dangerous. Of those, 45 in 100 did commit a homicide or attempted homicide within two years - much higher than the 1 in 100 among the general population of probationers and parolees.

The predictors that mattered most were age, age at first contact with adult courts, prior crimes involving guns, being male, and past violent crimes.

But there are major concerns about the new system: it could lead us to spend too much energy on "high-risk" parolees and ignore the needs of "low-risk" people, denying them opportunities for support and services.  And would an increased reliance on this system mean more minor technical violations for parolees considered high-risk?

"The main ethical concern," said Richard Bonnie, a law professor at the University of Virginia, "is the possible unfairness to the 'selected' offenders."

If the high-risk people do get more supervision, it means they face a greater risk of being caught in a technical violation that will send them back to prison. Should such power be relegated to a computer?

While this is a worry, the system currently sends thousands parolees back to prison for violations, and the chance of being caught seems fairly arbitrary. Perhaps it does make more sense to focus on the violations of parolees with violent histories. Another red flag: officials considered including race in the equation - an extremely scary thought - but decided against it. I wonder if there are other questionable factors included in the computer's calculations.

I worry that the use of this system would be abused and take on weight greater than individual, human assesments. But is there a place for it in our system? Would it make the system more fair and more targeted? The jury's out.