– Imagine a police department that doesn't just lock people up after they steal, shoot or rape, but instead uses predictive analytics to target young people expected to become adult criminals — and tries to prevent it.

If it sounds a little sci-fi for a Minnesota city, authorities here acknowledge that their proposal, expected to focus on about 30 youths, is unusual and comes with a host of challenges.

Residents have already raised questions over whether the move could lead to racial profiling and numerous legal and logistical complications, but Police Chief Roger Peterson is defending the project as an innovative way to diminish crime.

The effort, using technology from IBM, would not necessarily focus on kids who are already racking up time in juvenile court.

Instead, authorities would examine other indicators of trouble, such as skipping school. Asked if police would examine such factors as a person's neighborhood or the stability of their home environment, Peterson stressed that the examination would focus only on behaviors. Police would then suggest a referral to the Boys and Girls Club or a similar program that aids at-risk youths.

Former Olmsted County Attorney Ray Schmitz said the program raises many questions that Rochester didn't consider when it approved the program. Those include what standards they'll use to identify the juveniles, and what will happen if the police refer one to a social organization and the child or his family declines the intervention.

"Is that a role for police?" he asked of the decision to target juveniles for intervention. "Or is it a role for social services agencies?"

In recent years, cities around the country have turned to some form of predictive policing, combining databases and geospatial technology to trace patterns in crime. Such examinations, for example, allow departments to increase the number of cops in areas where more crime is likely to happen. Critics have raised concerns about violations of privacy and the appropriate bounds of government surveillance.

A report last year by the nonprofit RAND Corp. stressed that predictive policing only shows the risk of a future event happening, not an actual prediction of the future, and noted that the results were only as accurate as the underlying data that departments were analyzing. Researchers warned about some pitfalls, noting, "The very act of labeling areas and people as worthy of further law enforcement attention inherently raises concerns about civil liberties and privacy rights."

More efficient arrests

Rochester's predictive policing dates to 2011, when it launched an IBM program called Infosphere Identity Insight to scrutinize the most frequent offenders.

Culling information from databases of jail bookings, warrants and other records, the program also enhances an officer's ability to look up people's associations with one another. That means that an officer would be able to tell before approaching someone without a criminal record that their boyfriend is a felon.

At police headquarters last week, crime analyst Joseph Koncur displayed the computer system on a screen, showing how it revealed networks of people.

A thief reported at Gander Mountain was described by a witness as a white woman between ages 30 and 40, yet the license plate on the car she was seen jumping into was registered to an elderly Asian woman. Using the system, he said, they were able to see that the owner's son's girlfriend matched the description.

IBM, a major employer in Rochester, has sold its software to many cities ranging from Memphis, Tenn., to Charleston, S.C. The company markets the system as leveraging "advanced algorithms specifically optimized to recognize nefarious individuals and organizations despite their sophisticated attempts to mask their identity, their unscrupulous relationships, and their activities."

Building on the same technology, the city even developed a smartphone app that allows officers to do a more detailed search on a suspect before walking up to the person at a traffic stop or somewhere else out in the field. Cops now have access to information about not just the registered owner of the car, but people one or two degrees of separation from that person who are on probation or who have been involved in a violent crime or weapons violations.

All of this has made the department more efficient at making arrests, according to Peterson, the police chief.

"If we get that effective in shutting off the supply of those serious, prolific offenders, we'll be doing a lot more good for the community. We'll certainly be doing a lot more for those kids."

Privacy, profiling concerns

Peterson and the police department are offering scant details about the next phase of the program that involves juveniles because the city is still trying to hammer out exactly how it would work. It would likely need cooperation of other agencies, like the school system.

But the police chief said that it would start with a close look at Rochester's most prolific adult offenders. Officers would review commonalities among them and then identify a field of young people who are not yet in the criminal justice system but share some of the same early behaviors as the adult offenders.

Peterson vowed not to green-light the pilot project until the city finishes assembling a seven-member oversight commission this year.

Kolloh Nimley and other critics, especially data privacy advocates, say too many questions remain.

"Who has access to that data, how long will the data be available, where is it stored, what's the protection policies?" said Nimley, a community program specialist for the Council on Black Minnesotans. "All those things are a concern."

She added that she also worries that targeting people at such a young age could involve profiling people of color.

But Peterson insists that a more data-driven approach actually prevents racial profiling, and that race won't be taken into account. He said the department does not plan to hold onto any data after referring young people to outside programs.

"I think the more information that gets out there, the more comfortable people are going to be," Peterson said. "We're not creating some monster to follow kids around their whole life."