GovLoop

The Power of Predictive Policing in Surf City

When the term ‘predictive policing’ comes up, what movie immediately comes to mind?

If you said The Minority Report you aren’t alone. According to Santa Cruz Deputy Chief of Policy Steve Clark, that’s a common reaction. In reality, the technology currently being tested by law enforcement today is just as fascinating as science fiction – but without the dystopian terror.

Deputy Chief Clark recently joined a GovLoop online training to tell the Santa Cruz Police Department story of using big data analytics to help police officers become more efficient and effective crime fighters. An archived copy of the virtual training session, hosted in partnership with IBM, can be viewed in its entirety above.

For now, we’ve provided a recap of Deputy Chief’s predictive policing story.

A Policing Conundrum in Santa Cruz

In the beachside city of Santa Cruz, California, the police department had a problem. In the last few years, the department found itself facing an increasing demand for its crime-response services. This demand had a secondary effect on the department: any time spent responding to crime meant less time mitigating future events through what Deputy Chief Clark calls proactive policing.

“Policing time for us is a zero-sum equation unless you add more staff,” explained Deputy Chief Clark. “And I don’t know if any of us are adding staff in these lean economic times.”

This dynamic can produce a feedback loop in which an engaged police department has less time to mitigate future events, which can lead to an increase in demand, which ultimately leaves even less time for proactive policing. Unabated, the cycle can quickly deteriorate beyond control.

“For us, we had to find a way to be much more efficient with the resources we had to keep our communities safe,” said Clark. “We started to look at something more productive.”

The Solution: Big Data Analytics

The Santa Cruz Police Department, under Deputy Chief Clark’s direction, decided to use its massive library of crime, community and geographic data as variable inputs for a model that they would ultimately use to predict specific criminal ‘hot spots’ for the day.

“In our case, we had data, but we just weren’t using it effectively,” said Clark. “It wasn’t until we improved our decisions and actions – using [big data analytics] – that we really started to see some benefits.”

The department has since adapted a mathematical model used to predict earthquakes – the Earthquake Aftershock Model (ETAS, which stands for Epidemic Type After Shock sequence) – to help figure out the diffusion and reenactment of crime in the community.

What Does the Model Look Like?

From Deputy Chief Clark’s slides: Image courtesy of Dr. Jeff Brantingham and Predpol

The triangle above represents, at a high level, the different sources of data used in the model.

These factors are then plugged into the model, which creates an algorithm that is used to chart out on a map the area with the highest probability for crime on that particular day (see image below).

From Deputy Chief Clark’s slides: A snapshot of ‘hot spots’ identified by the predictive policing model.

Deputy Chief Clark contrasted this method with the traditional approach to law enforcement, which involved using a historical model for mapping out crime in the city. “The problem is that it is yesterday’s crime,” said Deputy Chief Clark. “Your city is actually a living, breathing thing. While we are busy chasing yesterday’s crime in that hot spot, that hot spot might not be there today.”

This gets at the heart of the goal of predictive policing: using empirical evidence to break up the sometimes-inefficient policing habits that build up over time.

“A simple way to describe it is this: we are just using data to see what’s really out there,” explained Clark.

“Sometimes in our jobs in government, we can see things differently from how they really are,” he continued. “It is only when we focus and really look at the data and analyze it, that we see that our cities or states or communities are not really moving in a way that we think they are.”

Deputy Chief Clark used himself as an example. He said he could take you on a ride-along and immediately take you to a place where he knows he can make an arrest. “Every police officer out there knows the spot where they can go right now and probably go arrest somebody,” he explained.

“The thing is that we fall into these habits where we begin to see the city in this way,” said Deputy Chief Clark. “[We] aren’t really doing things that are actually affecting crime or the quality of life in our community.” As law enforcement officers – or any government employee for that matter – fall into that trap, something is needed to shake them out of that pattern. That is the power of predictive policing.

The Results

The Santa Cruz Police Department has seen early success since the launch of its predictive policing program in 2011. Burglaries were down 11 percent from the previous year, robberies down 8 percent and auto theft recovery up 22 percent.

Additionally, the department found that newer police officers who started their careers in the predictive policing era had a faster learning curve and increased problem-solving skills compared to those who didn’t benefit from the tool.

“We don’t tell the officers how to do police work,” Clark explained. “We just give them a pointer as to where they have an opportunity to mitigate the most probable crimes that are happening out there.” The officers already have the toolbox – the tool just provides them with a tool for more efficiently employing their training and experience.

The Future of Analytics

The model used by Santa Cruz is being adapted all over the world, from Los Angeles to England to Uruguay. “This is a growing trend,” said Deputy Chief Clark. “No matter who you are or what you do in government, the use of data, and the ability to analyze it and have it impact your decisions and actions is not going away.

“In ten years we are going to look back an wonder, ‘How did we do the job without these tools?’”

Additional Resources

Exit mobile version