Jerry Dyer is betting he soon will know what criminals will do before they decide to break the law.
If Fresno's police chief is right, local crime-fighting — and the civil-liberties debate — will never be the same.
Dyer is overseeing the arrival of predictive policing in Fresno and throughout Fresno County. His police department is lead agency for a state grant that will pay for six trained crime analysts — something local law enforcement has never had on this scale.
Three analysts will be in Fresno. A fourth will go to Clovis. One will be based in Kingsburg and focus on the county's eastern half. The sixth will be based in Fresno and concentrate on the county's western half.
The analysts will have many duties, but foremost will be using powerful computers, sophisticated software and huge amounts of information to make good guesses on where criminals might strike next.
The analysts might also predict who is likely to be the next bad guy.
Predictive policing, Dyer says, "will look at where crime has occurred and what occurred and what time it occurred so you can logically predict where crimes are likely to occur the next day."
Predictive policing in one sense is as old as the eternal good-vs.-bad fight. Cops always have used common sense and crime patterns to figure out where the next stakeout might bear fruit.
But that was before the "Big Data" revolution. "Big Data" is shorthand for supercomputers capable of speedily analyzing billions or even trillions of pieces of information.
Correlations can emerge that would have been beyond the imagination of analysts from the previous century, let alone gumshoe detectives slogging away with pencil and paper. These Big Data correlations already are proving useful to business and government.
Controversial, as well. Recent revelations that the National Security Agency is monitoring the phone records of millions of Americans have stunned much of the nation.
Government officials say the NSA's monitoring helped disrupt dozens of terrorist plots. Critics worry that never-ending technological advances will lead to a "Big Brother" world where privacy vanishes. Big Data correlations might also lead to a racial or ethnic stereotyping no less dangerous to democratic ideals because it's based on impersonal data rather than institutional bias.
The Fresno Police Department and its local allies aren't in the NSA league. They aren't crunching hundreds of huge databases or eavesdropping on every cell-phone chat.
But in many ways the modern Big Data-Big Brother world has arrived in Fresno, and done so rather quietly.
Dyer is remodeling a suite of rooms on police headquarters' second floor to serve as a confidential control center. The chief envisions wall-to-wall computer screens. Estimated cost: $150,000, funded with grants and donations (cash and labor).
Monitors elsewhere in headquarters already display video from 150 cameras located throughout the city. These cameras record every movement of anyone within their range. Similar cameras 3,000 miles away helped nab the alleged Boston Marathon bombers.
Fresno County Sheriff Margaret Mims has her own high-tech center in the basement of her department's downtown headquarters.
There are doubters. The top cop in Coalinga sees little value in predictive policing for his town.
And it's not clear whether local elected officials fully grasp what is happening. For example, Fresno City Hall has City Council-approved policies for video policing but none for predictive policing.
But this new reality won't be stopped.
"It's going to be exciting," Dyer says. "If we continue to lose resources, and have fewer and fewer officers, then we have to make sure we make the most of the officers we have."
Leaving the buggy age
This won't be the first time for crime analysts at the Fresno Police Department.
Not too many years ago, Dyer says, the department would tap a couple of community service officers on the shoulder and tell them to take a bit of analytical training. Presto — they were deemed crime-data experts.
Community service officers were the same people who drove out to someone's house to take a stolen-car report that wasn't worth the time of sworn officers.
Then, starting in 2008, City Hall's finances went over the cliff. The community service officers/crime analysts disappeared in waves of budget cuts.
The department didn't stand still on the technological front. Dyer every month hosts a two-hour, in-house statistical analysis called CrimeView that is heavy on numbers and computer-generated crime patterns. But Dyer didn't have expert number-crunchers on staff. The result was an operational hole. "Police officers don't like analyzing data," Dyer says. "They want someone to analyze the data and tell them where to go."
The City Council fixed that in June when it told Dyer to negotiate the details for part of a $1 million state grant for the 2013-14 fiscal year.
The grant is a piece of Gov. Jerry Brown's effort to quiet critics throughout the state over realignment — his emptying of state prisons and the subsequent chaos that is causing in local jails and their communities.
Nearly half of the grant went to Fresno County District Attorney Elizabeth Egan so she could start an electronic filing system for criminal complaints.
More than $500,000 will go toward the six analysts. The Fresno Police Department was picked as the project's "fiscal agent" — in essence, a clearinghouse to make sure the money goes where it should.
Clovis will hire and supervise its analyst. The other five will be hired by Fresno, three to be under Dyer's control. The other two will be supervised by law enforcement chiefs in western and eastern Fresno County.
Deputy Chief Keith Foster in a report to the City Council wrote that realignment has led to violent and habitual criminals being dumped in places like Fresno.
The bad guys pay no attention to city boundaries. All cities have money woes.
The six analysts, Foster wrote, "will allow us to effectively implement a regional effort to track and monitor career criminals."
Foster only hinted at one of the key ideas behind predictive policing: Criminals, like all humans, are creatures of habit.
Full-moon crime rises
Fresno generally has more robberies under a full moon than a new moon.
"I'm not saying the criminals become werewolves," George Kikuchi says with a smile.
But the assistant professor of criminology at California State University, Fresno knows he's right — statistics prove it.
Kikuchi qualifies his finding. The extra light of a full moon may lure more potential victims to entertainment spots, or persuade them to stay longer.
But Kikuchi's point is one of psychology: people aren't entirely free agents.
"Because we're dealing with human beings and human behavior, there's always room for uncertainty," Kikuchi says. "Criminology is not natural science where prediction is almost perfect — if you add chemical A to chemical B then chemical C is produced. But nonetheless we can still estimate if event A is more likely to occur than event B.
"To some extent your behavior is patterned."
Crime fighters are combining this ancient wisdom with Big Data.
Todd Wilson, chairman of Fresno State's Computer Science Department, says that Big Data, like "cloud computing," is a vague term more concerned with generating pop-culture buzz than identifying a new concept.
"It's used to refer to many different but related ideas that center around getting useful information from data sets that are typically very large, unstructured, and automatically generated, where the focus is more on 'emergent' properties of the data rather than with the individual records," Wilson says.
By "unstructured," Wilson says, he means data that is not organized in an immediately useful way — unlike, for example, a payroll database. A payroll database needs perfection. A single missing Social Security number causes problems for the payroll clerk.
Big Data, on the other hand, "is more about aggregation," Wilson says. "A mistake in the data or a missing record, as long as it was relatively isolated, would hardly be a significant concern, since it is overall patterns and trends that are of the most interest."
Wilson emphasizes that computers have been crunching data for a long time. But, he adds, stunning progress on several fronts is producing a new world in technology.
The "convergence of low-cost storage, fast processing, and the 'digitization' of much of our lives and the external world now make it possible for these techniques to be used to organize the data and generate information from it in ways that were not possible before," Wilson says.
The lay of the land appears clear: Everyone's every move seems to be digitally recorded; government seems to have access to everything; people are largely predictable; computers can figure out anything.
That's the hype of Big Data in general, and predictive policing in particular. Could it all just be nonsense?
No, says Wilson: "Much of the hype is justifiable."
You are being watched
Fresno police have embraced a slice of Big Data for nearly a decade, with results far from perfect.
There are 150 video-policing cameras on poles and rooftops throughout town. They work nonstop, producing more than 25,000 hours — nearly three years' worth — of recorded video every week.
The system began years ago with a few cameras in high-profile places, like Roeding Park and regional shopping centers. Video policing took on a new character with the addition of 75 cameras in 2006.
Dyer expects to add 10 to 15 cameras a year until the department has 250.
There have been successes. A camera in October 2010 caught a glimpse of a pickup driven by a man who had kidnapped an 8-year-old girl. That snippet led to the girl's rescue the next day by Good Samaritan Victor Perez.
Dyer says patrol officers responding to a crime often ask headquarters: "Did a camera get anything?"
The challenge is taking full advantage of the mountain of video data. A frustrated Dyer made this clear at the department's Aug. 14 CrimeView presentation.
"We have to find a way to take better advantage of these video cameras," Dyer told his top officers.
Staffing shortages and strategic confusion are the big problems.
The cameras run 24/7, but most of the video is kept for only about a week. Police cadets used to keep an eye on the monitors, but for money reasons they've gone the way of community service officers. Dyer says officers on light duty — perhaps recovering from on-the-job injuries — sometimes watch live feeds.
For much of the day, though, no one is watching all 150 cameras. That's no problem if the system's aim is to identify the license plate of a fleeing criminal's car an hour after the fact. It's a big problem if the aim is to nab the criminal in the act. "We're just not maximizing them," Dyer said at CrimeView.
The department can't seem to figure out what it wants the public to know about video policing.
Dyer says cameras deter crime if everyone in an area knows they're around. Prominent warning signs are the key, he says. But the signage in downtown and southwest Fresno, neighborhoods with the most cameras and the biggest need for them, is terrible, he adds.
Dyer says he will change that.
At the same time, Dyer has been reluctant to identify specific locations of all cameras. Part of the reason is security — some cameras cover sensitive spots. Part is tactical — criminals don't need to know the system's coverage holes.
Video policing's future is as amazing and unpredictable as the Big Data revolution itself. For example, The New York Times recently reported how computers and surveillance systems may one day be able to scan crowds and automatically identify people by their faces.
The Fresno City Council recognized seven years ago that video policing's growing scale raises valid worries about Big Brother.
The council created a position that in essence is a video-policing auditor. Ideally, the position is to be filled by a former federal judge with expertise in civil liberties issues. The auditor would field citizen concerns, review video-policing operations and report annually to the council on whether the system is playing fair. In other words, no cameras peeping through bathroom windows.
This concept didn't get off to a smooth start. The council is still awaiting its first report. Oliver Wanger has been auditor since September 2012. The well-known former federal judge says he hopes to deliver a report by the end of October.
Wanger says he hasn't come across any civil liberties violations so far.
The people running Fresno's video-policing program "take the rules that confine them very seriously," Wanger says. "I'm very happy to see that level of responsibility."
As to the Big Data revolution in policing, Wanger says the community must ask a basic question. "To what use is the data going to be put?"
Wanger has an answer: "We don't know. All we know is the data is being collected."
Lots of data
The predictive policing experiment almost certainly will suffer growing pains.
Dyer says analysts hired by his department should be in place this month. They will step into a network of computerized intelligence that includes county, state and national links.
The Fresno Police Department's CrimeView system has been 10 years in the making. It's no technological slouch.
For example, Dyer in his office can type a few commands on his desk computer and a large wall screen instantly displays the last known home address for every parolee in Fresno busted for burglary. A few more key strokes and he adds the location of each burglary in Fresno over the past 24 hours, or two weeks or month.
When the two clusters of icons — burglary-obsessed parolees and burglary victims — converge, Dyer says, you've got a crime "hot spot" and potential suspects.
There is no shortage of CrimeView variables at Dyer's disposal. Homicides, rapes, robberies, car thefts, aggravated assaults, felons on probation, car accidents — they and much more can be plotted on a map of Fresno.
Sheriff Mims has much the same capability at her headquarters.
The grant's analysts are supposed to super-charge this digital capability. They're to do it with their unique expertise and cooperation among the law enforcement agencies they serve.
"Ideally, we'd like to have people develop an expertise within a particular crime or within a particular geography," Dyer says of the three analysts focused on Fresno.
The databases to be crunched are in the public domain or part of a government network, Dyer says. The information need not be directly connected to criminal justice to be of use. Business databases, for example, can be mined.
"It's not enough just to look at the crime," Dyer says. "We need to look at it from a problem-solving perspective. For example, is it a recycling center in a particular neighborhood that is contributing to a crime in that neighborhood?"
But analysts aren't coming to town just to tell Dyer that a bunch of robberies in the Tower District means it's probably a good idea to keep an eye on the neighborhood around the Tower Theatre.
They're coming to shake things up. Dyer can't explain exactly what that means, at least not yet. He says the analysts will take orders. But he also expects them to use their creativity to help Fresno police broaden their use of Big Data analyses.
"Individuals who have a gift for analytics — they can't get enough," Dyer says. "They want to analyze more and more. What we have to do in policing is say, 'Here's the information we want you to analyze and it must be for this specific purpose.'"
That purpose is to put a dent in the most serious crime: Murder, rape, robbery, assault, property crimes.
Fresno is headed into new territory. "The data we have tells us where the crimes are likely to occur," Dyer says. "It's going to also require human input to determine who might be committing those crimes tomorrow or the next day."
Digital world rights
Police already have a good handle on who commits many of Fresno's crimes.
More than 3% of all gang members in America reside in Fresno County, even though the county has just a fraction of the nation's population. Fresno by itself is home to more than 10,000 validated gang members, with as many as 10,000 others affiliated with gangs.
Society and the law give police a monopoly on the use of lethal force. The Fresno Police Department almost every year is rocked by controversial officer-involved shootings. Community activists keep a constant and critical eye on Dyer.
An analytical crime-fighting model is coming into this complex world. This model most likely will spur the police to make increasingly refined guesses on the next "hot spots." This Big Data model, Dyer acknowledges, on occasion will have police picking specific people as possible perpetrators before they act.
There are 94 known gangs in Fresno. Many of the most active, police officials say, are African-American, Hispanic and Asian gangs. But that's only one side of the coin. The vast majority of African-Americans, Hispanics and Asians are law-abiding. And plenty of whites commit crimes.
Here's the root of the matter: Will crime-fighting Big Data have a tendency to target innocent minorities?
One expert says that is among the many important questions about predictive policing.
Andrew Guthrie Ferguson, an assistant law professor at the University of the District of Columbia, wrote a paper last year called "Predictive Policing: The Future of Reasonable Suspicion."
Predictive policing could be useful, both for arresting criminals and for deterring them with a police presence, he says in a phone interview. But he has concerns about how it will affect the Fourth Amendment that guards against unreasonable searches and seizures. He says it's only a matter of time before a lawyer challenges a court's admission of evidence based on a predictive calculation and unproven technology.
"How do we know it reduces crime? Just because police chiefs have embraced it doesn't mean it's true," Ferguson says.
Ferguson says predictive policing is based in large measure on extrapolations from crime data. But not all crime is reported or recorded.
Ferguson says data-driven "hunches" are no more reliable than an officer's "hunch," which traditionally has been legally insufficient to justify an arrest on reasonable suspicion.
Ferguson's key concern is whether computer-generated crime forecasts will lower the threshold for reasonable suspicion. He says this could lead to increased racial and class profiling as well as pat-downs and arrests that otherwise would have been perceived as unwarranted.
Ferguson says the public doesn't know the full implications because there have been no appellate decisions on the legality of predictive policing.
If a burglar breaks into a house and sees similar style homes in the neighborhood, Ferguson says, "there is a statistical likelihood he will go from house to house."
But this same analytical model doesn't work for violent crimes which "are more interpersonal than opportunistic," he says.
Probable cause and reasonable suspicion require more than demographic probabilities, Ferguson says. There must be something specific to the defendant to create the probable suspicion to arrest him or her. This could be a furtive gesture, an informant's tip or excessive nervousness.
Imagine, Ferguson says, a judge asking an officer to explain his reasonable suspicion for stopping someone, only to have the officer say a computer told him to do it.
Says Ferguson: "How are you going to cross-examine a computer?"
Covering new ground
Former Judge Wanger says there are no easy answers.
"OK, you've got all this data — what are you going to do with it? That's where the rubber meets the road," he says.
Wanger says he finds today's law enforcement agencies to be better informed and observant of people's rights. Still, he recommends that predictive policing efforts be accompanied by clearly identified objectives.
Fresno City Council Member Oliver Baines, a Fresno police officer for nearly 12 years, says elected officials should get a good idea of predictive policing's pros and cons before drafting binding guidelines.
"We may be boxing something prematurely," Baines says.
Fresno City Manager Bruce Rudd says police analytics won't stereotype. Its purpose is merely to "address a problem before it becomes a bigger issue," he says.
Sheriff Mims, whose department also provides law enforcement services to the west-side town of San Joaquin, says Big Data analysis will deliver innovation, not bias.
"Racial profiling is bad in any organization," Mims says. "But profiling is another word for common sense — looking for the right person."
The six grant-funded analysts are supposed to work together as well as for individual departments. However, it's still not clear how this teamwork will occur. Will law enforcement agencies just naturally get along? Or will turf rivalries and institutional indifference carry the day?
Coalinga Police Chief Cal Minor says the analysts are a waste of money for his small west Fresno County city.
"In our city the numbers are too small to form any statistical opinions," Minor says. "There may be a pattern if a couple of people or group is committing the crime, but most are people wandering about looking for the opportunity."
Reedley Police Chief Joe Garza recalls watching the department's analyst 25 years ago charting crime patterns with "pin-maps." When a month ended, Garza says, the analyst removed all the pins and got ready for a new round of pins.
The task was the essence of time-killing tedium in a pre-computer age, he says.
Garza likes the Big Data world.
"Analysts do not have a 'crystal ball' and cannot foresee crime occurring or who might commit it," he says. "But what they do have is the ability to notify supervisors of where crime is likely to occur."
Clovis Police Chief Janet Davis already has one analyst on staff. The grant-funded analyst should be on duty in November, she says.
"We are excited about having an additional crime analyst on our staff," Davis says.
Kikuchi, the Fresno State assistant criminology professor, says data analysis can enhance fairness.
"Without data, the only thing police departments may be able to say is that based on their experience, they think more crimes are occurring in a neighborhood," Kikuchi says. "But that's a subjective perception, which may not be a convincing argument."
As a general rule, Dyer says, violent crime occurs more often in low-income neighborhoods.
Property crimes are found throughout Fresno, he says.
Predictive policing is doing nothing more dangerous than "taking data and doing something more purposeful with it," Dyer says.
Wilson, chairman of Fresno State's Computer Science Department, says it's too late for an either-or debate. The technological Pandora's Box has been opened — oceans of information, super computers and curious humans are permanently on the loose.
"Just as you can't isolate the point where consciousness arises from our neurons, you can't say where the 'scary' effects arise from individually innocuous data," Wilson says.
No one knows how Big Data will make peace with liberty, Wilson says.
"We just have to keep exploring, and using our moral sense as we do."