A college professor's two-and-a-half year project has brought the cutting edge of police work out of the shadows.
Since 2011, Silicon Valley-based software firm Palantir has helped the Los Angeles Police Department analyze data, ranging from license plates photos, to rap sheets, traffic tickets, listings of foreclosed properties and more.
The company, which also works with government agencies such as the CIA and FBI, is quietly transforming how police operate. Palantir doesn't reveal how many clients are using its tools, but police departments in both California and New York have previously worked with the company.
Sarah Brayne, a sociology professor at the University of Texas in Austin, conducted more than 100 interviews of officers and civilian employees. She went on ride-alongs in patrol cars and a helicopter, and watched data analysts answer queries from detectives. Brayne also observed divisions adopt the new technologies.
Her results were published online in the American Sociological Review last month.
Experts say that Brayne's work is a window into the future of law enforcement. It illuminates the promise big data holds for making police work more efficient. But it also shows its perils: how data, which is generally thought to be objective and fair, can exacerbate biases.
Civil rights lawyers and advocates for minority communities, among others, have criticized the use of data by police departments. Nineteen cities have considered legislative proposals related to police surveillance, according to the ACLU.
The Los Angeles Police Department and Palantir both declined to comment for this article.
What follows are six of Brayne's most striking findings.
1. Surveillance today is unprecedented
Just as Facebook made it easy for you to track your friends, Palantir simplifies law enforcement's ability to monitor potential criminals. Its platform is similar to a social network.
Basic information such a person's name, gender, school affiliations are entered. Officers sometimes record this data on note cards when they stop citizens on the street, say, following a traffic accident. Information is collected not only on the individual, but the people with them at the time. Additional data comes from government agencies, or data the department purchases from private companies.
Once a person is in the system, officers can receive automated alerts on their smartphones. For example, an alert may be triggered when an individual's car is seen driving into a specified neighborhood. The Los Angeles Police Department has integrated cameras installed on its cars and along streets into Palantir's system. The cameras photograph license plates, feeding their time and location into the system.
Another example: If a suspected bank robber's vehicle is caught on camera going near a specified bank, a police officer may receive an alert.
Data once held in technical silos, requiring multiple searches, is now centralized and more easy to access to aid investigations.
2. Citizens without police contact can be tracked
Brayne found that Los Angeles' databases also included individuals who haven't had direct contact with law enforcement. For example, just having a link to a person of interest is enough to be in the city's Palantir system. This could be a simple as being in a fender bender with a gang member.
There's even a "lover" category in Palantir -- a romantic relationship with a person of interest could lead you to be in a database, too. Once you're listed, officers can receive alerts about you.
"I'd caution against the thinking that if you have nothing to hide, you have nothing to fear," Brayne told CNN Tech. "That logic rests on the assumption of the infallible state. It rests on the assumption that actors are entering information without error, prejudice or discretion."
For example, African Americans are seven times more likely than white Americans to be wrongly convicted of murder, according to research from the National Registry of Exonerations.
3. Sharing data helps cops do their jobs
The Los Angeles Police Department covers a huge area — 500 square miles, the equivalent of 22 Manhattans. Distant yet related crimes may not be connected to each other.
But in Palantir, it's easier for detectives to see connections. An officer can sign up for alerts about a specific type of crime with keywords such as "marijuana, robbery, male, 5-foot-11." They may learn that a colleague is trying to solve a similar type of crime.
There's also a benefit in planning operations, Brayne found. If the cops are going to search a home, they can search Palantir for relevant data. Officers might be warned about a gun registered next door, a gang associated on the block or a warrant issued for a nearby crime.
In one example, Brayne recalled how officers zeroed in on a car involved in a series of copper wire thefts. Officers chose to search for license plates photographed near where the wire was stolen, during a relevant time window. This gave them a list of cars present at all three locations.
4. Not all officers love the new surveillance tools
From the outside, police departments may seem like a monolith, working in lockstep to vacuum up data and monitor citizens. But Brayne found the department's more than 12,500 employees have varied opinions on big data systems. Not every division in the Los Angeles department is using Palantir, and some don't even know what it is. Others are wary of it.
"All we're doing right now is, 'Let's just collect more and more and more data and something good will just happen.' And I think that's kind of wishful thinking," a civilian employee in the department told Brayne.
The rise of surveillance also means that cops can be watched like never before. A shift to police cars automatically reporting their location every five seconds wasn't welcomed by the police union, Brayne noted.
5. Big data has a role in inequality
Individuals in low-income, minority areas were more likely to have their "risk" measured by police, the findings showed.
In 2011, Los Angeles started a program in such a neighborhood that identified chronic offenders, who were given a point value and ranking. Points were awarded for a violent criminal history, gang affiliations, prior arrests with a handgun or for being on parole or probation. Each time an officer stopped an individual on the list, a point was added to their score. The practice has been explained in police department documents but not detailed in media reports before now.
The danger, Brayne warns, is that data can become a self-fulfilling prophecy. An individual with a high point value is likely to receive more police contact. More interactions will only raise their score, leading to even more interactions with police.
According to a January 2017 letter from Los Angeles police chief Charlie Beck, 12 of the department's 21 patrol divisions used the quantified risk assessment program at the end of 2016.
6. Laws haven't kept up with technology
Big-data policing changes the game on warrants and constitutional violations, too. Cops are required to have a warrant to gain certain insights, but now they can use powerful databases to deliver the same findings without one, Brayne reported.
Privacy is also an issue. Civilian employees told Brayne they were considering new sources, such as call data from pizza chains and address information from contact lens rebates. Every scrap of data -- whether collected by government agencies or private companies -- can be potentially useful for law enforcement.
What kind of future do we want?
The LAPD's use of Palantir is a reminder of the implications of data. Technology can be a double-edged sword.
Andrew G. Ferguson, a professor at the University of the District of Columbia law school and author of the forthcoming book "The Rise of Big Data Policing," believes Brayne's findings reveal the future of policing. He says a national conversation must be opened, and the public is largely in the dark.
The New York Police Department, which has revealed little about its previous use of Palantir, declined to comment to CNN Tech.
For smaller police departments, technology such as Palantir's may be too expensive, according to Brayne. A recent Wired report said that California law enforcement agencies spent over $50 million with Palantir since 2009.
"We need to start talking -- before the technology gets too far and too advanced -- about whether we're comfortable with this growing data collection about the citizenry," Ferguson said.