Computer Scientists Think They Have The Answer To Police Misconduct

Police officers secure the scene of a shooting involving a police officer near a church south of downtown Memphis, Tenn., Wednesday, Jan. 27, 2016. CREDIT: AP PHOTO/ADRIAN SAINZ
Police officers secure the scene of a shooting involving a police officer near a church south of downtown Memphis, Tenn., Wednesday, Jan. 27, 2016. CREDIT: AP PHOTO/ADRIAN SAINZ

Gut-wrenching video footage of police officers killing unarmed black people or using excessive force has become an everyday reality. With each video, law enforcement officials insist these killings are rare and committed by a few bad actors. The problem is, no one really knows for sure.

But two new data projects funded by the Knight Foundation are primed to change that. Groundbreaking new tools could enhance police accountability by making once-scattered police data open to the public.

Why Police Data Is So Elusive

“Right now, there’s not even a comprehensive look at all 50 states and what data is collected [by police] and what they do with that data,” said Vignesh Ramachandran, a data journalist and managing editor at Stanford’s Peninsula Press.


Police data in any form is hard to come by. Media organizations and federal law enforcement agencies are the primary sources for data on police conduct, such as officer-involved shootings.

“The anecdotal stories are always very compelling, firing us all up.”

But the information collected by the government typically relies on voluntary self-reporting. That system often leads to information being withheld, incomplete, and inconsistent across law enforcement agencies.

Even recent attempts to make government-collected police data more robust, fall short by relying on volunteered information rather than a mandate. The White House launched its police data initiative last year, which allows law enforcement agencies to voluntarily share data regarding officer-involved shootings, incidents, assaults on police, and 911 calls through the Public Safety Open Data Portal.

The FBI is planning to revamp its highly criticized database that tracks fatal police shootings by 2017. The system, which only recorded deadly gun incidents from 3 percent of the nation’s police department, will soon include all serious injuries and deaths caused by police officers, expanding the database beyond officer shootings. The system, however, will still rely on voluntary reporting and may only capture a fraction of incidents.


“The anecdotal stories are always very compelling, firing us all up,” Ramachandran said. But backing up those stories with the raw data enhances those accounts in a new way by statistically showing how the rest of country experiences police stops and pinpointing disparities.

Getting The Numbers On Driving While Black

Fatal confrontations with the police don’t always involve suspicion of a violent crime like armed robbery. They occur during routine traffic stops: Walter Scott of Charleston, South Carolina was shot by a police officer after being pulled over for a broken brake light. Sandra Bland was violently wrestled to the ground by a Texas police officer who pulled her over for improper signaling. She was later found hanging in a jail cell three days after the arrest.

Scott’s and Bland’s deaths after traffic stops escalated were only brought to light because civilians filmed them. But exactly how often these kinds of encounters go wrong is still a murky question.

“Most of the time it’s innocuous when people are pulled over…”

Ramachandran is trying to make it possible for the public to answer that question. He works with an interdisciplinary startup project at Stanford University called Law, Order, and Algorithms, which is compiling the demographic data police officers collect during routine traffic stops.


“Most of the time it’s innocuous when people are pulled over for speeding,” said Sharad Goel, an assistant computer engineering professor at Stanford University, who also leads the Law, Order, and Algorithms project. “At the same time it’s not clear what’s happening the rest of the time.”

“At the same time it’s not clear what’s happening the rest of the time.”

Racial disparities are particularly stark in how traffic law is enforced. In McKinney, Texas, where a young black girl at a pool party was assaulted by a police officer in 2015, white people were were disproportionately ticketed less often than African Americans. Nearly half of all tickets McKinney police issued from 2013 to 2014 were for black drivers, when African Americans only count for a third of the population.

In the case of Walter Scott, officer Michael Slager argued through his attorney that the North Charleston Police Department’s quota system was to blame. The department’s policy mandates that officers make three traffic stops each day and failing to do so could result in a loss of promotion or earned time off. But that same policy was shown to disproportionately target low income residents of color.

The Stanford project, which received a $310,000 grant from the Knight Foundation, has collected information for more than 50 million highway traffic stops across 11 states through public records requests. When it’s up and running, the database will allow media and the public to see where stops are happening, who’s involved, the reason for the stop, and pinpoint down to a county level where discrimination could be a problem.

But the group is still in the midst of cleaning up the data it has already and won’t have preliminary results until the Spring. “It sounds like measuring discrimination would be easy but it’s difficult,” Goel said. “There are a lot of gray areas.”

Piercing The ‘Blue Wall’

Discrimination in police encounters spans more than traffic stops. One of the only ways to document any kind of police misconduct is to file an official complaint.

Every police department has misconduct complaints, but filing and getting access to them isn’t always easy: There’s often a layered bureaucratic process to file complaints, requiring citizens to take off work to get a signed affidavit. And depending on the state, those records aren’t public. Only 12 states require complaints to be a part of the public record, compared to 23 states that keep them confidential, according to data compiled by WNYC.

Only 12 states require complaints to be a part of the public record.

The Chicago Police Department has been at the center of the debate. The department has been fraught with controversy over discriminatory practices and for failing to discipline police officers who get complaints. The Citizens Police Data Project — a joint venture run by the Experimental Station and the Invisible Institute — is trying to address the issue and released a data tool in November that allows the public to access four years’ worth of Chicago police complaints. But getting the metadata for those 56,000 complaints has hasn’t been easy.

“The story of our data project began about a decade ago, when our founder Jamie Kalven was reporting on the last high-rise housing projects in Southside Chicago,” said Alison Flowers, an investigative journalist with the Invisible Institute, a media production company in Chicago. “He interviewed a resident Diane Bond who said she was sexually assaulted by police and repeatedly harassed,” by a group of officers nicknamed the Skullcap crew.

With Kalven’s help, Bond filed a complaint and a lawsuit against five officers in the Chicago Police Department in 2004. During the course of the trial, Bond’s lawyers asked for a list of police officers who received the most complaints — four of the five accused police officers were on it. But those documents were sealed until 2014, when the Illinois Appellate Court ruled that police misconduct records should be publicly available.

“The work that has brought us to this point has been lean and largely unfunded,” Flowers said, indicating that the legal battle is still going on. Only records from the past four years were released after the 2014 court decision, but all archived misconduct records were requested.

“It would be easy to say we’re anti-police; that’s really not the case.”

CPDP’s database doesn’t reveal the content of each complaint but allows users to search for currently employed CPD officers by name or badge number to see the reason for the complaint, such as evidence tampering, discharge weapon, use of force, or profanity. A preliminary analysis of the data shows less than 3 percent of complaints end in disciplinary action and the more complaints an officer has, the less likely he or she will be disciplined.

Despite a spotlight on its legal battles, the Chicago police trying to purge misconduct complaints. The Fraternal Order of Police demanded the city destroy complaints older than five years. There’s now a fight over whether records dating back to 1967 could be made publicly available — and ultimately a part of the Citizens Police Data Project — or permanently deleted.

“It would be easy to say we’re anti-police; that’s really not the case,” Flowers said. “We want this tool to promote public safety and to weed out the officers who are contributing to a culture of cover-up in the Chicago Police Department.”

But the data has some good news: About 80 percent of CPD officers receive fewer than four complaints over the course of their career. Ninety percent have fewer than 10 complaints. “It’s not exactly a few bad apples, but it’s not the whole crop,” she said. “It’s barrels of bad apples and they need to be disciplined, some need to be terminated.”

CPDP is still waiting for archived complaints but in the meantime is planning to use their $400,000 grant from the Knight Foundation to build an app that makes it easy for people to file complaints against police officers and incorporate other datasets into the program.

So far, the data has highlighted a concentration of misconduct complaints in the predominantly black South and West sides of Chicago, Flowers said. The project plans to incorporate other demographics — race, gender, socioeconomics, geography, test scores, and public health data — to get a better look at the “abandonment in these communities.”

Data has the potential to improve police interactions with the public nationwide. Police departments already use historical crime data and heat maps to pinpoint where the next 911 call might come from. But data projects focused squarely at police conduct could empower citizens and improve police policy in a way media stories alone can’t.

“We’re living in a complex age with so much data. There is a need now for projects that use data and allow people to make decisions about their own lives,” the Knight Foundation’s VP of media innovation John Bracken told ThinkProgress. “People can learn about the safety of their own bodies and their own person…We’ve always known that these problems exist but both of these projects take us to a place [based on evidence].”