Three weeks into his new job as commanding officer of Manhattan’s 20th precinct, Captain Timothy J. Malin stared at a map on his computer screen, puzzled. It showed his jurisdiction carved up by streets and parks, with the southern edge encased in an ominous shade of red.
For decades, the New York Police Department has used real-time statistics to chart spikes in violence and calibrate police activity across the city. This map, however, displayed not crime data but something new in the arsenal of police metrics: public approval. The crimson on Malin’s map indicated that some residents in his precinct, the Upper West Side—one of New York City’s wealthiest and safest neighborhoods—reported feeling little trust in his officers. It was Malin’s job to figure out why. “I look at that, and I am like, ‘Okay? What’s causing this?’” he said.
In April, the NYPD informally introduced its public opinion monitor, also known as the “sentiment meter,” during CompStat, the weekly meetings in which top brass interrogate precinct commanders about crime trends. Precincts now receive a monthly “trust score” along with rankings that measure overall satisfaction with police performance and how safe residents feel. The data is culled from questionnaires administered through about 50,000 smartphone apps, including Candy Crush and WeatherBug, as well as traditional landline calls. Facebook and Instagram began to advertise links to the surveys in June.
Since late 2016, the NYPD has been co-developing the sentiment meter with Elucd, a Brooklyn-based startup. (Pronounced e-loo-cid, as in “elucidate”). It’s meant to help residents get more comfortable dealing with police, officials said. “Crime numbers in New York City now are at record lows,” said NYPD Police Commissioner James P. O'Neill. “If we are going to push those numbers down even further, we have to make sure that we have the trust of the 8.6 million people that live in New York City.”
In the age of viral police shootings, generating public trust of cops has become even more difficult. “The relationship between police and the communities they serve is at a crisis in many cities,” said Michael Simon, Elucd’s 37-year-old co-founder. “And they have no way of measuring what success is right now. It’s shocking.”
Simon headed Barack Obama’s data analytics team during the 2008 campaign, and his strategy, or “precision targeting,” helped identify pockets of undecided voters who were then courted by the candidate.
The NYPD approached Simon four years ago to see if he could measure what New Yorkers think about local officers. The partnership started out with a survey of 17,340 New Yorkers, but eventually the team of data experts built the sentiment meter. The police already use algorithms for tasks such as predicting who is likely to get shot and spotting threats on Twitter. Elucd is a tool to help cops get hold of the intangible: human feelings.
Though it’s still in a development phase, the sentiment meter has already drawn the attention of the Los Angeles Police Department; Elucd has been measuring public opinion of the police department across the California metropolis since October. The city of Grand Rapids, Michigan has also signed on, and starting August 1st, Elucd will be working with the Chicago Police Department.
But even the person who ordered the tool’s creation, former NYPD Police Commissioner William J. Bratton, expressed concerns about how to effectively use it. “The challenge is going to be: how to get it down to the officer level, where the officer doesn’t see this as a danger to him, where he’s been evaluated by these anonymous citizens that are commenting on what’s going on in his area of patrol.”
The meter was first shown to New York precinct commanders in the spring of 2017, but many mid-level police supervisors haven’t figured out how to respond to the monthly satisfaction, safety and trust scores. Beat cops have yet to be introduced to the sentiment program, and it’s been discussed during some CompStat meetings.
During a recent tutorial session with a team from Elucd in his station-house office, Timothy Malin wondered about the significance of the low trust score data from the southern part of his precinct. He had some theories: Maybe it was the arrests of pro-immigration protesters in front of the Trump International Hotel & Tower across from Central Park? Or maybe it was tied to the 180 thefts, within a half-mile walk, from several Duane Reade drug stores? “What activities do police do that actually affects sentiment?” he asked. “Great, you are showing me this data. We have to figure out what can I tell my cops.”
Based on the company’s data, the average trust score for New York City during the first three months of 2018 was about 9 percent lower than in Malin’s precinct. The sentiment meter, Simon explained, doesn’t flag a specific number, high or low; it hunts for fluctuations of public attitudes. “In our world, if people feel roughly as safe as they did yesterday that’s a good thing,” he said.
Elucd sends its surveys, packaged as pop-up ads, to hundreds of thousands of smartphones each month. Robocalls to landlines are also made to target older residents. Elucd’s latest advance is getting Facebook and Instagram to run their ads linking to the survey. The meter breaks down responses by precinct sectors, 297 geographical areas that the NYPD created in 2015 as part of its push to give beat cops more autonomy within the pockets they patrol.
Elucd says its data science, protected from public scrutiny because of “proprietary rights,” can guarantee survey results in a given police sector will match the census demographic breakdown of the sector by race, age-group and gender. The tool can also prompt the survey to appear in English, Spanish, Russian, or various Chinese dialects. For the LAPD version, Elucd added Korean and Tagalog.
More than 200,000 unique respondents have filled out the questionnaires since October 2016, according to Elucd. In Los Angeles, where the meter was launched during the fall, the count is already more than 50,000. The survey is 10 questions long, and is constantly adjusting. A December questionnaire asked whether respondents felt safe in their neighborhoods, whether local officers treat residents with respect and whether cops take action based on facts, not personal biases. Respondents are also asked to share their age, gender, race and whether they live in the area where they are taking the questionnaire.
Elucd says little about how survey responses translate into measurable scores, and cites privacy issues when asked about who is completing questionnaires. Police officials, nonetheless, say they are confident that Elucd’s data science works. “What I see makes sense. I am not a mathematician. I am not a statistician, but we’ve asked many deep questions about their methodology,” said O’Neill.
But mid-level supervisors say that top-brass has left them rudderless.
“If precinct commanders are telling you they don’t know what to do with that information, it’s quite obvious that the department needs to do more education,” Bratton said.
NYPD officials say they are simply giving cops the freedom to come up with their own creative solutions, an approach echoed by police in Los Angeles. “They are responsible for their area of the city,” said LAPD Deputy Chief Sean Malinowski.
Grand Rapids, Michigan, population about 200,000, is Elucd’s smallest client. It’s also the only law enforcement agency with a clear response plan. Police Chief David Rahinsky says he will assign more foot, bike and horse patrols to neighborhoods with negative perceptions of police. Those places will also see more gimmicky events, such as officers giving away car-seats and backpacks, along with “police car karaoke,” where officers invite kids to sing over a patrol car loudspeaker. “In my mind, there’s a lot we can do with this tool,” Rahinsky said.
In Elucd’s cramped WeWork space, near the foot of the Brooklyn Bridge, a millennial-friendly office co-share with craft beer and La Colombe coffee on tap, Michael Simon fielded questions about his creation. Simon’s staff of 10 — with more positions to fill — includes mainly data scientists and software engineers. No one has a policing background.
“If I can be totally candid, one of the reasons that NYPD has been slower to move is because they don’t know what to do in response to it,” Simon said. “No one has any real training on how to build trust in the right direction.”
The NYPD’s curiosity about attitudes towards cops grew from a 2013 shift in city politics. New Yorkers had just elected Mayor Bill de Blasio, who promised to end aggressive police encounters, known as stop-and-frisks, and to improve public trust of cops. In New York, Bratton had been the architect of CompStat, and during the aughts, he moved on to be police commissioner in L.A. De Blasio brought him back to New York and to a new conundrum: “Crime is down dramatically, quality of life has improved so dramatically, why are we still getting negative sentiments from certain neighborhoods, from certain streets, from certain people?” Bratton said.
The news cycle began to fill with Black Lives Matter protests against police brutality in city after city. A gunman from Baltimore, angered by the headlines, traveled to Brooklyn and fatally shot two NYPD officers sitting in a parked patrol car. Police polling became a NYPD priority. “You need to know, as close in real time, ‘is this getting better? Or getting worse?’” said John Linder, Bratton’s longtime consultant who served as the bridge between the police and the pollsters up until June.
The NYPD has no contract yet with Elucd, but gives the startup unprecedented access to its management and to its anonymized crime data. Elucd also works for the LAPD without a contract, and is trying to raise private money to fund the tool. Likewise, a spokesman for the Chicago Police Department said money from yet to be named foundations will pay for the meter. Grand Rapids, thus far, is the only police department that has officially contracted with the startup, using drug forfeiture funds to pay $150,000 for two years of service.
Elucd says it has a budget in the multi-million dollar range, which includes a combined $1 million in seed money from incubator Y Combinator and the Omidyar Network, an investment firm run by the founder of EBay. The NYPD connected Simon and his business partner, Saul Shemesh, to the nonprofit New York City Police Foundation, which has given more than $289,000 towards the endeavour. The New York City Mayor’s Office of Criminal Justice awarded Elucd $1.4 million last spring to test out the sentiment meter in public housing developments.
Excitement over the sentiment metrics is not universal. An enclave in the 30th Precinct, which spans north Harlem up to the Dominican community of Hamilton Heights, ranked in the top five during the first three months of 2018 for its rising trust scores. But the precinct captain, Lourdes Soto, said she was unaware of her precinct’s achievement, and hadn’t told her beat cops about the tool. “I don’t know how they collect the data,” said Soto during a public precinct meeting in March. “I have to explore it first.”
Nine miles away in the 7th Precinct, which covers the Lower East Side, there were also concerns. Elucd identified an eastern section of the precinct as having one of the city’s lowest scores for trust, and a declining safety rating. The area is on the outskirts of a trendy Manhattan neighborhood clustered with several public housing complexes and a concentrated population of Latinos and Chinese immigrants. Store workers and residents were puzzled that the Lower East Side ranked so low compared to higher crime precincts.
“Something is off,” said 38-year-old Selina Balestier, a recent MBA graduate from a cop family who lives in Vladeck Houses, a public housing development where her mother is the resident association president. “How are you verifying the integrity of this data?”
“We need to come up with ways to better explain this,” Michael Simon conceded. Later, however, he defended the lack of transparency: “We promised our respondents that they would be anonymous.”
Elucd’s secrecy can be costly. The Research and Evaluation Center at John Jay College of Criminal Justice shaved off nearly a third of Elucd’s $1.4 million grant and hired a think tank to dig into Elucd’s data science. The National Opinion Research Center issued 18 recommended changes. Elucd adopted a few of the suggestions and said the feedback, for the most part, was outdated.
Still, the city of New York is expected to give Elucd another $1 million to keep the NYPD’s sentiment meter running for at least another year. As New York continues to build out the tool, Elucd gets to keep its big selling point — an alliance with the world’s most famous police department. “What we are trying to do here: is to increase outreach, increase engagement, create relationships, and build on those relationships,” O’Neill said. “I think this is important information.”
