Controversial Policing Software Draws Criticism in New Orleans

After news reports about the city’s contract with Palantir Technologies surfaced this week, the public and civil rights groups have voiced concerns about the technology.

by Matt Sledge, Ramon Antonio Vargas, The Advocate / March 2, 2018

(TNS) — Newly revealed details about sophisticated software the New Orleans Police Department has been using in its fight against street violence have stirred concern among civil liberties advocates and also drawn pushback from local law enforcement officials about the true extent of the program.

The Verge, a website focused on tech news, reported Tuesday that New Orleans had contracted with a Silicon Valley company called Palantir Technologies since 2012 to provide software that allows officials at the department to analyze crime networks and trends.

Various publicly available reports and websites mention Palantir's involvement at the department as part of Mayor Mitch Landrieu's “Nola for Life” murder reduction strategy. But officials have never actively publicized that aspect of the NOPD's approach to violent crime, and few outside of the Mayor's Office or the Police Department seem to have been aware it existed.

Aside from the perceived air of secrecy involved, critics worry the software — dubbed Gotham — could lead to biased policing if not properly overseen, focusing police attention on certain individuals only because a computer suggests they might be a threat, without other evidence to back that up.

City officials insist that Palantir’s software, which the city was licensed to use at no cost, has not violated anyone's constitutional rights. Police said they did not use the program to search social media accounts or to try to predict when or where particular crimes might occur.

Instead, they said, the software’s primary use was to provide a faster way to sift through information that is already available to police but is spread throughout different areas of the department.

Analysts regarded it as a one-stop shop for pulling up and cross-referencing information from field interview cards, police reports, court documents or 911 dispatch data to aid investigations by discovering unseen connections among victims, suspects or witnesses.

They did acknowledge that a pair of city crime analysts used the technology to assess the risk that individuals might fall victim to, or participate in, gun violence. One of those former analysts said little was done with those risk assessments, however.

Still, City Councilman Jason Williams said he was concerned by the lack of transparency about a company whose work has proven controversial elsewhere. He pointed out that little information has been made public about how effective the program has been.

“I would not take a pill out of an opaque, amber bottle that doesn’t have a label on it,” said Williams, who is also a prominent defense attorney. “Right now, that is what this is.”

Landrieu’s office has not responded to questions about the program’s future in New Orleans.

The initial two-year arrangement was extended several times but lapsed last week, and it is not clear whether the agreement will be renewed. The agreement did not have to get City Council approval because there was no money changing hands, Landrieu's office said.

Controversial company

Palantir has attracted attention for both its technological pedigree and its unusual connections in government and politics.

Peter Thiel, the billionaire investor who helped bankroll Facebook in its early days, co-founded Palantir in 2004 with the help of funding from the Central Intelligence Agency’s venture capital firm. Its past work has included helping the Pentagon predict where insurgents in Iraq and Afghanistan might plant explosives.

Palantir ended up doing business in New Orleans after connecting in about 2011 with James Carville, the prominent political consultant and longtime local resident. Carville said Wednesday that the company mentioned an interest in pursuing a philanthropic project, and he suggested tackling New Orleans’ notorious violent crime problem.

That help came in the form of the Gotham software platform, which the city began using in 2012 through a data-sharing and licensing agreement.

Carville has since become a paid consultant for Palantir, which the Verge report said later tried to sell its services to Chicago police by citing the work it had donated to New Orleans.

Nonetheless, Carville insisted his only motivation in recommending Palantir was to help a city struggling with gang violence, especially at a time when New Orleans was reeling from crimes such as the 2012 double murder that claimed the life of 5-year-old Briana Allen.

“I think free is good, and murder is bad,” said Carville, who declined to say how much he’d made from his arrangement with Palantir. “That was my guiding principle.”

Yet the application of advanced software to the world of policing, where officers are supposed to be guided by legal requirements like probable cause, has proven controversial in other cities.

Perhaps nothing is as controversial as the concept of “predictive policing.” Jessica Saunders, a researcher at the nonprofit RAND Corp., said she applies a two-fold definition to the concept: first, that algorithmic or mathematical functions predict either when and where a crime will happen or who will be involved, and second, that police or other authorities act on that information.

If they could be predicted with total accuracy, there would be little opposition to stopping crimes before they happen, Saunders said. But she added that algorithms being tested in some cities today do a poor job of predicting rare events like homicides, and may even amplify human bias.

Mapping crime

Current and former city officials said that in practice, Palantir’s program was used for a limited set of applications that stopped short of “predictive policing." They suggested that the department's reliance on the program was overstated in the Verge report.

The software's primary users between 2013 and 2016 were a pair of crime analysts tasked with modernizing the use of data at the Police Department, which still relied on handwritten reports just a few years ago.

Jeff Asher, a former city crime analyst who now works for the Jefferson Parish Sheriff’s Office, said he was the first analyst to use the Palantir program extensively. The software’s primary virtue was in collecting the “firehose” of data produced by the department and making it easily available, he said.

“The software provides a platform for evaluating all this data that ordinarily would live in 16 different databases," he said. "It just lives in one software program."

One application of the program was “tactical,” according to Asher, who previously wrote a blog about crime for The Advocate.

For example, a detective might call an analyst with a nickname from a witness or a confidential source that was then plugged into Palantir to search police databases, Asher said.

In a more strategic use, Palantir was also used to develop graphic charts of whom gang members associated with and who might be in a gang with them. Yet it still fell to detectives to prove that those associates were actually gang members, Asher said. “The detective would have to find actual evidence," he said.

Palantir "is helping you look at data you’ve already collected,” he said. “It’s not producing new information. It’s not used for probable cause or anything like that.”

Asher and the analyst who followed him, Zach Donnini, also experimented with more advanced applications of the software.

In one May 2016 email obtained through a records request, Donnini tells a Palantir employee that Donnini has created a “gang-member scorecard” in Excel. The scorecard ranks people “according to the number of gun-related events (weighted according to severity) with which a person is associated,” the email says.

The city denied a public records request from The Advocate and other outlets for a copy of the Excel spreadsheet, citing an exception to public records law for “security procedures” and “investigative techniques.”

NOPD spokesman Beau Tidwell on Thursday said the agency is still "determining what use to make of the data set" Donnini created.

As the emails make clear, Donnini had to go outside of Palantir’s software to generate the scorecard, and Courtney Bowman, a “privacy and civil liberties engineer” for Palantir, warned Donnini of the pitfalls of the scorecard approach in an email reply.

“I have some serious concerns about instituting a ranking or numeric scoring approach,” Bowman said. “The looming concern is that an opaque scoring algorithm substitutes the veneer of quantitative certainty for more holistic, qualitative judgment and human culpability.”

Bowman noted that Chicago police instituted their own scorecard, which they called a “heat list.” The list was supposed to identify the people who were most at risk of gun violence, but it also generated concern that police were using gang member “scores” to guide arrests.

Palantir didn't respond to a request for comment about its relationship with the NOPD.

Saunders, the RAND Corp. researcher, said a study she conducted of the initial version of the “heat list” in Chicago found that it did a poor job of predicting the victims or perpetrators of violence. Yet it seemed to correlate to increased arrests for list members, likely because police used it to generate rosters of suspects.

'An academic exercise'

Asher, the former city crime analyst, said he created his own list of those at most risk of falling victim to gun violence in New Orleans. He said he hoped it could be used to guide the delivery of social services. But Asher said he never ranked the individual members of the list, as Donnini appears to have done. He added that Excel played just as large a role as Palantir in generating the analysis.

Asher said he opposes using secretive scorecards to guide arrests. “That’s a terrible way to approach it, because you need community trust, you need community buy-in,” he said.

Saunders said that if nothing was done with risk assessments or with gang violence scorecards, the use of Palantir’s software would not fit her definition of “predictive policing.”

“What you’re describing to me sounds like more of an academic exercise,” she said.

In any event, a number of local officials said authorities’ reliance on Palantir has been overstated.

Ben Horwitz, the NOPD’s director of analysis, and Ronal Serpas, who was the department's superintendent from 2010 to 2014, said the software was one of multiple tools that anti-gang specialists had in their arsenal. Another tool includes officers going online and looking at any social media pages they suspect may have clues about crimes.

Software-wise, Serpas said his officers were much more versed in an IBM-made program called COPLINK. Purchased with federal funds, that program shares some similarities with Palantir. It does less but is also more intuitive, so many more rank-and-file officers had licenses and training to use it, Serpas said.

Asher, as well as a number of other people familiar with the investigations, disputed the notion that Palantir was central to the racketeering cases that resulted in convictions of people charged with belonging to the notoriously violent “39’ers” and “110’ers” gangs. The groundwork for those cases had been done before Asher, the first Palantir analyst, was hired in August 2013, they said.

Donnini, meanwhile, left the NOPD in January and had not been replaced before the most recent extension of the agreement with Palentir lapsed last week.

Officials said the lack of a replacement for Donnini was a strong hint that Palantir wasn’t seen as essential by the NOPD, though Horwitz did say the agency has succeeded in being able to replicate some of the program’s functions without needing the software.

Fear for the future

Despite the assurances from police brass, several officials, defense attorneys and civil liberties advocates expressed concern about the software this week.

There are worries, for instance, about who could gain access to any of the data that have been gathered into the system, although their agreement prohibited both the city and Palantir from sharing data from the program with anyone else, except when compelled by a court order or in broad terms when used for promotional purposes.

While police officials pointed to a number of easily accessible websites and public reports mentioning Palantir’s involvement in New Orleans, it has received none of the abundant publicity that officials have invited for other aspects of their crime-fighting strategy.

Craig Mordock, a libertarian defense attorney who has defended people accused of gang activity, argued that if Palantir factored into any racketeering indictments, it should have been disclosed to defense attorneys.

He also questioned whether the use of Palantir in gang investigations — at any stage — amounted to guilt by association.

Mordock said, “You need some independent information, some probable cause, to get on the police’s radar, and just who a program says you associate with isn’t it.”

©2018 The Advocate, Baton Rouge, La. Distributed by Tribune Content Agency, LLC.


Tech Startups Can Help Revitalize Both Ends of Your Ballot