Jaquan Lugo stood stone-faced and somber inside a circular, wood-paneled courtroom on a Thursday afternoon in Paterson, New Jersey, as Superior Court Judge Donna Gallucio considered her options.
Just four days prior, the 22-year-old and two other men were arrested in Paterson, accused of six counts of attempted murder and various gun charges after a predawn drive-by shooting left a 17-year-old girl with a life-threatening wound near her lung. An off-duty officer heard the shots just a few blocks away and gave chase to the fleeing vehicle, a 2002 Jaguar, as someone inside the car fired back at him. Lugo and two of the car’s other occupants—Kashief Davis, 24, and Andre Green, 20—allegedly got out of the car and tried to escape on foot before they were caught and brought to Passaic County Jail.
The victim’s friends and family crowded into the courtroom benches to hear the decision on Lugo’s fate, but Judge Gallucio wasn’t there to determine his ultimate sentence. She was, instead, deciding whether Lugo should spend the months leading up to his trial in the county jail or at home.
Standing shoulder-to-shoulder with his client, Lugo’s lawyer, Gregory Aprile, argued for pretrial release, imploring Gallucio to consider one crucial factor: A new algorithmic tool that purports to predict a defendant's likelihood to reoffend, or to fail to appear in court, ranked Lugo as fairly low-risk. On an escalating scale of 1 to 6, it rated Lugo a 2 for failure to reappear and a 3 for likelihood of reoffending.
“They aren’t arbitrary numbers,” Lugo's attorney said of the so-called Public Safety Assessment, or PSA. “It was the result of millions of statistics from around the country.”
This may seem like an unusually technocratic approach to public defense. But it’s not so unusual anymore, at least not in New Jersey, where the state has recently undergone a holistic technological transformation of its arcane court system, all in the service of eliminating the use of bail statewide.
New Jersey is far from the only state government taking a critical look at the centuries-old bail system in America. Politicians on both sides of the aisle, from California senator Kamala Harris to Kentucky senator Rand Paul, argue that bail sets up a two-tiered justice system, in which the wealthy can buy their way to freedom while the poor remain locked up until their day in court comes. In 2016, the Department of Justice, under President Obama, also issued a Dear Colleague letter to state and local courts around the country, advising them that courts “must not employ bail or bond practices that cause indigent defendants to remain incarcerated solely because they cannot afford to pay for their release.”
As it turned out, that described a large percentage of people who have spent time in New Jersey jails, according to one 2013 study by the New Jersey Drug Policy Alliance. The advocacy group found that some 75 percent of New Jersey’s jail population at any given moment was simply awaiting trial, and 40 percent of jailed people were there because they couldn’t afford $2,500 or less in bail. On average, people spent 10 months in jail before even getting to trial. Meanwhile, because New Jersey prohibited even the most violent criminals from being detained without bail, judges often had to set exorbitant bail amounts to keep violent offenders off the streets; sometimes, those people made bail anyway.
“You had a situation where if you had money in New Jersey, no matter how serious your offense was, you could pay and walk away pending trial,” says Roseanne Scotti, the Drug Policy Alliance’s senior director in New Jersey. “If you didn’t have money, no matter how minor your offense was, you sat in jail for months.”
That system also meant people could pay for-profit bail bondsmen a small fraction of the 10 percent of their bail they needed to pay to get out of jail, only to owe even more money to the bondsmen over the longterm. Not only did that create a predatory industry but, says Passaic County Assignment Judge Ernest Caposela, “A lot of dangerous people were making it out on bail.”
Driven by advocates like Scotti, as well as the American Civil Liberties Union, New Jersey Governor Chris Christie signed the so-called Bail Reform and Speedy Trial Act, which went into effect on January 1 and is designed to virtually eliminate bail in the state. Of all of the attempts to curb the use of bail nationwide, New Jersey's approach is perhaps the most audacious. Pulling it off has required the state to harness the power of tech, not only to move people through the system more quickly but also to analyze who is least likely to pose a risk to society upon release.
Just months in, the experiment has already made an impact. New Jersey saw a 19 percent reduction in its jail population overall between January 1 and May 31 of this year, with just eight people being held on bail throughout the entire state over that time period. Others are either being released with certain conditions or detained without bail.
This shift has also prompted a number of lawsuits, including one filed by the mother of Christian Rodgers, a 26-year-old man who was allegedly murdered by a man named Jules Black, just days after he was released from jail without bail earlier this year. That suit targets both Christie as well as the Arnold Foundation, the nonprofit organization that designed the PSA tool.
Perhaps unsurprisingly, the case is backed by the bail-bond industry, including reality star Duane Chapman, better known as Dog the Bounty Hunter. They argue that tech tools like the PSA offer a dangerously inadequate way of distributing justice. Scotti argues the bail-bond industry cares more about its bottom line than the public well-being. Now, as states across the country look to tech tools to reform their jail and prison systems, New Jersey’s experiment illustrates both the promises and pitfalls of using technology to determine who does and doesn’t remain behind bars.
Something Like Progress
“Today is a great day,” says John Harrison, clasping his hands together. Harrison helps run the county’s newly created pretrial services division. Housed in a charmless, cubicle-filled office adjacent to Paterson’s historic courthouse, it’s the group responsible for running risk assessments on every person who enters the county jail.
The source of Harrison’s delight? Today his team will move 23 people—more than average—through their pretrial hearings. They include a woman accused of prostitution, another accused of credit card theft and burglary, another accused of child endangerment, and a man charged with assault and disorderly conduct. One by one, their faces will appear on a television screen inside Judge Abdelmageid Abdelhadi’s courtroom, where he will rattle off their rights, their charges, and their PSA risk scores with the quick-tongued elocution of an auctioneer. Last year, this would be the part where the judge sets bail. Now, after the reform efforts, it’s where he tells each defendant whether they’re being released today no strings attached, released today with some type of monitoring, or whether the prosecutor is filing to have them detained until trial. The PSA score is one of several factors he considers. Within three minutes flat, he’ll wish each defendant good luck, before calling the next in line and reciting the same script. In Paterson, a once-illustrious industrial town now riddled with crime, that counts as progress.
When Harrison started working for the courts back in 1995, this whole process dragged. When a new person came into the jail, Harrison and others like him would have to go to the court’s physical Rolodex machine and pull index cards on a given defendant’s record, in order to help the judge make bail decisions. That task was made all the more complicated, given the fact that so many defendants go by a slew of aliases that could be tough to keep straight. “God help you if the Rolodex machine broke,” Harrison says.
The longer it took to compile all that information, the longer people waited in jail before they ever even got a chance to plead their cases.
Today, when law enforcement arrests someone in New Jersey, they take his fingerprint and enter it into their Livescan system. That system automatically sweeps both the FBI’s database and the statewide database for a person’s complete criminal record. Last year, the state began using an IBM tool to search through its 40 million records and rule out possible duplicates, to ensure that one John Smith isn’t picking up some other John Smith’s rap sheet. Likewise, if two records for John Smith are virtually identical, except their birthdays are one digit off, the system will look at the statistical likelihood they’re actually the same person. The IBM tool also searches for possible aliases, where records filed under different names might, say, contain the same Social Security number and date of birth.
Agencies decline to release information about algorithms used for criminal justice, social welfare, and education.
Cities are using techniques borrowed from public health and machine learning to figure out what to do with people after they get arrested.
Even as the president promises a crime crackdown, a deep data dive finds jail populations are soaring fastest in areas that backed him heavily.
To illustrate just how useful that is, Harrison pulls the rap sheet of a man who goes alternately by Thomas Ali, Barry Simpson, and a handful of iterations of those names, whose fingerprint matches 70 separate arrests. In the old days, it could have taken Harrison’s team weeks to track down that information. Under the new system, it takes about two minutes, including the time it takes to calculate the PSA score. Mr. Ali (or Simpson) scored a 6-6, the highest risk on both the failure to appear and the likelihood of committing a new crime scales. It’s unlikely he would walk free before his trial, even under bail reform. But, the thinking goes, condensing that decision-making process into a matter of minutes gives the courts the time they need to assess other defendants---ones who might very well walk free--more quickly.
The judiciary’s IT department has added other time-saving tools to the mix, too. It built a so-called virtual courtroom, so judges can hold pre-trial hearings on weekends when the courts are closed. Now, the team is tinkering around with voice-recognition technology that can save judges time when filling out detention orders. “We’re trying to look at the technology from the standpoint of eliminating as much of the clerical functions of whoever touches the case as possible,” says Judge Caposela.
Trial and Error
By far the most controversial element of the state’s technological transformation is the risk score itself. Similar assessments have popped up across the country, from Miami to San Antonio, put to use for everything from bail reform to decisions on which defendants most need mental health assistance. Not all of these algorithms are created equal. One ProPublica investigation found that a tool called Compas, which was used in sentencing decisions, overwhelmingly rated black defendants higher risk than white defendants.
“Algorithms and predictive tools are only as good as the data that’s fed into them,” Ezekiel Edwards, director of the ACLU’s criminal law reform project, recently told WIRED. “Much of that data is created by man, and that data is infused with bias.”
The Arnold Foundation, which designed New Jersey’s PSA tool, now used in several states and dozens of local jurisdictions, attempts to sidestep that problem by vastly limiting the number of risk factors it considers to eliminate racial or gender indicators. The Foundation analyzed 1.5 million pre-trial records from across the country and narrowed its algorithm down to look at just nine risk factors: the person’s age at the current arrest, whether the current offense is violent, pending charges at the time of the offense, prior misdemeanor convictions, prior felony convictions, whether those prior convictions were for violent crimes, prior failure to appear in the past two years, prior failure to appear instances that are older than two years, and prior incarceration sentences. Unlike other tools, it doesn’t weigh factors like education, income, or employment, any of which might disadvantage certain demographic groups.
“An effective risk assessment must be gender and race neutral,” says Judge Caposela, one of the PSA’s early evangelists in New Jersey. “The more risk factors you have, the less likely you’ll be able to eliminate gender and racial bias.”
Even so, Leila Walsh, a spokesperson for the Arnold Foundation, cautions that the PSA scores are meant to serve merely as a baseline for the courts. "The decision about what to do always rests with the judge," says Walsh. States including New Jersey often couple the PSA with another set of parameters that could, for instance, flag defendants who have been charged with domestic violence, or who have been re-arrested while out on bail in the past.
The Arnold Foundation's stripped-down risk assessment has still faced a fair bit of backlash. As WIRED recently reported, researchers have criticized the Foundation for making municipalities sign a confidentiality clause. Peter McAleer, a spokesperson for New Jersey’s courts, says the state has no such agreement with the Arnold Foundation.
This lack of transparency has become central to lawsuits surrounding the use of the PSA. Jules Black, the man accused of murdering Christian Rodgers, had been in and out of the New Jersey county jail system 28 times since 1994, according to the suit. His most recent arrest was for unlawful possession of a firearm. During a press conference about the case, Dog the Bounty Hunter questioned why a man with such a record would be released. “The Arnold Foundation has a questionnaire. Guess what? You must not have asked the right question,” he said.
Even Judge Caposela acknowledges there’s some truth to that. The PSA takes what he describes as a “neutral view” of gun possession. Because it was trained on data from across the country, and because some states have far more lax gun regulations than New Jersey does, the PSA doesn’t consider mere gun possession as an outsized risk. It wasn’t until after the Rodgers murder that the state's attorney general issued new guidance, directing New Jersey prosecutors to seek pretrial detention in any gun-related cases.
"We extend our deepest condolences to the Rodgers family for the tragic death of Christian Rodgers," Walsh says. She acknowledges that the PSA is not a "perfect system," but neither, she argues, is bail. "The traditional for-profit, money bail system is deeply flawed, unjust, and inefficient," she says. "We should not allow those who make their living in the for-profit, money bail industry to use tragedies to deflect attention from the urgent need for reform."
There’s an argument to be made that an over-reliance on the algorithm may have impeded the court’s decision to release Black. Then again, one could counter that under the old system, Black would have made bail regardless. He had, after all, already been in and out of jail 28 times. He had bought his way out before. He could probably do it again. Under the new system, that’s not how it works.
Which brings us back to Lugo. While his attorney asked the judge to consider his client’s relatively low PSA score, the prosecutor reminded the judge that the PSA system also flagged Lugo as “no release recommended.” That’s because of the violent nature of the crime—attempted murder—and the gun charges. "Of paramount concern is safety to the community," assistant prosecutor Nubar Kasaryan told the judge.
He reminded the judge of Lugo’s record, which includes a child-abuse conviction for which he was just released from state prison in March. And he explained that the drive-by shooting victim is still in critical condition at a local hospital. Judge Gallucio pursed her lips and furrowed her brow, before deciding that Lugo’s release would “pose a significant risk to the community.”
Lugo and the other alleged shooters will have to await their day in court behind bars. But the same law that will keep them there will also ensure that others accused of lesser crimes may not have to.