New Jersey’s quest for unbiased bail is righting the scales of justice
Algorithms review millions of case files to reduce bias
It was nearly a decade ago when the New Jersey judiciary came to a critical conclusion: bail is broken.
A troubling number of pretrial defendants were spending days, even weeks, in New Jersey jails awaiting charges. The reason? They could not afford bail.
Even a few hundred dollars was a barrier. A study conducted by the New Jersey Court System found that while waiting for their court dates, 15,000 defendants served jail time with assigned bails of $2,500. This meant the defendants were unable to pay 10% of that amount ($250) to afford temporary release from state custody.
In fact, research conducted by the US Department of Justice found that three in five people in United States jails are there awaiting trial. Since 2000, 95% of all jail population growth in the US has been attributed to pretrial defendants, those individuals who have been accused but not convicted. And yet, they remain in jail until their trial—often for reasons more to do with the balance of their bank account than for the charges related to their arrest.
For years, the US bail system has been criticized by legal experts for having practices that can be unjust, unequal and prejudice. Critics of bail say it’s a system that is unfairly harsh on lower-income Americans, and that disproportionately affects racial minorities. For example, Black defendants experienced a 25% increase in the odds of pretrial detention relative to white defendants, according to a 2015 study in the academic research journal Criminology & Public Policy, published by The American Society of Criminology.
And the potential damages of a rampant bail system are considerable, because any amount of jail time harms the lives of those imprisoned. After just eight hours confined, the imprisoned are more likely to lose their jobs or homes, according to a study of people incarcerated in the New Jersey Court System. After days or weeks, lives can begin to unravel.
For Jack McCarthy, chief information officer for the New Jersey Court System, and Kristin Rangel, the agency’s project manager of database administration, the process needed an overhaul. Could the New Jersey judiciary use technology to leverage data and reduce the number of pretrial defendants jailed because of low-income status?
Managing pretrial risk
Over the past seven years, McCarthy and Rangel have tried reshaping the New Jersey courts’ bail system by managing the risk of pretrial defendants, using IBM data management tools to help do so.
Risk assessments are based on a research-tested algorithm developed by the Laura and John Arnold Foundation that works in tandem with IBM data solutions. To create the risk assessment, the system culled 40 million records and 200 million cases, surfaced at the instant of a judge’s click. These risk reports were used to provide scoring and release recommendations to aid a judge’s final decision.
The risk assessment system could create recommendations in three minutes—aiding New Jersey courts to meet target deadlines for case disposition. In the place of bail, the system now provided guidance reliant on the defendant’s data, including pending charges, prior convictions and previous failures to appear in court, to help officials determine release.
From inception to launch, “show the math” was the team mantra. Information about how the algorithm works and criteria used to derive risk assessments were made publicly available. From every angle, the court was working to eliminate bias. “We sought to automate steps, eliminate the interpretation of data and take all bias out of decisions as much as we could,” McCarthy said.
Maintaining quality and mitigating bias in the courts’ datasets was also a priority, because release recommendations could only be as trusted as the defendant data that derived those recommendations.
Scanning millions, saving millions
The New Jersey Courts System turned to IBM data management tools to reliably retrieve case records, which helped maintain the quality of the defendant data used to create risk assessments. The IBM tools presented a 360-degree view of defendant information, which gave New Jersey administrators a vantage point from which to mitigate bias in their datasets and contributed a sense of trust in the risk assessment’s findings and recommendations.
No matter if the defendant entered through New Jersey’s criminal, civil or family court—no matter if the defendant’s data was scattered across these judiciary divisions digitally—the IBM technology linked these systems and brought forth a more complete view of each defendant for the judge’s consideration.
The goal was a uniform system that worked for everyone. “To make the world a better place, we need to standardize as much as we can,” Rangel said. “Ensuring data is processed the same way every time.”
Within seven years, New Jersey experienced a near 40% drop in jail population with no measurable increase in crime rate, while the courts saved USD 10 million.
While a court entirely devoid of bias remains a mission of the future, the route to eliminating bias has been made smoother thanks to the technology the New Jersey Courts System has pioneered. “The road is paved, now you only need to drive on it,” McCarthy said.
Today, the risk assessment technology that started life in New Jersey courtrooms has inspired digital transformation across the whole of New Jersey government.
An exciting prospect for this forward-looking team, who, no matter where the AI journey takes them, in their hearts know the destination. “We’re driving to the ultimate goal of fairness,” McCarthy said.