A coalition of civil rights groups is calling for an end to the use of computer risk-assessment tools such as the one New Jersey uses in pretrial release decisions, saying these can worsen racial disparities in the criminal justice system and lead to greater incarceration.
At the same time, though, New Jersey’s recent move away from a system that depends on monetary bail has led to a drop in the number of people detained while awaiting trial. The state’s new bail reform also allows for a hearing where those who are ordered held can argue for their release, one coalition member noted. In the first six months of this year, nearly two-thirds of those defendants initially ordered detained wound up being released pending trial, according to data from the state Administrative Office of the Courts. The number of people held pretrial dropped by almost 30 percent from the start of criminal justice reform through June 30.
Still, that does not mean civil rights activists support New Jersey’s use of an algorithmic-based risk assessment. The state’s courts still have not released data on the number of people detained according to race and ethnicity over the last several years, in order to compare the pre-reform and post-reform periods. Without that data, it is impossible to see whether there has been a change in the racial disparity of those being held for what could be a significant amount of time while awaiting trial, which could take up to two years under the law.
“We need good data so we can measure what the system has done for the racial disparities,” said Alexander Shalom, senior supervising attorney at the American Civil Liberties Union of New Jersey. “If it’s done nothing, that’s unacceptable. If it’s made the disparity worse, that’s even more unacceptable. Or it may have improved matters.”
The wait for good data has been ‘too long’
“It’s been too long a wait,” continued Shalom, a supporter of the reform who has also represented defendants in pretrial detention-related cases that have helped refine the system. “We are ready to hear what the AOC has to say. If it’s working, we need to know. We know a lot of other states looking to New Jersey want to know.”
Peter McAleer, a state judiciary spokesman, said the office is continuing to work on putting the data together, adding that gathering comparable data for years prior to 2017 has been a major challenge.
New Jersey has the greatest black-white racial disparity among its prison population in the nation, according to the Sentencing Project.
The coalition of more than 100 civil rights, digital justice and community organizations operating under the Leadership Conference Education Fund supports an end to monetary bail, which winds up keeping many people in jail not because they pose a danger, but because they cannot afford to post bail. But they released a shared statement of civil rights concerns over the growing use of risk assessments to determine pretrial detention.
Vanita Gupta, president and CEO of The Leadership Conference Education Fund, urged that courts forgo the use of assessments and instead focus on removing other racial biases in policing and the criminal justice system.
“Pretrial assessments must not be seen as a panacea for criminal justice reform,” she said. “The algorithms are only as good as their data inputs. Biases in the data sets many not only be replicated, but they may be exacerbated.”
The group supports due process to make decisions about release and not the use of algorithms, but urges courts using these tools to follow six principles to ensure their proper use. Two of those hold that if these tools are to be used, they should be designed and implemented in ways that reduce and ultimately eliminate racial disparities and should not automatically allow pretrial detention without a full hearing. New Jersey’s tool does not explicitly reduce racial disparities, although that may have been its result, but the state does require a full hearing in detention cases.
“Pretrial detention reform that addresses the injustice of people being jailed because of their poverty is urgently needed, but substituting risk assessment instruments for money bail is not the answer,” said Monique Dixon, deputy director of policy and senior counsel for the NAACP Legal Defense and Educational Fund. “Biased policing practices in communities of color result in racial disparities in the data risk assessment tools rely on, making black and brown people look riskier than white people. Pretrial detention reform must include solutions that will reduce the disproportionate incarceration of people of color, not worsen it.”
Algorithm calculates danger
These tools, including the one from the Laura and John Arnold Foundation in use in New Jersey, include multiple pieces of data about a person to forecast an individual’s likelihood of appearance at future court dates and his risk of re-offending.
McAleer said that the more data points a tool uses, the more likely it could be racially biased, and
New Jersey’s Public Safety Assessment uses the smallest number of data points of any state. All but one of the nine data points relate to the defendant’s current and prior charges and failures to appear, with the other point the defendant’s age. Other tools may use less relevant data, including whether a person has failed a drug test in the past, in recommending whether a person is released or detained.
Under the 19-month old criminal justice reform, which New Jersey voters passed as a constitutional amendment overwhelmingly in 2014, the algorithm calculates what kind of a danger a defendant would be to others and the likelihood of his appearing for trial. Depending on the PSA score, a person may be released with no conditions, with the need to check in by phone or in person with a pretrial services officer, or with an ankle bracelet and a limit on his ability to leave the home. Alternatively, the defendant may be held until trial.
Jeffery Robinson, deputy legal director for the American Civil Liberties Union, noted that New Jersey’s tool is used as part of a broader process that includes one or more hearings during which a judge has the discretion to accept the PSA’s recommendation for release or detention or make his own decision.
‘Troubling’ imposition of monitoring conditions
“I’m hesitant to say New Jersey should not be looking at its tool, but other tools are worse,” said Andrea Woods, an ACLU staff attorney. “New Jersey’s reforms provide a robust due process, including a right to discovery in pretrial detention hearings. Care is taken to vindicate people’s constitutional rights.”
But Logan Koepke, a policy analyst with Upturn, an organization focused on equity and justice in the design, governance, and use of digital technology, said it is “very unclear to say the pretrial population went down because of criminal justice reform,” given that the decline in New Jersey’s pretrial jail population predated the implementation of bail reform. McAleer said the earlier decline in the jail population was due to the implementation by the courts of some of the principles of the reform prior to its full implementation on January 1, 2017.
Koepke also called “troubling” the imposition of strict monitoring conditions on a significant number of those who are released. The latest AOC statistics show that about 43 percent of those released in the first six months of 2018 had to make a weekly phone or in-person contact with pretrial services personnel and 1,375 of those also were subject to some electronic monitoring. This is “causing the state to spend a lot of money.”
Shalom said New Jersey’s reform program, while imperfect, is an improvement over the “deep racial bias built into the bail system.” There was no way to effectively address that bias without systemic reform.
“No matter what the PSA tool does, there will be ways to improve the system,” he said. “Now we have a system that relies on robust due process and a broad presumption of release.”
Can tools really determine risk?
The data used in these algorithms can be wrong or prejudiced, said Rachel Foran, managing director of the Brooklyn Community Bail Fund. She said the New York City courts’ risk-assessment system judged 88 percent of her fund’s more than 3,000 clients a flight risk and recommended their detention. But on their release — the fund paid their bail — 95 percent made all their court dates.
“Our results … strongly call into question the ability of any such assessment tool to determine risk; the one used in New York City couldn’t have gotten it more wrong,” Foran said. “Presumptively innocent people should be free to return to their communities and fight their cases from a position of freedom, not held in a cage whether because of the inability to afford bail or because of a biased risk assessment instrument.”
The Laura and John Arnold Foundation issued a statement in which it supports at least some of the principles, particularly transparency of use and public accountability, as well as the desire to reduce racial disparities in the criminal justice system and reduce unnecessary pretrial detention. But the foundation said its assessment is useful in helping judges make their decisions.
The coalition’s “description of risk assessments as tools that ‘can defer the responsibility of determining who to detain pretrial and who to release’ misconstrues the role of risk assessments,” the statement reads. “Risk assessments, such as the Public Safety Assessment (PSA) developed by LJAF, do not make pretrial release decisions or replace a judge’s discretion. They provide judges and other court officers with information they can choose to consider — or not — when making release decisions. We believe — and early research shows — that this type of data-informed approach can help reduce pretrial detention, address racial disparities and increase public safety.”