There’s widespread agreement among justice reformers that decisions about whether to jail people accused of crimes before trial should not be based on whether they have enough money to post bail.
But the growing use of an alternative approach that uses “risk assessment” tools based on data about an individual’s past history in the justice system has come under fire from civil rights advocates. They argue that such apparently neutral science-based methods not only preserve the baked-in racial disparities of the justice system—but can in some cases exacerbate them.
A paper commissioned by the Pretrial Risk Management Project of the John D. and Catherine T. MacArthur Foundation, offers a powerful defense of the civil rights critique, mustering arguments that suggest even the most carefully calibrated algorithms ultimately are weakened by the structural inequalities that have long beset the U.S. criminal justice system.
The paper, written by David G. Robinson and Logan Koepke, respectively managing director and senior policy analysis of Upturn Inc., a Washington, DC-based nonprofit, as part of the MacArthur Foundation’s Safety and Justice Challenge, argues that risk assessment should for all practical purposes be abandoned.
Instead, it says, courts across the nation should end the use of pretrial detention altogether for individuals arrested for nonviolent and relatively minor offenses—a practice that according to some estimates keeps some 450,000 Americans who haven’t been convicted of anything behind bars daily.
At best, the authors wrote, “policymakers can ensure that pretrial detention is limited to a narrow set of serious charges, and that for people who are so charged, a specific and thorough hearing is required before detention of more than a few days can be imposed.”
The authors concede that it may be politically difficult to both phase out money bail and reverse the increased use of risk assessment tools as a substitute for bail without fundamental changes in the way the U.S. justice system operates at local levels—beginning with law enforcement.
But they argue that it is essential to ensure that risk assessment doesn’t distract attention from efforts to end inequities in incarceration.
“Many in the civil rights community fear not only that risk assessments may be harmful (or offer little benefit), but also that much of the current wave of reformist energy concerning pretrial justice will be spent implementing these systems (that may yet have to be reformed years later), rather than advancing other changes that would have greater benefit for liberty and racial equality,” they wrote.
The debate over the use of risk assessment has been building for some time, but the Upturn Inc. paper may be the most serious critique advanced so far—since it’s combined with an alternative “vision,” as the authors put it, that calls into disrepute the entire practice of pretrial detention.
The key elements of the alternative vision include:
- Ensuring that detentions before trial are rare and only used when there is clear evidence that an individual will flee the court’s jurisdiction, or will endanger public safety if released from custody;
- Assessment of “dangerousness” should not depend on a risk assessment score, but on a transparent process involving all the stakeholders in a local justice system—including community members who can serve in an oversight capacity to ensure racial bias is eliminated in the decision-making;
- Where risk assessment tools are employed, they should only serve as guidelines to determine whether a further hearing before a judge on the individual’s case is necessary.
In fact, many judges have said that is how they use risk assessment in their courtrooms today: as a guide that aids them in applying their own experience and discretion (or “gut feeling”) in making decisions.
But such “gut” decisions, taken by thousands of judges over the decades, are precisely what fueled the emergence of scientific scoring and algorithms in the first place. Opponents of the money bail system have argued that judges with crowded dockets are too often swayed by prejudice, community pressure or aggressive District Attorneys—sometimes purposefully setting a high bail out of personal pique.
Research has shown that the poor are doubly victimized by the system. Forced to stay behind bars because they or their families can’t afford even the minimal amounts to spring them from custody, they may lose their jobs and face disruptions in their personal lives that may force them into a downward spiral of criminal behavior, even if their original offenses were minor or they were completely innocent to begin with.
Using a system that “scores” individuals charged with an offense on the basis of algorithms derived from their criminal, employment and personal history to determine whether they can be released, risk assessment advocates say, is a fairer way to determine whether they can be released.
Civil rights advocates say that’s mostly wishful thinking.
The authors offer the example of New York City, where a risk assessment tool was redesigned to utilize arrest data between 2009 and 2015. Those happened to be years when police “stop, question and frisk” practices were at their height, resulting in the detention of a disproportionately high number of young black males often for minor offenses.
Those recorded encounters with cops increase the odds of an individual receiving an algorithmic score that indicated he was a higher risk.
“Arrest statistics are greatly influenced by what citizens choose to report and what police departments and officers choose to enforce, both of which can vary widely from one neighborhood precinct to another,” the authors wrote.
But an even more telling criticism of algorithmic tools raised by the paper was what sort of “risk” they actually measured.
“Failures to Appear” Don’t Constitute a Flight Risk
In most cases, the tools used a composite score based on relatively minor violations, such as missing a previous court appointment, as a reason to raise doubts about an individual’s likelihood of making a court date. In fact, such “failures to appear” could be the result of something as simple as a missed bus, an inability to arrange child care, or forgetfulness—none of which suggests the individual is a real “flight risk,” the authors said.
“A series of research studies suggests that simple reminders—either delivered by postcard, voicemail, or text message—can dramatically increase the rate of appearance at court dates,” they said..
The authors cited research showing that the majority of people given such additional nudges will show up in court.
Algorithmic tools that measured how well individuals are likely to meet their obligations, rather than how likely they will fail—tools that take into account the level of community support services available to help them—could offer a far better measure of fair and unbiased justice, they said.
The authors maintained that most tools make no attempt to forecast the outcome that needs to guide judges, namely “the accused person’s likelihood of success after available steps are taken to maximize the likelihood of successful appearance and non-arrest throughout the pretrial period.”
“This matters,” They added, “because there is good reason to believe that supportive services and, in some cases, monitoring conditions can and do materially reduce risk.”
The authors conceded that supporters of risk assessment tools are not likely to give up on an approach that has been considered an effective antidote to America’s much-attacked money bail system.
Nevertheless, they said, it was important to disabuse those supporters of the idea that there was no other alternative.
“Concrete and powerful alternatives to both money bail and group-based actuarial assessment of risk are available, and need to be considered,” they wrote.
The full paper can be read here.
Stephen Handelman is editor of The Crime Report. He welcomes comments from readers.