The Laura and John Arnold Foundation is testing its algorithmic pretrial risk assessment tool — called the Public Safety Assessment (PSA) — in 40 jurisdictions across the country, and despite the fact that those jurisdictions haven’t been using the tool very long and do not have much data on the tool’s efficacy, the foundation told InsideSources it plans to make the tool available nationwide via an undisclosed national technical systems provider this fall.
Criminal justice reform advocates herald pretrial risk assessment tools like the Arnold Foundation’s as a way to fix injustices wrought by the money bail system.
Because the money bail method often discriminates against poorer individuals, many advocates say evaluating an individual’s criminal history and likelihood to appear in court through an algorithm is a fairer, more efficient way of determining who gets released pretrial.
But many of the jurisdictions using the Arnold Foundation’s tool haven’t been using the PSA for much longer than a couple years, and don’t have sufficient data on how efficient, effective or fair the tool is.
The state of New Jersey and Lucas County, Ohio are the only jurisdictions with preliminary data, and while it is positive — showing an increase in pretrial releases and a decrease in pretrial crime — neither jurisdiction has been using the PSA for longer than three years, and in Lucas County’s case, the data was collected only a year after implementation.
New Mexico’s Bernalillo County has been using the PSA for a year and does not have any data available. New Mexico’s Republican governor, Susana Martinez, has said the tool was implemented with “devastating results” as it was “letting dangerous criminals back out on the streets to terrorize communities.”
In San Francisco, an individual who was on felony probation, had been re-arrested, then released pretrial based on a judge’s discernment after evaluating the Arnold tool, later killed a 71-year-old man. More than a third of those released in San Francisco as a result of the assessment were either booked on a new offense or failed to appear, according to a study.
Because there is little data, there is no way to know if other jurisdictions in different cultural and crime climates will see the same success as New Jersey or Lucas County. Furthermore, there is still a fear that the PSA and other similar tools perpetuate racial biases and encourage the pretrial release of dangerous criminals.
Judges ultimately make the final decision in a pretrial hearing, so no matter how heavily the PSA and similar tools influence the decision, the onus of responsibility falls on the judge, not the tool, suggests the Arnold Foundation. In other words, anyone can make a pretrial risk assessment tool and not face legal consequences for the kinds of decisions it fosters.
Many pretrial risk assessment tools are “black boxes” and don’t clarify what kind of factors are taken into account and how the risk assessment is weighted.
The Arnold Foundation has publicly released its criteria and research process for its PSA. The PSA relies on these nine elements:
- Age at current arrest
- Current violent offense
- Pending charge at the time of the offense
- Prior misdemeanor conviction
- Prior felony conviction (misdemeanor or felony)
- Prior violent conviction
- Prior failure to appear in the past two years
- Prior failure to appear older than two years
- Prior sentence to incarceration
The number of prior convictions, of course, are weighted heavier than the other criteria. The problem is, there’s plenty of evidence showing that African-Americans are more often wrongfully convicted of violent crimes than whites — which means that the data going into the tool is inherently skewed.
Jeremy Travis, the foundation’s executive vice president of criminal justice research, told InsideSources that the foundation cares deeply about eliminating racial bias in the criminal justice system.
“We have an abiding concern for racial justice,” he said. “In designing it, we took care to make sure it was race neutral. We do not consider, for example, neighborhood of residence or income or race. But we do look at the data and what the data shows.”
Travis acknowledged that criminal justice data is inherently skewed, but said that if an individual has a wrong conviction on his or her record, he or she should note that in the pretrial hearing.
“If someone’s criminal history is either inaccurate or they want to assert before the judge that a conviction five years ago was a wrongful conviction, they should do that through their lawyer,” Travis said. “There absolutely has been a history of racism in the criminal justice system. The current system against which we have launched this reform effort is inherently discriminatory.”
At the end of the day, though, the PSA does not determine whether an individual is released pretrial: it is ultimately the judge’s decision, and the tool is merely an aid.
“We believe it helps them make better decisions,” Travis said.
Travis also said there’s been a very high demand for the PSA from other jurisdictions despite its newness.
“We have 600 requests for the PSA and we want to work closely with jurisdictions to test it and learn more about the implications and refine it,” he said.
Right now, the foundation partners with jurisdictions to run the PSA on jursdictions’ existing technical systems. But because of high demand, Travis said, it no longer makes sense to manage the PSA this way, which is why the foundation will employ a tech company.
“We had a competition going, but we’re not yet in a position to name that organization,” Travis said.
The foundation is not alone in its efforts to reform pretrial justice: U.S. senators Rand Paul (R-Ky.) and Kamala Harris (D-Calif.) introduced a bill last summer to require states to “replace or reform” their money bail systems.
California’s Senate Bill 10 would end money bail in the state and require jurisdictions to use algorithmic pretrial risk assessment tools.
The National Public Pension Coalition (NPPC), which has accused John Arnold of “attacking” public pensions with his public pension reform efforts, says the Arnold Foundation helped write the bill and claims only the foundation’s PSA fits the bill’s requirements.
“According to the text of the legislation, only his foundation’s system would meet the requirements of the bill,” a NPPC blog post states. “SB 10 would essentially award the Laura and John Arnold Foundation a contract worth tens of millions of dollars a year to implement their system in California. How convenient for the Arnold Foundation that their proposed system is the only one that works according to the proposed legislation they helped to write!”
At the same time, PSA’s algorithm is much more transparent than other risk assessment tools, despite some criminal justice reform advocates fearing the PSA perpetuates racial biases.
“Let’s keep our eye on the bigger picture here: in some jurisdictions, some are assigned to a bail schedule, and that is a blatant violation of the constitutional requirement that bail be set to a person’s ability to pay,” Travis said. “What we want to do is introduce into the current money bail system a way that judges can assess the person standing in front of them, the risk of them coming back or not and whether they’ll commit another crime or not.”