At a time when Democrats and Republicans in Congress can’t agree on just about anything, there is one issue that unites them: the urgent need for criminal justice reform.
A Senate bill on the issue has attracted an impressive 37 co-sponsors from both sides of the aisle. The Sentencing Reform and Corrections Act has gained support from figures as politically diverse as the Koch brothers and President Obama for its goals of reforming mandatory minimum sentences, reducing prison populations, and rehabilitating prisoners.
One aspect of the bill, however, has attracted much less attention: It would require the government to rate federal prisoners’ risk of committing a future crime, and treat them differently according to those ratings. The bill calls for the attorney general to develop a new formula for predicting future behavior or adopt an existing tool. The Bureau of Prisons would then use the algorithm to score and classify inmates.
Inmates who receive “low risk” scores — and those who manage to lower their scores over time — would be allowed into rehabilitation and education programs that would shave time off of their sentences. “High risk” inmates would not be eligible.
Such tests are increasingly popular around the country. They are used to make decisions about everything from bail to sentencing. The scores are meant as a counterweight to the vagaries and biases of human decisions.
“A good risk assessment acts as an anchor against our personal biases,” says David D’Amora, at the Council of State Governments Justice Center. “God forbid we go back to what we used to do — because with the best of intentions, we used to make decisions that had no evidence behind them.”
Yet the formulas are often not transparent. ProPublica recently investigated one popular tool sold by a for-profit company and found that it’s frequently wrong and is biased against blacks.
So what’s the evidence the assessments of prisoners will work in the ways the bill envisions? Not much, according to an analysis by Federal Public & Community Defenders, a group made up of attorneys who represent federal defendants who can’t afford counsel.
“The system described in the bill is novel and untested,” the group concluded in a little-noticed white paper that was presented to a federal task force last May.
The proposal to score federal inmates was first developed by Sens. Sheldon Whitehouse, D-Rhode Island, and John Cornyn, R-Texas, in a prison reform bill last year. The senators, who co-sponsored the bill, have said they modeled the plan after seeing successful versions of it in their home states.
But as the federal defenders’ paper points out, the Rhode Island and Texas programs the senators point to work in a different way than the system the bill proposes. Inmates in the two states can earn reduced time through their personal performance and disciplinary records, not risk assessment scores.
The report argues that the scoring of prisoners would be particularly problematic because factors that go into risk assessment calculations tend to correlate with socioeconomic class and race. A fairer and more effective approach, the group says, would be to make recidivism-reduction programs available to all inmates equally.
The Senate reform bill does not specify the use of any particular risk assessment tool, and those who support the bill say that it has built-in protections against errors and discrimination. The current bill would mandate regular testing and adjusting of the risk assessment tool, including analyzing the results for evidence of racial bias. It also would require that the risk assessment tool be statistically “validated,” that is, calibrated to work specifically on the population that will be assessed with the tool.
But the bill does not specify that the validation be done independently. (The tool that ProPublica analyzed had been validated by the company that developed it.) Nor does the bill specify how soon such testing would have to happen before the tool gets put to use, other than “as soon as is practicable.” The federal defenders argue that the tool should be fully tested and corrected before it gets used at all.
Cornyn has said prisoners should be separated into two groups: “hardened” criminals, who should be kept isolated from the rest of society; and “low-level” offenders who truly want to change and thus deserve help.
“I’m not naive enough to say that this is something we’re going to be able to do for 100 percent of the people behind bars, that’s just not true,” Cornyn said on the Senate floor in April.
Richard Davidson, a spokesman for Whitehouse, said it’s important to assess prisoners on “dynamic” rather than “static” factors. The idea is to focus on things that prisoners will be able to change while they are incarcerated, rather than things like past criminal history that can often reflect societal bias.
The federal defenders say even so-called dynamic factors are very hard to change in a prison setting.
Risk assessments often include questions about defendants’ family relationships or employment status, explained David Patton, who chairs the legislative committee of the federal defenders group. “Will it be possible to validly measure those things for somebody who is institutionalized?” he asked. “We just don’t know that such a tool can be developed, or if it can, whether it will exhibit similar racial biases of current tools.”