Skip to content
Link copied to clipboard
Link copied to clipboard

Pennsylvania’s proposed risk-assessment algorithm is racist | Opinion

Risk assessment tools (RATS) are meant to predict people’s future behavior. RATs in the criminal justice system are almost always trained on criminal and police records, data from a society that overpolices, prosecutes, and punishes black and brown people.

Dorothy Johnson-Speight, pictured here in 2017, is a community organizer who lost a son to murder and founded Mothers in Charge. She testified last year about her opposition to the Commission’s RAT, and why those who have lost loved ones deserve individualized sentencing, not RATs.
Dorothy Johnson-Speight, pictured here in 2017, is a community organizer who lost a son to murder and founded Mothers in Charge. She testified last year about her opposition to the Commission’s RAT, and why those who have lost loved ones deserve individualized sentencing, not RATs.Read moreCAMERON B. POLLACK / Staff Photographer

Over the last two years, scores of people have testified to the State Sentencing Commission about a proposed “risk assessment tool” meant to be a part of sentencing decisions. We testified because the risk assessment tool (RAT) designed by the commission is racist and will contribute to Pennsylvania’s mass incarceration crisis.

Risk assessments are algorithms, formulas meant to predict people’s future behavior. RATs in the criminal justice system are almost always trained on criminal and police records, data from a society that overpolices, prosecutes, and punishes black and brown people and communities nationwide.

To its credit, the commission’s public process is rare; drafts of the tool were published for all to see. This has meant that those impacted by mass incarceration, data scientists, prosecutors, defense attorneys, national academics, survivors of harm, and other stakeholders have been able to speak against the racism baked into the tool, and about how the RAT will impact them.

But the tool on the table is simply inaccurate. It fails to predict future violence. When the commission contracted with Carnegie Mellon University to evaluate it, CMU data scientists recommended that the commission discard its attempts to predict who would commit a crime in the future and just focus on nonviolence; the RAT is far more accurate at predicting people who could return to the community safely. The commission did not accept that recommendation, giving no meaningful justification.

The proposed RAT uses prior convictions to predict future ones. That’s troubling in a state where overpolicing, bail, and overprosecution mean black men are nine times more likely to be incarcerated than white men. The tool uses gender in a way that is likely unconstitutional, opening up Pennsylvania to lawsuits.

Worst of all, the RAT proposal removes the recommendation for a post-sentencing investigation. Instead of encouraging a judge to dig into a person’s humanity through interviews and investigation, the proposal instead directs judges, after applying the first RAT, to then apply a second RAT — whatever algorithm the county probation and parole office uses to determine community supervision level.

The commission is pushing these probation tools as “risk/needs/responsivity” tools, or RNRs, meant to support a person’s needs when convicted. But these tools are nothing of the sort. Some ask invasive questions about “demeanor,” or factors correlated with homelessness, addiction, and race. These second RATs then generate yet another inaccurate risk label; one that leads to oversupervising people for years, putting high barriers in front of their success.

Philadelphia’s probation RAT is one of the least transparent and most problematic in Pennsylvania. As the ACLU noted in August testimony, the tool has historically used zip code as a proxy for dangerousness in one of America’s most segregated cities. High-risk labels assigned by the Philly tool are a part of why our jails are full of people accused of violating supervision, a big driver of Philly incarceration at a time of marked reform.

Dorothy Johnson-Speight, a community organizer who lost a son to murder and founded Mothers in Charge, testified last year about her opposition to the commission’s RAT, and why those who have lost loved ones deserve individualized sentencing, not RATs: “I believe had the individual who killed my child had access to stronger supports, both he and my son would be here making a positive impact on our community.... I am standing against the utilization of a tool that will likely expand actions of mass incarceration in the state."

Pennsylvania has the chance to lead by rejecting this risk assessment tool, and by moving to repeal the bill that called for one, pushing instead for individualized justice for each person impacted by this system. Even former State Sen. Stewart Greenleaf, who authored the bill to put risk assessment in Pennsylvania sentencing, has rejected this proposal.

Civil rights organizations, scientists, communities, and the nation’s most brilliant thinkers on criminal justice have rejected algorithms as a path to fairness and justice. Voters care about algorithmic bias. We are paying attention.

The sentencing commission can join us now by rejecting this tool and working to repeal the mandate today.

Hannah Sassaman is policy director at the Media Mobilizing Project, which is a member of the #No215Jail Coalition.