After nearly a decade, Pa. sentencing commission adopts risk assessment tool over objections of critics

STATE COLLEGE, Pa. — The Pennsylvania Commission on Sentencing voted Thursday to adopt a controversial risk assessment tool designed to help county judges determine sentences and divert low-level offenders into alternative sentencing programs.

The 7-2 vote, held during the commission’s quarterly policy meeting in State College, caps off nearly a decade of work on the algorithm, which uses empirical data including age, gender, and conviction history to predict a person’s risk of committing new crimes. 

The tool is aimed at curbing prison populations by making sentences more consistent and identifying candidates for diversionary programs. 

But groups including the American Civil Liberties Union have opposed it from the get-go, saying it would violate constitutional liberties by reinforcing race and gender biases. In July, 27 leading criminal justice researchers said similar tools’ reliance on distorted data does little to reduce jail populations and unfairly penalizes minority defendants.

Sentencing algorithms like the one under development in Pa. are ‘deeply flawed,’ researchers say

Members of the public reiterated those concerns on Thursday during a brief public comment period preceding the vote.

Mark Houldin, an attorney and former policy director for the Defender Association of Philadelphia, asked the commission to heed the warnings of a Carnegie Mellon University research team, which said the tool identified high-risk offenders with dubious accuracy.

The commission tweaked the tool in response to the Carnegie Mellon report, Deputy Director Matthew Kleiman said Thursday. But it rejected the team’s recommendation that it restrict use of the tool to “low-risk” offenders.

Pennsylvania’s General Assembly ordered the sentencing commission to develop the tool back in 2010. The panel has since published five drafts and heard feedback in 19 public hearings.

Sheila Woods-Skipper, the Philadelphia Court of Common Pleas judge who chairs the commission, said Wednesday that the road to the vote has been “a long and tedious process.” 

“We have attempted to [be] transparent about the process and responsive to changing the tool based on criticism from the public,” Woods-Skipper told the Capital-Star Wednesday. “We would recognize it’s not a perfect tool, but is it a tool that accomplishes our goals and does what it’s intended to do and provide additional info in sentencing?”

The most recent round of hearings took place in Philadelphia, Pittsburgh, and Harrisburg in August, when opponents rejected a proposal to use county-based probation screenings to determine recidivism risk in some cases.

The latest version of the tool, circulated at the commission meeting on Thursday, no longer calls for county probation offices to perform so-called risk and need assessments, which currently help them determine levels of supervision for people on parole. Instead, it advises county courts to obtain “additional information” about certain offenders prior to their sentencing.

That advisory will apply to offenders deemed both low and high risk, in order to avoid stigmatizing people in the latter category, commission vice-chair Rep. Todd Stephens, R-Montgomery, said Thursday.

But since it’s just an advisory, courts don’t have to follow it. 

“Our recommendation to the court is just that — a recommendation,” Mark Bergstrom, the commission’s executive director, said Thursday. “The court can reject that if they wish to.”

Commissioners say the findings of those investigations should help courts divert defendants out of prisons and into alternative sentencing programs.

Critics like Houldin, on the other hand, worry that any hint of a “high risk” assessment score will dog defendants during sentencing.

“Judges are going to figure out what the high-risk group is and what the low-risk group is,” Houldin said. “You can’t unring that bell.”

Some commission members expressed concern Thursday that counties would use unreliable assessment tools while gathering more information on offenders.

Helene Placey, executive director of the County Chief Adult Probation and Parole Officers Association, also said it would put a “significant” burden on county probation and parole officers who would have to perform the investigations.

Placey urged the commission to defer their vote and ask the Legislature to repeal the mandate.

Sentencing commission leaders expect the tool will be challenged in court. Unless the General Assembly passes a resolution rejecting it, the instrument is set to take effect on July 1, 2020, when it will be incorporated into the sentencing software that county courts use to generate sentence guideline forms. 

Starting in January, the commission must conduct a six-month training and orientation for judges and court staff on how to use it. 

The proposal approved Thursday also requires the commission to review the instrument every year to study its impact on sentencing patterns and the investigative methods employed by courts. They must also revalidate it every three years to ensure its accuracy.

Thursday’s meeting in State College was the commission’s final one before it is repopulated in October. At that time, it’s expected to lose at least three members who were appointed by Gov. Tom Wolf. 

As part of an omnibus bill passed in June, lawmakers in the Republican-controlled Legislature stripped Wolf, a Democrat, of his appointing powers to the commission and redistributed them to legislators.

Stephens, who was reappointed to serve another term on the commission starting in October, said he felt a “huge weight off [his] shoulders” with Thursday’s vote.

The former prosecutor said commissioners have narrowed the scope of the assessment tool significantly since 2015, when they released its first draft for public comment.

Earlier versions of the tool took variables such as an offender’s home county into account when calculating their risk assessment scores, Stephens said. The commission scrapped that version after advocates warned it would penalize minority offenders living in counties with crime rates.

The commission also narrowed the applications for the algorithm, Stephens said. He hopes that its current purpose — to help judges learn more about the people  — will prove beneficial once the tool takes effect.

“All we’re doing is recommending a judge get more information about a defendant, which is universally accepted as a good practice,” Stephens said.

Sentencing officials are trying — again — to build an algorithm to predict future crimes. Critics still have big concerns