GateHouse Media
August 2, 2019
Over the past two decades as crime rates fell and
incarceration rates rose some have argued, according to University of Virginia
Professor Brandon Garrett, “that one way to begin dialing down ‘mass
incarceration’ without simultaneously jeopardizing the historically low crime
rate is to put risk assessment back into sentencing.”
According to Garrett, risk assessment in sentencing is not a
new phenomenon. In California, more than a century ago, lawmakers introduced
indeterminate sentencing - whereby an offender is given a minimum sentence and
maximum sentence and is only released from prison after the minimum, and prior
to the maximum, upon demonstrating that he or she is a low risk for
reoffending.
It has recently been estimated that at least 20 states have
begun to incorporate risk assessment into the sentencing process “in some or
all cases.”
However, this week, 27 leading criminal justice researchers
signed a letter denouncing the use of sentencing risk algorithms in courtrooms
across the country. They report the tools’ reliance on distorted data does
little to reduce jail populations and unfairly penalizes minority defendants,
reported the Pennsylvania Capital-Star.
The letter suggested that current risk assessment
instruments are unable to distinguish one person’s risk of violence from
another. In statistics, predictions are made within a range of likelihood,
rather than as a single point estimate.
Some predictive algorithms appear to make estimates of
recidivism accuracy at between 5 and 15%, where studies have demonstrated that
predictive models can only make reliable predictions about risk for violence
within ranges of 20 to 60%.
The letter continues, “As a result, virtually everyone’s
range of likelihood overlaps. When everyone is similar, it becomes impossible
to differentiate people with low and high risks of violence. At present, there
is no statistical remedy to this challenge.”
As far back as 2014, then U.S. Attorney General Eric Holder
warned that risk assessments might be injecting bias into the courts. At the
time, he called for the U.S. Sentencing Commission to study their use.
“Although these measures were crafted with the best of intentions, I am
concerned that they inadvertently undermine our efforts to ensure
individualized and equal justice,” he said, adding, “they may exacerbate unwarranted
and unjust disparities that are already far too common in our criminal justice
system and in our society.”
According to a recent report by the Partnership on
Artificial Intelligence, “using risk assessment tools to make fair decisions
about human liberty would require solving deep ethical, technical, and
statistical challenges.” The report found that tools currently available and
under consideration by jurisdictions across the country fail to adequately
address those concerns.
In Pennsylvania, the Board of Probation and Parole, the
agency charged with making parole decisions, uses various assessments including
risk, sex offending, mental health and drug and alcohol. Risk assessments tools
coupled with parole guidelines are thought to provide uniformity to the board’s
decision making process.
Just this past April, Pennsylvania Governor Tom Wolf said,
“Pennsylvania can lead the nation with bold bipartisan reforms to probation and
parole.”
Within three months, according to WHYY-FM, parolees from
Pennsylvania prisons committed six murders.
The first, on May 23, a parolee allegedly strangled his
girlfriend’s mother, then set her home in Hershey on fire to cover it up.
On June 29, police say another parolee murdered his
girlfriend’s two-year-old daughter in Baltimore, Maryland. About a week later,
a parolee from New Castle, just north of Pittsburgh, allegedly murdered his
girlfriend’s eight-year-old son.
On July 14, 2019 an off-duty Pittsburgh police officer was
shot and killed, allegedly by a parolee. Just four days later, another parolee
allegedly stabbed his sister and niece in Lancaster.
Pennsylvania’s Secretary of Corrections John Wetzel told
WHYY-FM, “Listen, I think in every way we need to use data, measure it, infer
what we can from the actual data, learn what we can from individual cases, and
not knee-jerk.”
Predicting human behavior is a tricky business. While it may
have a place in the criminal justice system, blind adherence to predictive
tools can have deadly consequences.
Matthew T. Mangino is of counsel with Luxenberg, Garbett,
Kelly & George P.C. His book The Executioner’s Toll, 2010 was released by
McFarland Publishing. You can reach him at www.mattmangino.com and follow him
on Twitter @MatthewTMangino.
1 comment:
Brandon Garrett is a law professor at Duke University, not University of Virginia.
Post a Comment