Paper short abstract:

In July 2020, the Sentence Risk Assessment Instrument was implemented in Pennsylvania courts to evaluate "the relative risk that an offender will reoffend and be a threat to society." Through interviews, I probe how judges interpret and use the tool's recommendations in their sentencing decisions.

Paper long abstract:

In July 2020, the Sentence Risk Assessment Instrument was implemented in Pennsylvania criminal courts to evaluate "the relative risk that an offender will reoffend and be a threat to society." Recidivism risk assessment instruments, which estimate an individual’s risk of rearrest for a future crime, are often presented as a data-driven strategy for progressive judicial reform – a way of reducing racial bias in sentencing, abolishing cash bail, and reducing mass incarceration. In much of the United States, these risk scores inform judges’ decisions including bail, pretrial release, and sentencing. However, little is known about whether and how risk assessment promotes these progressive goals in practice. Growing empirical evidence suggests that risk assessment can increase racial disparities in judicial discretion because judges may selectively disregard risk scores along racial lines and are more likely to agree with recommendations to detain defendants. Given the stakes, it is essential to understand not only risk assessment tools' construction and fairness, but also their effects on algorithmic fairness in practice – that is, how the instrument's recommendations interact with judicial discretion upon deployment. In interviews with 20 judges who use the Sentence Risk Assessment Instrument's recommendations in their felony and misdemeanor sentencing decisions, I probe how judges interpret and conform to the tool's recommendations, as well as how they weigh the role of recidivism risk in their decision-making overall. I also discuss the effects the tool has had on judges' day-to-day work and their attitudes about the controversy the tool has generated.