Racial Bias and Amelioration Strategies for Risk Assessment

Go Back

3:30 PM - 5:00 PM

---

For over 40 years justice agencies have relied on risk assessments to guide decisions regarding diversion, case management, supervision, and placement practices. Despite widespread use, it was not until a 2016 ProPublica investigation “Machine Bias” that risk assessment developers, researchers, and justice agencies were directed to examine the potential for tools to extend systematic biases. By design, prediction models remove idiosyncrasies that can result from the inconsistent use of discretion, scoring individuals with equity, and based on established predictors of risk. However, the items used to create prediction models are difficult to disentangle from cultural and regional biases. With that said, common predictors used, such as prior criminal convictions, are the result of law enforcement patterns and strategies that are argued to possess their own implicit biases. In 2021 the Pennsylvania Department of Corrections (PDOC), Vant4ge, and researchers from the University of Nebraska –Omaha, began the process of developing a customized version of a prominent risk assessment –the Static Risk and Offender Needs Guide-Revised (STRONG-R) –for the Pennsylvania incarcerated population and those under parole supervision. The STRONG-R is unique in that the prediction models, its items, and response weights are adjusted to account for the specifications of the population in which it is to be implemented. Unlike traditional off-the-shelf assessments, the STRONG-R takes advantage of prior applications, designing the Pennsylvania version of the tool from a development sample.