Are risk assessments used in adult corrections racially biased? Do they unfairly label people of color as higher risk for future crimes?
These are important and often uncomfortable questions, raised most recently in a laudable ProPublica analysis of the COMPAS risk assessment used in adult corrections sentencing decisions.
Aside from the question of what decisions risk assessment should, and shouldn’t, be used for, there is a question of whether risk assessment itself perpetuates inequity. The answer: Yes. And no.
Risk assessment, at its best, can reduce bias in decision making. At its worst, it can propagate bias.
This is exactly why using risk assessment for sentencing in adult corrections is troubling. The adult corrections system is often racially biased and overwhelmingly punitive in nature. Risk assessments used at sentencing are being used to drive punishment as the risk score moves higher.
In a racially biased system, or in a society that has inequities, risk assessment will create a disproportionate impact on a particular group, in this case, African Americans.
The key fact we need to keep in mind is that before risk assessment has any chance to influence any individual’s corrections involvement, a long list of other factors has already established inequity in the system.
- Federal housing policies like redlining made it difficult for some communities to sustain intergenerational economic prosperity;
- Financial redlining created pockets of underserved communities with few resources that have been linked to higher arrest rates;
- Policing practices have unfairly targeted black and African-American communities;
- Drug sentencing laws have had disproportionate impacts on black and African-American communities;
- Implicit biases affect how threatening African Americans are perceived to be by decision makers, as compared to whites.
- A history of institutionalized disenfranchisement of African-American communities has eroded the representativeness of democracy.
- The destructive myth of the African-American “super predator” has shaped criminal justice policy and practice.
Until and unless these types of structural and implicit inequities are resolved and untethered from the adult corrections system, use of fair and equitable risk assessment tools will not be sufficient to transform adult corrections in the United States.
People who care about how decisions are made in justice settings often ask whether it is an ethical problem if African-Americans are more likely to get a high-risk assessment score. The worry is that higher risk scores lead to negative outcomes.
This is a valid concern, and it holds true if the risk assessment is used to direct punishment, prison sentences, or punitive probation practices (see the ProPublica article for good examples). In this case, the issue is not with the risk assessment, but the punitive intervention the corrections system and the community are investing in.
That means risk assessment can be an essential part of non-punitive decisions in adult corrections, including pre-trial release, targeting alternatives to placement, parole review and release, etc. The positive uses of risk assessment are, in fact, a strong argument in support of justice reform—not that we should need any further convincing.
Advocates, funders, researchers, justice system workers and, most of all, every person who cares about a fair, effective justice system should want well-constructed and well-tested risk assessment as part of the ideal system that they are working to create.
We all deserve to be treated equitably, afforded opportunities and resources as needed, and empowered to contribute to our communities. Risk assessment can be part of that vision.
Jesse Russell, PhD, is the chief program officer of the National Council on Crime & Delinquency. NCCD works to improve outcomes for at-risk children, adults, families, and communities by bringing research and data-driven decision making to the juvenile and adult criminal justice, child welfare, and adult protection systems. Dr. Russell welcomes comments from readers.