Effect of adopting Virginia’s risk-assessment tool on racial sentencing disparities

Determine whether adopting the risk-assessment tool used in Virginia felony sentencing increased the disparity between sentences imposed on Black and White defendants with comparable recorded risk and guideline characteristics.

Background

The paper discusses evidence that, conditional on a given risk score, judges sentenced Black defendants to terms 12–24% longer than White defendants. It distinguishes this disparity associated with judges’ discretionary decisions from disparities associated with the risk score itself.

The cited study estimated that the change in disparity following adoption of the tool was small and statistically insignificant, but the paper emphasizes that the study did not establish whether adoption increased the disparity. Thus, the causal effect of adopting the tool on the disparity remains unresolved.

References

Whether adopting the tool worsened this gap is separate and more cautious: the estimated change is small and not significant; the study does not establish whether adopting the tool increased this disparity.

— Cybernetic and Epistemic: A Missing Vocabulary for Trustworthy Agentic Delegation  (2610.00961 - Lumbroso, 1 Oct 2026) in Section 3.5, “A negative case: cybernetic oversight without epistemic reconstruction” (Subsection \ref{sec:negative})