NCJ Number
221033
Journal
Criminal Justice and Behavior Volume: 34 Issue: 12 Dated: December 2007 Pages: 1638-1658
Date Published
December 2007
Length
21 pages
Annotation
Using the Violence Risk Appraisal Guide (VRAG), relative operating characteristic (ROC) statistics are exemplified, with a discussion on criticisms of actuarials and ROCs as measures of accuracy.
Abstract
It is concluded that some criticism of actuarial instruments for the prediction of recidivism among seriously violent offenders relates to implicit or explicit values-based concerns about the morality of various forms of preventive detention and incarceration. It is proposed that true scientific caution reflecting all the available evidence and the relevant values entails enshrining actuarial methods in public policy and legislation. Actuarial tools for the risk of recidivism have existed for many decades. Actuarial assessments are distinguished from nonactuarial methods in that they are developed via specific tests of the relationships between constituent items and outcomes, and they combine variables to maximize predictive accuracy and efficiency. This article uses the Violence Risk Appraisal Guide (VRAG) as an example of several issues important to the value of actuarial assessments in forensic practice. One issue pertains to base rates (the proportion of offenders who violently recidivate) and how they should be evaluated. Other matters concern implicit and explicit value judgments. All these issues arise in sex offender civil commitment but also clearly pertain to violence-risk-related decisions in all forensic populations. The purpose is to provide nontechnical support for users of actuarial risk assessments. Figures, tables, references