ABSTRACT
Introduction In the criminal justice system, prediction is widely used to classify individuals into different risk levels, and assign some form of treatment in the system based on this classification. For example, the magistrate must decide whether to grant bail based on the risk of new offenses, the correctional authority must decide on the level of security to place the inmate, and the probation officer must decide whether to place the individual in treatment for substance abuse or vocational training (Gottfredson, 1987). Prior to the 1970’s, these decisions were usually made subjectively, using the judgment and experience of administrators. Starting in the 1970’s, there was an accelerating movement towards more objective, actuarial risk assessment tools, a movement that has led to a large literature on risk assessment and at least four distinct gene rations of risk tools (Gottfredson & Moriarty, 2006). However, there has been comparatively little discussion on the determinants of potential classification systems for risk-based decisions outside of the criminal justice context (i.e. employment decisions) for individuals who have a criminal record.
