A Computer-Made Device for Sentencing Decisions

Publisher Website
Google Scholar
Abstract
This paper reports a new device for predicting success on pro bation, developed in a study of 933 male probationers in Ontario. Discriminant function and related methods, association analysis, simple summation, and predictive attribute analysis were used, and their virtues and weaknesses compared. Predictive attribute analysis proved to be most efficient but least stable, and simple summation least efficient but most stable. On cross-validation the differences between the various techniques tended to disappear. The new device consists of a table and a form on which the probation officer notes eight predictive attributes of an offender as present (+) or absent (-). This form is moved along the table until all the pluses and minuses agree. The risk category and the probability of success on probation can then be read off the table. The device is at present applicable only in Ontario and its life expectancy is short. A more lasting device with wider application could be constructed if cooperation from several jurisdictions is obtained and the standardization sample is representative of all offenders before the court for whom probation is available.

This publication has 0 references indexed in Scilit: