NCJ Number
81335
Journal
Law and Human Behavior Volume: 5 Issue: 4 Dated: (1981) Pages: 263-274
Date Published
1981
Length
12 pages
Annotation
This study develops statistical analogs of the opposing 'public protectionist' versus 'civil libertarian' models of social control for predicting violent behavior among samples of defendants found incompetent to stand trial in New York State.
Abstract
The study sample was composed of two distinct subgroups: persons indicted for a felony prior to their incompetency determinations who already had specific psychiatric evaluations for dangerousness and persons unindicted for their felony charges and requiring psychiatric determination of competency to stand trial. The basic prediction equation was derived retrospectively from the indicted sample. This decisionmaking model was applied to the unindicted sample. In these predictions, the probability rates were set to approximate the two different social control models. Thus, although the analyses are based on retrospective data, they simulate a prospective design, which may represent the closest approximation to an 'in vivo' study of actual clinical decisions. Using the more conservative 'public protectionist' approach (equal probability) yields a high percentage of false positives (i.e., persons identified as being at high risk but failing to exhibit the predicted assaultive behaviors). On the other hand, this model produces relatively few false negatives, (i. e., persons predicted not to be dangerous who are). When the prediction model is altered to reflect the actual level of assaultiveness of the indicted group (Bayesian probability), the overall accuracy increases -- only 102 persons were predicted to be assaultive as compared to the 155 so predicted in the other model. The policy implications of this decrease are that one-third fewer persons would have been placed in higher security, higher staffed, higher cost units. However, such a strategy would also have 'cost' 21 persons classified as not dangerous, who were. The data show the direction and volume of the errors produced from contrasting social control models reflecting ideological positions of clinical and legal practitioners. The advantages of this statistical approach are a more precise estimate of the costs, personnel and societal, of each decision and the possibility of a more scientific, cumulating product from the regular demands for clinical predictions of future behavior. Tables and 35 references are given.