THE LIKELIHOOD OF VIOLENT BEHAVIOR - PREDICTIONS, POSTDICTIONS, AND HINDSIGHT BIAS

The effects of variations in base rate information and temporal context on predictions of violence and hindsight bias were investigated. Two hundred and seventy undergraduate students estimated the likelihood of violent offenders committing another violent offense based upon fictitious case history information. The subjects were given 1 of 2 population base rates of the likelihood of re-offending (30% or 70%), and a recidivism estimation question that was phrased either predictively, predictively with a given outcome of either recidivism or no recidivism, or postdictively. Although different offenders were perceived as differentially likely to reoffend, estimated likelihood of recidivism was affected by neither base rate nor temporal phrasing of the estimation task. A second study of 182 students, using shorter case histories, again found no base rate effect. Hindsight bias effects were weak and inconsistent in both studies. These results suggest that insensitivity to base rate information is a very robust phenomenon whereas the hindsight bias effect is restricted to certain types of tasks and postdictively phrased questions.