A Brunswik lens investigation of the profile interpretation, pure statistical, and clinical synthesis methods of predicting absenteeism
Willoughby, Frederick William
MetadataShow full item record
The ability of humans to make judgments and predictions about their environment has been studied in the laboratory through a variety of tasks (e.g., making diagnoses, predicting improvement in therapy, job performance, dangerousness to self and others, and recidivism). The accuracy of human (clinical) prediction in comparison to actuarially-derived (statistical) prediction has been heavily debated since Paul Meehl's controversial publication Clinical versus Statistical Prediction in 1954. Support for the superiority of statistical prediction methods over clinical prediction methods has been overwhelming. Researchers have suggested that further research should focus on developing procedures that would enable the human judge to make more accurate predictions. The present study investigated human judgment in the task of managers predicting absenteeism of employees. Three hypotheses were proposed. The first hypothesis predicted that strict application of Bayes' formula (pure statistical method of prediction) would significantly exceed the predictive accuracy of managers who obtained training in Bayes' formula (clinical synthesis method) as well as managers who did not receive this training (profile interpretation method). The managers in the clinical synthesis condition, however, were expected to be significantly more accurate than managers in the profile interpretation condition. The second hypothesis predicted that managers in the clinical synthesis condition would have significantly higher achievement, consistency, and matching indices as proposed by the Brunswik lens model, than managers in the profile interpretation condition. Finally, in the third hypothesis, it was predicted that managers in the clinical synthesis condition would be more appropriately confident in the accuracy of their predictions (i.e., better calibrated) than managers in the profile interpretation condition. The results found that there were no differences between the pure statistical and clinical synthesis methods of prediction; these two methods, however, were significantly more accurate than the profile interpretation method. Secondly, the clinical synthesis condition resulted in a significantly higher achievement index than the profile interpretation condition. Finally, managers in the clinical synthesis condition were better calibrated than managers in the profile interpretation condition. Discussion of these results included possible explanations of the findings and suggestions for further research.