Concluding Comments About the Special Issue on Hierarchical Inference.

The article presents information on the articles published in the December 1973 issue of the journal "Organizational Behavior and Human Performance." The articles reflect growing awareness that many important probabilistic inference tasks are hierarchical in the sense that they involve complex conditioning patterns among events. Each event forms a link in an often lengthy inferential chain leading from observables to primary hypotheses whose relative likeliness is to be determined. Three articles in this journal are concerned with various formal issues in hierarchical or cascaded inference. A general model by Scott Barclay and Clinton W. Kelly for hierarchical inference shows rather clearly just how complex the essential conditioning patterns among inferential events can be. The Schum-Kelly article examines a special case of hierarchical inference in which the reliability or credibility of sources of inferential information is an issue. This paper illustrates the absolute necessity for careful prescriptions about the independence and conditional independence of events.