A threshold model of relevance decisions

摘要

This article presents a model of relevance decisions as being one of many classes of decisions made by the receiver of a message (document). The relevance decision model has the underlying assumption that a relevance threshold is a fundamental characteristic of every individual. The purpose of the model is to predict active relevance-related behavior which is assumed to follow from the internally made relevance assessment. Two predictor variables, total efficiency, and total value, are operationally defined. The criterion variable, relevance, is assumed to be dichotomous but expressable on a continuous interval scale. The use of discriminant analysis is offered as a means for deriving a relevance threshold. Predictions about a document's relevance can be made by determining whether a new document's criterion score is above or below the relevance threshold. Emphasis is then placed on predicting relevance-related behavior which would follow the internal decision rather than merely verbally classifying new documents. Strengths and limitations of the model are examined.