On Aggregation, Generalization, and Utility In Educational Evaluation

ion and operation; those who concern themselves with details too far down or up the line will lose their effectiveness. The Stanford Evaluation Consortium (1976) suggested that producing this type of information should be a prime concern: Evaluation [in their suggested model] becomes a component of the evolving program itself, rather than disinterested monitoring ... Formal reports to outsiders are reduced in significance, and research findings become not conclusions but updatings of the system's picture of itself. (p. 18) To summarize, information is more likely to have meaning to a decison-maker if it is from an organizational level close to that at which he or she operates. Thus, the cross-levels hypothesis is supported by the semantic criterion, and we can conclude that errors in interpretation are likely to be introduced by aggregation and generalization across level. Cross-levels Evaluation and the Behavior