TOWARDS A BETTER CONCEPTUAL-FRAMEWORK FOR UNDERSTANDING RELEVANCE FOR INFORMATION-SCIENCE RESEARCH

Our current understanding of the concept of relevance in information science is limited and incomplete. The traditional approach to defining relevance is to try to find one characteristic or set of characteristics that serves to delineate the relevance criteria in a given judgement, e.g. usefulness, novelty. This has been called a "monothetic approach," and is characteristic Tefko Saracevic's overview article of relevance [1]. Although worthwhile, this approach does not fully account for how relevance judgements are made, and why there are variations in what is relevant. Relevance judgements vary in terms of amount of precision applied, in the criteria used and the level with which they are applied. On the one hand, information researchers assume that precise criteria exist in the minds of most information seekers, where in actuality, levels of imprecision and precision exist. On the other hand, the criteria for relevance are heterogeneous and different criteria, or sets of criteria, can be applied simultaneously. This paper shows that relevance judgements can be based on vague criteria, are polythetic in character, and are based in part on the environment, task, and background of the judgement maker. Most existing research on relevance judgements has not fully taken into account the necessary, diverse considerations in assessing relevance judgements nor how the concept of relevance evolves in human cognition. Using Eleanor Rosch's notions of natural categorization, basic-level categories and prototypes, George Lakoff's linguistic theory and some phenomenological notions of Maurice Merleau-Ponty, this paper will show how relevance judgements are based in a basic-level category of relevance acquired through experience, how subcategories of relevance evolve in relation to information systems, how these can be invoked with levels of precision or imprecision, and how they can be based on different sets of criteria that can be applied simultaneously or within the same context, these criteria being close to the prototype or having "family resemblances" with core criteria or sets of criteria. Through such a conceptual framework, a firmer foundation for continued research in relevance will be established.