Properties-based retrieval and user decision states: User control and behavior modeling

As retrieval set size in information retrieval (IR) becomes larger, users may need greater interactive opportunities to determine for themselves potential relevance of the resources offered by a given collection. A parts-of-document approach, coupled with an interactive graphic interface and control panel, permits end users to tailor the information seeking (IS) session. Applying the model described by the author in a previous paper in this journal, this paper explores two issues: whether a group of information seekers in the same research domain will want to use this type of IR interaction, and whether such interaction is more successful than relevancy ranked lists, based on the general vector model. In addition, the paper proposes the use of gradient space as a means of capturing end users' cognitive states -- decision-making points -- during a parts-of-document-based IR session. It concludes that, for a group of biomedical researchers, a parts-of-document approach is preferred for certain IR situations and that gradient space provides designers of systems with empirical evidence suited for systems analysis.