| The work presented in this dissertation is interdisciplinary research in human-computer interaction at the intersection of information retrieval (IR), cognitive psychology, and computer science. My research focuses on the use of relevance feedback (RF), a query reformulation tool. I present two experiments that investigate (1) how end-users with and without a general background in information retrieval use a state-of-the-art, best-match, ranked-output information retrieval mechanism interactively to develop effective information filters for document retrieval from a large scale, full-text document collection and (2) how the availability of relevance feedback and the level of user knowledge about and the control of the output of the relevance feedback mechanism impact user behavior and performance.; Ten searchers who had a background in IR were observed in a first study; 64 complete novices took part in a second experiment that systematically varied user knowledge and user control of the feedback mechanism. Behavioral and performance data suggest that (1) iterative query formulation is a depth-first search process, that (2) users with and without a background in information retrieval can, with little training, interact effectively with a best-matched, ranked-output IR mechanism, that (3) relevance feedback tends to improve performance, and that (4) increased levels of user control over relevance feedback shortens the search process while resulting in equal or better performance. |
