A Nine Month Report on Progress Towards a Framework for Evaluating Advanced Search Interfaces considering Information Retrieval and Human Computer Interaction

This is a nine month progress report detailing my research into supporting users in their search for information, where the questions, results or even their environment requires functionality beyond the scope of keyword search alone. While keyword search environments, such as Google, have become the familiar standard for search on the web, research has shown that users, who may not know which keywords to use, adopt coping strategies because they have no other methods to express their query (Pirolli and Card, 1995). Typically, these coping strategies involve submitting tentative queries with general keywords to learn more about the meta-data, to inform a more specific keyword search later. More recently, the term Exploratory Search has been used to describe search environments that provide alternative search functionality for when users may have poorly defined goals or complex questions, have insufficient pre-search knowledge, or may be using a system with poorly defined or unpredictable indexing (White et al., 2006a). While early exploratory search interfaces have been developing for some time (Hearst, 2000), recent reports have discussed the need to find metrics for evaluating their success (White et al., 2006b). As these more interactive models of search provide increasingly versatile combinations of functions, Marchionini suggests that such success is not achieved by simply adding more features but by combining them to produce synergetic designs(Marchionini, 2006). To evaluate exploratory search interfaces, designs need to be measured in terms of their support for known search tactics. The research below investigates the history of Information Retrieval (IR) research into the human element of interactive search, to develop a framework that produces a measure for such support. Then, the weaknesses of rich search designs can be identified and mitigated, before they are put forward for complex and expensive user studies. This report describes the progress of this investigation into using InformationSeeking models of users, their needs and their search behaviour, to design a