An initial examination of ease of use for 2D and 3D information visualizations of web content
International Journal of Human-Computer Studies
We present a discussion and initial empirical investigation of user interface designs for a set of three Web browsers. The target end user population we identified were experienced software engineers who maintained large web sites or portals. The user study demonstrated the strengths and weaknesses of two conventional 2D browsers for this target user, as well as that of XML3D, a novel browser that integrates an interactive 3D hyperbolic graph view with a more traditional 2D list view of the
... . A standard collapse/expand tree browser and a Web-based hierarchical categorization similar to Yahoo!, were competitively evaluated against XML3D. No reliable difference between the two 2D browsers was observed. However, the results showed clear differences between XML3D and the 2D user interfaces combined. With XML3D, participants performed search tasks within existing categories reliably faster with no decline in the quality of their responses. It was informally observed that integrating the ability to view the overall structure of the information space with the ability to easily assess local and global relationships was key to successful search performance. XML3D was the only tool of the three that efficiently showed the overall structure within one visualization. The XML3D browser accomplished this by combining a 3D graph layout view as well as an accompanying 2D list view. Users did opt to use the 2D user interface components of XML3D during new category search tasks, and the XML3D performance advantage was no longer obtained in those conditions. In addition, there were no reliable differences in overall user satisfaction across the three user interface designs. Since we observed subjects using the XML3D features differently depending on the kind of search task, future studies should explore optimal ways of integrating the use of novel focus + context visualizations and 2D lists for effective information retrieval. The contribution of this paper is that it includes empirical data to demonstrate where novel focus + context views might benefit experienced users over and above more conventional user interface techniques, in addition to where design improvements are warranted. 3 Risden et al. IJHCS 5/18/00 Introduction Web browsers and search engines provide users with a narrow keyhole onto the vast content of the World Wide Web. Most interfaces for accessing Web content employ minimally interactive two-dimensional techniques. There is promise in developing more effective browsers that incorporate richer three-dimensional interaction and visualization capabilities. Such browsers are becoming feasible for consumers because of the trend towards inexpensive high-performance graphics capabilities. In this paper, we compare a novel interactive browser with a 3D hyperbolic graph view to two legacy 2D browsers, and provide initial experimental evidence that the interactive 3D visualization of large amounts of information is indeed useful for certain tasks. However, our evidence also shows that users still prefer to use more familiar, 2D tools under certain task demands. We shall focus on browser designs that were created to help organize, make sense of, and manage large collections of Web pages, files, or documents available on the Web. We are particularly interested in collections that are not strictly hierarchical in structure, in which categories may have one or multiple parents. We explore the domain of Web site content generation and maintenance, tasks that our target user population deals with on a daily basis, often with nonhierarchical data sets. In the empirical study reported in this paper, we analyze the usability of three different browsers in detail for this genre of tasks and data. In this paper, we shall first provide key background and motivation for the design and development of XML3D, focusing on how this novel browser integrates an interactive view of the link structure of a site in 3D hyperbolic space with 2D list components. In addition, we will briefly cover the literature on graph-based visualizations. However, more detailed discussions of an earlier version of XML3D have been previously published (Munzner, 1997 (Munzner, , 1998a (Munzner, , 1998b , and a review of graph-based information visualization is not the goal of the current paper. Since our goal is to communicate the initial ease of use findings from comparing XML3D to alternative browsers, we describe in detail the user interface of the XML3D as it applies to the tasks of Web search, content generation and maintenance. Then, we demonstrate the relative strengths and weaknesses of XML3D when compared with existing Web visualizations. We present a user study that compares the XML3D Web visualization to two alternative 2D visualizations of Web content: the traditional collapsible outline tree browser (e.g., the Windows 4 Risden et al. IJHCS 5/18/00 Explorer-like user interface) and the Web site http://www.snap.com. The Snap Web site is an edited, hierarchical structure navigated from an overview page to more detailed but less comprehensive pages, much like Yahoo! and many other directory search sites. The contributions of this paper include a discussion of how to effectively integrate a 3D hyperbolic graph view with other 2D components based on empirical observation, and a detailed evaluation of the effectiveness of 3D hyperbolic layouts accompanied by 2D lists, particularly when compared to existing Web solutions such as collapsible tree browsers or standard Web portal site hierarchical layouts.