Exploring the Information Landscape

Elaine G. Toms
Graduate School of Library and Information Science
University of Western Ontario
London, Ontario
(519) 661-2111 ext. 8516 etoms@julian.uwo.ca

ABSTRACT

Exploring or browsing is a process of searching in which the user recognizes the object of the search when they see it a human-driven and unstructured process. To examine this process, three navigational aids are experimentally manipulated: method of access, method of suggesting items to explore and method of navigation. A fourth aid, cues that influence exploration, are also assessed. The object of the study is to characterize browsing, to understand what facilitates browsing in an electronic environment, and to suggest an abstract representation of browsing.

KEYWORDS

Exploring, Browsing, Full-text, Newspapers, Navigation, Menus, Fish-eye views, Similarity measures, Informativeness, Information searching, Experimental study

INTRODUCTION

In the past decade, searching for information in electronic systems has changed from an intermediary, third-party operation to one that is user-driven. This shift from a traditional to a contemporary system is reflected as well in a change in the type of data from bibliographic citations to full-text. Of interest to this study is the evolution in method. Users now identify the object of a search by recognition by navigating from screen to screen until they find what they need in a process of browsing or exploring [1,2,3]. Traditionally, users (via an intermediary) specified the object of the search in advance by entering a set of keywords in a process of querying. In fact, searching by recognition and searchingby specification are integrated in current systems; users employ one method or the other while seeking information. This study will concentrate on the less-researched, less understood and much-maligned browse method that emphasizes exploration.

The process of browsing is as much human-driven as it is machine-driven. Users are guided by the system. Because they receive instantaneous results, they cognitively encode the new information and use it as a lever in further searching, i.e., information begets information [4]. These systems are primarily, although not exclusively, textual, and thus this search method is influenced by centuries of reader-text interaction [5]. Guthrie's 5-component model of information search best illustrates this process from both a procedural and cognitive viewpoint [6].

Designing effective systems to accommodate browsing is difficult. Browsing success depends on the ability of a user to navigate the system (unlike querying which depends on sophisticated search algorithms). Hypertext has been widely used. Despite the extensive treatment of navigational problems [7], like orientation and digression, hypertext is only a partial solution. Numerous features, such as maps of hypertext links, electronic bookmarks and footprints,' have been added to overcome problems. As a result, systems often suffer from an excess of 'featuritis'. The most significant problem, however, is our lack of understanding of what users do when they explore, what influences how they browse, and what might facilitate that process.

In this study, I argue that information representation and type of navigational aid influences users' success in browsing. Because users rely on recognition to determine the search outcome, they need suggestions of what to explore. These suggestions may be similar or dissimilar to the information currently being examined. Simultaneously, users need to retain a sense of their spatial location within the system. To accommodate both requires a dynamic unconstraining device that facilitates exploration and a stable anchoring device that functions to orientate. An analogy to real world travel is the stabilizing influence of mileage signs with the diversion provided by signs signifying roadside attractions. The stabilizing signs serve as landmarks, provide route direction and hopefully over time lead to the acquisition of survey information [8]. Both can be formed by a representation of the information. A user's exploration at the local level, i.e., within a node of information, may be implemented by sequential paging, but the exploration is directed by cues from the content landmarks.

An electronic newspaper will be the vehicle used to examine browsing in this study. Research on electronic newspapers has focused on the personalized newspaper -- the daily me -- based on user profiles [9]. Thus, of secondary interest will be an assessment of browsing behaviour exhibited by readers of electronic newspapers. Characteristically, newspapers contain heterogeneous, free-form, and voluminous amounts of information unlike the vehicle used in many studies.Three types of navigational aids will be experimentally manipulated: 1) method of suggesting other items to browse based on a cosine similarity measure: one system-specified and one controlled by the user; 2) method of presenting stable orienting devices: a collapsible / expandable hierarchical menu and a filtering fish-eye view of the menu; and 3) method of navigating: use of sequential paging and use of hypertext links. In this 2x2x2 factorial design, menu presentation and method of navigating will be between-subject variables while method of suggesting items to browse will be a within-subject variable. A fourth aid, content cues that influence navigation at a local level, will not be manipulated but will be investigated. Additionally, the study will evaluate the applicability of an amplified version of Guthrie's information search model to browsing.

The purpose of this study is to: 1) characterize browsing in an electronic environment; 2) assess the impact of three navigational aids on browsing; and 3) study these variables in the context of an information search model.

METHODOLOGY

Sixty-four community college students will browse, search and read both current and back issues of an electronic newspaper four times over a period of four weeks. The database will contain the previous twelve months of the paper (negotiations are currently underway with a Canadian newspaper publisher). During the study, the database will be updated as issues are released so that subjects are examining the current issue of the paper. The database will start with about 250 mb of information and will increase by 20 to 30 mb during the course of the experiment, a potential confound that will be assessed. Four different interfaces (one for each experimental group) will be used for the study and each will vary according to the three types of navigational aids discussed above: 1) menu presentation styles, 2) method of listing items to browse, and 3) method of navigating. Additionally, three non-modal windows will be used so that both dynamic and stable navigational aids are always visible. The third will contain the test of the articles.

Subjects will do two types of tasks: browse the newspaper as is their usual practice and answer specific questions. These types of tasks will be devised based on search goals: implicit and explicit. While explicit goals are typical of many information system studies, implicit goals are unconscious goals that lie dormant in long-term memory or arise from a user's exposure to the information.

Data will be collected using two types of transaction log files. One will contain keystroke/mouse imprints while the second will contain a video' of the session, much like the results from Lotus ScreenCam. Users will walk through a 're-play' of the video, indicating user goals, identifying cues that influence navigation and evaluating information received.

Several types of analyses will be done including: 1) a multivariate analysis of variance of both subjective and objective performance measures such as informativeness [10] of the session, extent of exploration, and accuracy of the task; and 2) a path analysis to validate a model of browsing.

CONCLUSIONS

This study will reveal how users browse in large information systems: what patterns of behavior are exhibited, how performance varies across the two distinctive tasks and how different navigational aids influence patterns of use. This study is especially pertinent today. Searching through the digital haystack is becoming a daily ritual as information is increasingly being delivered in digital form to homes, schools and business. Browsing is without a doubt the key form of novice interaction with those systems. Hence improving our understanding of how users interact is an imperative as we become a self-service information society.

REFERENCES

1. Bawden, D. Browsing: theory and practice, Perspectives in Information Management 3, 1993, 71-85.
2. Cove, J.F. & B.C. Walsh. Online text retrieval via browsing, Information Processing & Management 24 (1), 1988, 31-37.
3. Marchionini, G. & B. Shneiderman. Finding facts vs. browsing knowledge in hypertext systems, Computer, 21 (1), 1988, 70-80.
4. Guthrie, J.T., & I.S. Kirsch. 1987. Distinctions between reading comprehension and locating information in text, Journal of Educational Psychology 79, 1987, 220-27.
5. Kerr, S.T.. Instructional text: the transition from page to screen, Visible Language 20, 1986, 368-92.
6. Guthrie, J.T. 1988. Locating information in documents: examination of a cognitive model, Reading Research Quarterly 23, 1988, 178-99.
7. For example: Smith, P.A. & J.R. Wilson. Navigation in hypertext through virtual environments, Applied Ergonomics 24, 1993, 271-78; Foss, C.L. Tools for reading and browsing hypertext. Information Processing & Management 25, 1989, 407-18; Kerr, S.T. Wayfinding in an electronic database: the relative importance of navigational cues vs. mental models, Information Processing & Management 26, 1990, 511-23.
8. Thorndyke, P.W. & B. Hayes-Roth. Differences in spatial knowledge acquired from maps and navigation, Cognitive Psychology 14, 1982, 560-589.
9. Ashton, E., & G. Cruickshank. The newspaper of the future. In Proceedings, National Online Meeting, 1993, New York, NY, Medford, NJ: Learned Information, 1993, 11-16.
10. Tague-Sutcliffe, J. Measuring Information: An Information Services Perspective. SanDiego: Academic Press, 1995.

Exploring the Information Landscape / etoms@julian.uwo.ca