Reconsidering information-seeking behavior as navigation leads us to ask what a path through knowledge could mean. we claim that these paths can be created to support useful tasks, and we offer principles to guide the design and presentation of paths that communicate.
A businessman en route to his office purchases a newspaper from a vendor. A student finds a book and checks it out from the university library. A woman searches for job listings using the World-Wide Web. Each of these individuals is participating in the process of information-seeking. Information-seeking is the domain of human behavior concerned with actively acquiring information from one's environment, usually to fulfill larger problem-solving goals [Marchionini, 1995]. For the businessman, that goal is more effective business decision-making that day; for the student, the completion of a research paper; for the job-seeker, employment. information-seeking is a continuous, circumstantial process that is a part of individual problem-solving behavior at every level of description.
information-seeking is also a problem-solving behavior in its own right. It begins when the information-seeker realizes that he is lacking some knowledge which will help him fulfill a higher-level goal. This in turn leads to the gradual formulation of an information need: some description of the knowledge that would allow completion of the higher-level task [Taylor, 1962]. This need, in turn, is translated into actions to retrieve media with the needed information.
An example. The student mentioned earlier is assigned a research paper in class. Having chosen a topic, say, nose morphology in vampire bats, he asks the professor for suggested sources; he mentions that there are two recent articles in the Journal of Vertebrate Zoology, but then rushes off to a faculty luncheon before he can provide any additional details.
The student goes to the university's library and uses its electronic catalog to retrieve the call number of the journal by a keyword search on ``vertebrate zoology,'' and takes last year's volume off the shelf. Scanning the tables of contents, he finds the articles in question.
The example points out some of the kinds of knowledge and skills necessary to successfully complete an information-seeking task in a library setting:
What makes effective information-seeking possible in this setting is the fact that the material in the library has been organized in a manner understood by its users, and that they can successfully apply knowledge of that organization to find material that meets their information need. In this case, the student was able to use the electronic catalog to match a journal title to a call number, and the layout of the library to match a call number to a location, to translate a need - ``articles in a recent journal volume'' - to a particular book in the stacks. It is precisely these kinds of structures that define what an information access environment is: a set of tools to allow a user to translate an information need into information-seeking actions to retrieve information that meets the need. As more and more information has become available and accessible in digital form, an important question has arisen that asks what kinds of tools our electronic information access environments should have.
A number of techniques have evolved for constructing electronic information access environments that organize on-line information. At first, the main organizational tool was the hierarchical directory structure provided by the file system of a personal computer or a time-shared mainframe. The user's personal information access environment consisted of his local disk or home directory, and information-seeking proceeded by moving from directory to directory and listing their contents. The user could distribute content to others by copying files to a shared directory or removable media. Organizing a large number of files required giving meaningful names to directories and subdirectories, and ensuring that a file was kept in the directory in which the user would most likely look for it. This method, the use of a hierarchical classification, has long been used outside of the electronic environment (for example, in library subject classifications [Immroth, 1980]).
Full-text indexing was developed to allow users to find documents containing words given by the user [Salton, 1968]. It has proven very successful at satisfying information needs in which the user knows words specific to documents meeting the need and unlikely to be in irrelevant documents. Because it requires only that the full text of the documents is available, it can be used in conjunction with other forms of organization, such as a hierarchical classification.
In the 1980's, the idea of linking individual documents or parts of a
larger document together into a hypertext began to be implemented on a
wider scale with the development of commercial authoring and browsing
tools (for example,
).
The standarization of hypertext markup languages and transmission
protocols, the distribution of free browser software, and the
phenomenal growth of the the Internet all contributed to the creation
of the World-Wide Web in the early 1990's. Hypertext is currently the
primary organizational method for published online content. The
techniques mentioned previously have each been successfully applied to
assist information-seeking on the World-Wide Web with site catalogs
and Web search engines.
Continuous advances in the underlying media and networking technologies have not only changed how information is accessed and organized, but the amount of information available to the average user. More and more of the growing body of human knowledge is available to a significant proportion of the population. At first, the issue was the the availability of information on-line: Am I able to access what I want at all? Now, however, the issue is how to find information that is most relevant to the task at hand: How do I use the organization of the World-Wide Web to meet my information need? Many individuals are concerned that their information-seeking strategies are insufficient - useful information is available to them of which they are unaware, or cannot find. One commentator calls this ``information anxiety'' [Wurman, 1989].
In response to these issues, we propose that reconsidering information-seeking behavior as navigation through an information space gives insight into how we should structure our access environments to assist the information-seeker - namely, to use the same abilities that people use to navigate in the physical environment. We also claim that these spaces should be designed to facilitate a specific information access task.
To clarify what is meant by navigation in the context of information access, it is useful to state a list of ground assumptions about information-seeking in electronic environments (see Figure 1-1).
From these assumptions, we can characterize information access as navigation through an information space consisting of locations, views into the space through the intermediary window, and moves that change one view to another. This way of describing the process gives us a new language for both the analysis of existing information access environments, and the design of environments that explicitly support navigation.
Each of the information access techniques described earlier - hierarchical classification, full-text indexing, and hypertext - can be thought of as defining a space of possible locations and moves. The next chapter examines how effectively (or ineffectively) those techniques provide affordances for navigation, and we conclude that effectively navigable information spaces must be explicitly designed with those affordances.
The question then turns to how we can take some collection of information and organize it into an effectively navigable space. Answering this question motivated the study of a kind of physical information space: educational museum exhibits. The analysis of seven educational museum exhibits and interviews with their developers and designers produced a collection of design principles for organizing information into spatial paths that communicate.
To test the validity of those principles, they are used to guide the design of two novel information spaces. The first, an early experiment in designing an information space, provides a survey view of the articles in the Journal of Artificial Intelligence Research for search, browsing, and retrieval. The second, the Course VI Information Space, provides both survey and immersive views of the subject listings for Electrical Engineering and Computer Science at the Massachusetts Institute of Technology.
The main points of the thesis are:
