Stephen Hirtle
University of Pittsburgh
School of Information Sciences
135 N. Bellefield
Pittsburgh, PA 15260
E-mail:
sch@sis.pitt.edu
URL:
http://www.pitt.edu/~hirtle/hirtle.html
Navigation in electronic environments is a particularly difficult problem given the multiplicity of navigational strategies, the varied goals of users, and the combinatoric problems associated with unrestricted topologies of networks. One approach to the problem, which we have found to be beneficial, is to compare navigation in the physical world with navigation in electronic worlds, with a focus on the underlying cognitive structures and the implicit metaphors that are adopted by the navigators of the space (Kim & Hirtle, 1995). Kim and Hirtle (1995) have argued that some of the difficulty in traversing virtual spaces, such as the World Wide Web, is due to the lack of identifiable neighborhoods and notable landmarks. However, with little additional effort, these items could be easily integrated into existing environments. Appropriate formal analyses can also lead to the development of intelligent views of a space, such as modified fisheye views and other "you-are-here" pointers for electronic worlds.
Research on cognitive mapping has examined the ability of individuals to acquire and use spatial information. The acquisition of spatial knowledge has been shown to be based on the use of organizing principles, such as the use of hierarchies, reference points, rotational and alignment heuristics and other related principles. These organizing principles, in turn, result in what Barbara Tversky has coined a "cognitive collage" of multimedia, partial information. Inherent within this collage is the ability to extract slices of information sources, such as visual cues, route information or linguistic labels. The collage necessarily operates at multiple levels, allowing one, for example, to discuss and plan a route, using highway systems or our own back alley with equal ease. In our own lab, we have shown that the need to structure space is so strong that subjects will impose hierarchies on an otherwise homogeneous distribution, which results in consistent bias of judgment (McNamara, Hardy, & Hirtle, 1989).
Given this background, a theory of spatial information must then include at least three levels: (1) a theory of the cognitive representation of space; (2) a theory of the use of spatial metaphors in the interface between user and computer; and (3) a theory of the storage of spatial data. The claim is made here and elsewhere (e.g., (Medyckyj-Scott & Blades, 1992), that spatial information will be useful to the extent that it mirrors the internal representation (level 1) and that the implied metaphor of the interface matches the adopted metaphor of the user (level 2).
The success of implementing or using spatial concepts will depend in part on the user's ability to understand or comprehend spatial knowledge. Therefore, the first step is consider how spatial information is processed and stored by individuals, not by the spatial information system. Likewise, users of information system typically adopt a physical metaphor for understanding and interpreting the command systems (Kuhn, 1993; Norman, 1988). Appropriate metaphors can lead to improved system usability, whereas inappropriate metaphors can lead to decrement in performance and user errors. The need for the formalization of metaphors for spatial reasoning, as proposed by Kuhn and Frank (1991), is necessary for the second level of the theory of spatial information.
Given this framework, numerous research questions are suggested. How is a virtual space like a real space? What are appropriate metaphors for the traversal of virtual spaces? What is useful spatial information, and in what format is the information useful? Such questions can only be answered through the development and study of user interactions.
As an example of how the three levels described above interact with one another and the resulting empirical investigations that are necessary to build a complete a theory of spatial information, consider the role of spatial cognition for the exploration of hyperspace. Several authors have noted that in many systems there is a serious problem of quickly getting lost (see Kim & Hirtle, 1994, for a review). This "lost-in-hyperspace" phenomenon occurs for several reasons. First, real space has real constraints, whereas hyperspace does not. Nodes might join in a strict linear order, a tree, a network, a cycle or any number of other topologies. Some topologies are indicative of a book, others of a museum, and others of an unorganized wilderness. You-are-here maps are either absent or uninformative when present.
One solution to this problem is to turn to the heuristics that people use to understand space. There are different types of spatial knowledge, such as route and survey knowledge. In simple spaces, individuals begin to acquire survey knowledge upon the first exposure to the space, whereas in complex spaces, such as hospitals, survey knowledge is rarely acquired even after years of experience. There are individual preferences for the presentation of spatial information. Thus, there is a need to develop different geometries or topologies to represent the vast differences in how space is encoded and accessed in a cognitive map.
Furthermore, aspects of the representation can be generalized to the characteristics of the physical space. For example, architects and urban planners have learned that undifferentiated spaces are harder to learn than rich environments. Even an idea as simple as using different colors on different levels of a parking garage will increase the likelihood of recalling upon return where your car was parked. Thus, aids in helping the user structure space and differentiate neighborhoods should lead to fewer errors and greater satisfaction with hypertext systems (Kim & Hirtle, 1994).
A second solution is to consider the second level of analysis and examine the metaphor that users adopt in hyperspace (Gray, 1990; Kim & Hirtle, 1994). Here the focus is on the relationship between the virtual space and the users' understanding of the virtual space. A critical observation is that the virtual space need not have a physical correlate to be easily traversed, and the inclusion of a physical correlate does not guarantee avoiding disorientation. For example, understanding the mapping of a video game that assigns the top row to the bottom row, and the left edge to the right edge is easily understood and visualized, even if it is physically impossible in real-space. Likewise, people may find themselves lost in a museum of interconnected rooms and the corresponding hyperworld would be equally disorienting (Foss, 1989). Thus, disorientation is often the result of either adopting the incorrect metaphor or the lack of an appropriate metaphor.
On-line aids, such as history trees, maps, and fish-eye views, can assist the user both in developing an appropriate metaphor and locating oneself in the virtual space. Pointers with some degree of redundancy will tend to more useful. However, the exact methods which prove to be of the most use in a given situation will depend on the structure of the virtual space and the preferences of the user. Rarely do most information systems build on both of these factors.
In our own lab, we have most recently begun to explore these hypotheses in hypertext navigation, by examining the role of (1) imposing structural cues in the virtual space and (2) presenting multi-modal information to a user wishing to locate a building on a university campus. In the first study, we contrasted navigation through a hypertext space with and without implicit neighborhoods defined. In the second study, users wishing to locate a building on campus are presented with a four-panel multi-media display that allows for efficient navigation and redundant perceptual and semantic cues. Together, these studies highlight the benefits and problems in generalizing about navigational behaviors between real space and virtual space.
In addition, we have also begun to focus on alternative geometries for modeling electronic worlds. For example, hierarchically structured geometries or those that distinguish between local and global relations might prove to be of greater value than traditional Euclidean geometries.
The need for greater use of spatial metaphors in information systems has been suggested. Furthermore, the need for understanding spatial cognition as a prerequisite to successful use of spatial metaphors has been argued. The multi-layered approach of considering spatial data, spatial interfaces, and spatial constructs on the part of the user, and the interconnections between these layers, remains a critical area for future study.
Hirtle, S. C., Sorrows, M. E., Cai, G., & Roberts, L. (April, 1997). Navigation in Real and Virtual Spaces, Paper presented at the Annual Meeting of the Association of American Geographers, Ft. Worth, Texas.
Hirtle, S. C. (1995) Representational structures for cognitive spaces: Trees, ordered trees and semi-lattices. In A. V. Frank and W. Kuhn (Eds.), Spatial information theory: A theoretical basis for GIS. Berlin: Springer-Verlag.
Hirtle, S. C., & Heidorn, P. B. (1993). The structure of cognitive maps: Representations and processes. In Garling, T. & Golledge, R. G. (Eds.), Behavior and Environment: Psychological and Geographical Approaches (pp. 170-192). Amsterdam: North- Holland.
Hirtle, S. C. (1991). Knowledge representations of spatial relations. Doignon, J.-P., & Falmagne, J.- C. (Eds). Mathematical Psychology: Current developments. (pp. 233- 250). New York: Springer-Verlag.
Kim, H., & Hirtle, S. C. (1995). Spatial metaphors and disorientation in hypertext browsing. Behaviour & Information Technology, 14, 239- 250.