Computer Science and Engineering
University of Michigan
1101 Beal Av.
Ann Arbor MI 48109-2110
In my current work, I focus on the identification of particularly interesting locations in a spatial information world and the introduction of landmarks -- either designated information objects or explicitly introduced navigational objects -- to guide navigation relative to these location. This work explores a means of increasing the navigability of an information world by decreasing the amount of information necessary and available to the navigator. Selection of interesting locations is based on the information content of the world without assuming that the information itself has any particular structure other than that implied by its spatial layout.
Navigation, Landmarks, View-Navigation, Pad++.
Navigation is the process of getting moving from one location to another by selecting among the options presented at successive junctions on the path between the two locations. Navigation is distinguished from traversal, the act of locomotion between locations, and steering, the process of controlling the locomotion. Navigation differs from querying in that the navigator chooses among the options presented. Querying is a process of choosing among options to be presented where the decision generally rests with some external agent.
Both navigation and querying are orthogonal to browsing and searching. Browsing and searching depend on the purpose of the activity. The purpose of browsing is generally to determine what is available and, possibly, to select something from what is available. The purpose of searching is generally to find something specific, which may or may not be available.
My interest in navigation is to understand what support people need to be able to navigate in computer-based information worlds and how to design and build such support. I am interested in general properties that apply to worlds ranging from personal workspaces to digital libraries to the World Wide Web, and that apply across a broad range of user tasks. My goal is to develop a theory of navigation that can be used to design navigational aids. That is, to understand what support is needed and what support can be provided to help people (or other decision-making agents) find their way around electronic worlds. The main focus of my work is on multiscale worlds -- worlds in which information can exist at different levels of resolution or scale.
My approach is to bring understanding of the psychology of navigation together with understanding of the properties of electronic worlds. This approach relies on two fundamental assumptions.
First, it is assumed that people should and will attempt to transfer real-world navigational knowledge and skills to the electronic world. This raises the important questions -- that I'm hoping others will answer -- of how electronic-world and real-world navigation are related, and of how differing properties of the two affect navigation.
Two differing properties, in particular, seem significant:
Second, it is assumed that electronic worlds have structure and regular properties that control navigability, and that these may be manipulated by the designers of worlds and of navigational aids. This raises questions of what structures are possible, what properties are desirable as well as how these may be manipulated. My work concentrates on the latter.
My work explores supporting navigation by selectively augmenting the world with navigational landmarks. I rely on empirical evidence that landmarks -- visually or conceptually distinct features -- are important to both real- and electronic-world navigation [Kaplan 1976, Watts 1994]. I draw on Furnas' theory of View-Navigation to guide the augmentation of the world. At present, I am developing and testing my ideas in Pad++.
Pad++ [Bederson et al., 1995] is an experimental multiscale information world. Its interactional metaphor is of an infinite two-dimensional surface that may be infinitely magnified. As in the physical world, objects have size and extent, and reside at particular locations on the surface. In addition to panning -- moving about the surface -- the view can be changed by zooming -- magnifying the surface. Pad++ supports geometric zooming, whereby the visual rendering of objects grow and shrink with the surface, and logical zooming, whereby the visual rendering of objects may be changed depending on their magnification.
View-Navigation theory [Furnas 1997] defines two properties of information worlds, view-traversability and view-navigability, and describes requirements that a world must satisfy in order for a user to be able to effectively traverse or navigate that world. In order to be effectively view-traversable the distance to be traveled between any two locations must small relative to the overall world. In other words, it must be possible to move quickly between two locations regardless of the relationship between them. In order to be effectively view-navigable, it must be possible, from any location, to determine a shortest path to any other location, and the amount of directional information at any location must be small relative to the overall world. In these terms, Pad++ is effectively view-traversable but not effectively view-navigable.
The approach I take to increasing the navigability of Pad++ is to select a subset of the locations on the surface and provide navigational aids that indicate paths from any location on the surface to those locations. I rely on three key observations:
The obvious selection of all locations at which information is located fails to be effectively view-navigable under the recognition that this is the very set relative to which navigation is taking place and therefore the set relative to which the amount of directional information must be measured. However, a variety of methods can be used to impart a navigable structure to this set. Under the assumption that people place related objects near each other, I use a simple clustering algorithm in which spatial proximity serves as an approximate measure of relatedness. The simplest variant results in hierarchically nested clusters or trees. Landmarks, either designated information objects or explicitly introduced navigational objects, are then created to mark internal and top-level nodes of this structure. These landmarks represent their associated cluster which, in turn, is assumed to represent a conceptually distinct entity.
I am exploring a variety of indicators to show the paths to these landmarks. At present, I am focusing on route-based indicators that, like road signs, show the direction (and in some cases, approximate distance) to particular landmarks. These are computed from the current view, which may or may not include information objects or landmarked locations. A number of these indicators are designed to take advantage of visual memory in hopes that users can manipulate the view so that the indicator matches their visual memory in order to navigate to a particular location.
Showing the locations of all landmarks at all times violates the view-navigable requirement of small amounts of directional information, so it is necessary to select a small set of landmarks of immediate interest. The "visibility" of a landmark is controlled by assigning it a scope, a region of the surface within which the landmark is visible. The size of the scope of a landmark is related to the area occupied by the information objects that it represents. I am exploring a variety of ways of determining what landmarks should be shown, such as limiting the scope in geometric space but not in scale, showing only landmarks in view or top-level landmarks, showing the closest landmarks, and various combinations of these.
The effect to the navigator is that conceptual groupings of the information on the surface are explicitly associated with individual landmarks, reducing the number of locations needed to navigate. Additionally, there is always directional information to at least one landmark, and the navigator can always zoom out to get information about more landmarks. And, as the navigator approaches a landmark, more navigational detail is revealed as "smaller" nested landmarks come into view.
I am currently a PhD candidate in Computer Science and Engineering at the University of Michigan. The focus of my dissertation is navigation in multiscale electronic worlds. Before starting the doctoral program, I spent six years in the software industry as a software engineer and user interface designer and manager. I hold a master's degree in Computer Science from the University of Washington.
Bederson, B. B., Stead, L., Hollan, J. D. (1994). Pad++: Advances in Multiscale Interfaces. Human Factors in Computing Systems CHI '94 Conference Companion, New York, NY: ACM Press, 315-316.
Furnas, G. W. (1997). Effective View-Navigation.Human Factors in Computing Systems CHI '97 Conference Proceedings, New York, NY: ACM Press.
Kaplan, R. (1976). Way-Finding in the Natural Environment. In Moore, G. T., Golledge, R. G. (Eds.), Environmental Knowing, Stroudsburg, PA: Dowden, Hutchinson and Ross, 43-57.
Watts, J. (1994). Navigation in the Computer Medium: A Cognitive Analysis. Proceedings of the Human Factors and Ergonomics Society 1994, Human Factors and Ergonomics Society, Inc., Santa Monica, CA, USA, 310-314.