Steven F. Roth
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh, PA 15213
Visage is a prototype user interface environment for exploring and analyzing information. It represents an approach to coordinating visualizations and analytical tools in data-intensive domains. Visage is based on an information-centric approach to user interface design which strives to eliminate impediments to direct user access to information objects across applications and visualizations. Visage consists of a set of data manipulation operations, an intelligent system for generating a wide variety of data visualizations and a briefing tool that supports the conversion of visual displays used during exploration into interactive presentation slides.
The VISAGE user interface paradigm takes an aggressively information-centric approach to the presentation of information to the user. The information- centric approach is the next logical step along the path from application-centric architectures to the modern document-centric approach. The distinctions among the three approaches hinges on differences in the "basic currency" through which the users interact with the system.
In application-centric architectures, the basic currency is the file. With the introduction of graphical user interfaces and the desktop metaphor, files became concrete visual objects, directly manipulable by the user, storable on the desktop or in folders, and--to a limited extent--arrangable by users and software in semantically meaningful ways. But the contents of those files were still out of direct reach of the user. With the document-centric interface paradigms the basic currency is no longer the file, but rather the document. The application's role is subordinated and the user focuses on direct manipulations of documents; users can almost literally "get their hands on" their documents.
The information-centric approach represented in Visage simply is a natural continuation of these trends. Visage abandons the primacy of the document wrapper as the central focus of user interaction in favor of the data element as the basic currency of the interface. Visage permits direct drag-and-drop manipulation of data elements at any level of granularity. A numeric entry in a table, selected bars from a bar chart, and a complex presentation graphic are all first-class candidates for user manipulations.
Hints of this approach may be found in a few existing interfaces; recent versions of Microsoft Word support the ability of the user to drag selected text from one place in the document to another. In Visage, this capability is promoted from a special-purpose feature to a ubiquitously available capability that can be used everywhere in the environment. It becomes part of the interface 'physic', empowering the user to perform directly unique actions that usually require knowledge of specialized interface "features."
In order to convey the basic elements of Visage, it is useful to consider some simple displays and operations in a detailed example. The example is based on one of the possible applications of this approach, and uses the next generation logistics tracking and planning systems. These systems are being developed to enable analysts to access information about the location, quantities, status, transportation, distribution, consumption of equipment and supplies and the people who need them worldwide. One of many displays that can be produced in the Visage environment is an outliner-style table. It provides a hierarchical perspective on tabular data and illustrates drill-down and roll-up capabilities in a familiar way.
Starting from any point in a data network users are offered a menu of alternative dimensions along which they may drill down. The user can choose to drill down further by selecting any of the relations from the pop-up menu, dynamically attached to the unit on which further information is desired, giving a more detailed organizational breakdown. This drill-down process can also occur across different relations. For example, it is possible for a user to drill-down from a division to the equipment it possesses and then from one type of equipment to the parts or supplies it requires and then to all the units that require a particular supply type. This is a process of turning a network or web of relations into a convenient hierarchical breakdown for navigational purposes. Users can select any attribute of the objects in the hierarchy that they want to have displayed, like its echelon or the number of people in the unit. As the hierarchy is expanded, the data attributes are expanded with it. Dynamic drill-down and expression of attribute values are fundamental in Visage and occur in all displays.
An important operation in this information-centric interface approach  is the movement or transfer of information throughout the Visage environment. Objects representing information are moved directly in groups or individually among visualization and application interfaces. In order to display some of the attributes of these units graphically, one simply drags the unit names from the outliner to an empty bar chart display. In this case, a bar chart shows the weight of supplies that the units require. Displaying unit supply weights in a bar chart makes it easy to select units requiring the most supplies - those with the longest bars. Focusing on the units with highest supply weights provides an opportunity to display the locations of just these units on a map, perhaps to determine the locations where supply points should be established. Units are dragged and dropped, same as before, to a map application. The map application is a product called MATT which was developed prior to and independently of Visage by Bolt, Beranek and Newman. This is an example of one of the primary goals of this work: to provide the mortar necessary to cement separately developed analysis tools into what the user will experience as an integrated work environment. By default, a map display shows longitude/latitude attributes of objects and the unit's associated military symbology.
By color highlighting (painting) a subset of units that occur close together in the center of the map, for example, (perhaps to identify a region where large quantities of supplies will be needed) multiple displays may be visually coordinated. Notice that painting an object in one display causes it to be similarly colored in all other displays. Together, the three displays show the selected units, the supplies they need, and where they are located.
Finally, this small subset of units can be rolled-up (i.e. aggregated) into a single object and named by a user, in this case, "Task Force A." It appears in the bottom of the outliner and its attributes are the appropriate totals of the those for the units it aggregates. The new aggregate can be treated as a single object for new drill-down operations. This new entity can then be drilled down to show other units, their support, inventory or other information. Note that this approach enables users to compose a complex query through a series of drill-down, drag, paint and roll-up operations. The last aggregate representing those units of a division which need the "most" supplies and are located in a small region to the north. In Visage such complex queries may be composed entirely by direct manipulation.
Also shown here is Sage, a knowledge-based presentation system that designs displays combining diverse information (e.g. quantitative, relational, temporal, hierarchical, categorical, geographic). We use Sage to automatically assume the burden of visualizing information for other applications and for computer- supported graphic design, especially for large data sets. Complete integration between Visage and Sage is the our goal during the next year.
The Visage environment has a simple briefing or "slide show" feature completely integrated with the rest of the application. As analyses are performed, text and graphics can be captured and saved in special frames called 'slides.' A slide is simple a frame designed to make it easy to "paste up" other frames and elements for visual presentations. The user simply "drags and drops" the desired displays onto the slide frame, where they are scaled appropriately. Note that elements on the slide do not lose their separate identity--they are still fully- functional interface objects that can be used for further analysis--even in the middle of a briefing. Collections of frames are accumulated in a "slide sorter" frame, used to sequence a presentation by simple drag operations, and the sequential display of each slide at full-screen size. Thus, the briefing function has been seamlessly integrated with those of data exploration and analysis.
The authors would like to acknowledge the contributions of Jeffrey Senn, Michael B. Burks, Philip Stroffolino, John Kolojejchick, Carolyn Dunmire, Noah Guyot, Hugo Cheng, Kenichiro Tanaka, Michelle Vincow, Joseph Mattis, and Bolt, Beranek, and Newman. This work was supported by contracts from ARPA and Army Research Laboratory, and Strategic Logistics Agency.
ęCopyright on this material is held by the author(s).