3D or Not 3D: "More is Better" or "Less is More"?

Keywords

Introduction

Diane Schiano (Moderator) is a cognitive psychologist at Interval Research Corporation whose work focuses on visual-spatial thinking.

More is Better? A Critical Look at a Common Assumption.

Until very recently, 3D displays were used primarily to render three-dimensional objects and data sets in specialized domains such as CAD/CAM and geometric modeling. But 3D representations are now commonplace in general-purpose analysis and visualization packages, often in the very controls provided for data manipulation. Interest in 3D visualizations - including some described here - has grown beyond the confines of the CHI and SIGGRAPH communities and captured the public imagination through the print and broadcast media, as well as such Hollywood blockbusters as Jurassic Park.

This innovative, entertaining approach has become so popular that its ultimate ubiquity in GUI design is rapidly becoming part of the received wisdom of the industry. The attractiveness of 3D displays may reflect many factors, from a simple technology-driven delight in surpassing the all-too-familiar WIMP interface, to an implicit belief in the inherent efficacy of more "natural" perceptual processes and more "realistic" visual displays. While much of the work is of real significance, a tendency to conflate style and substance is also common.

We believe the time is ripe for a considered examination of the implications of this growing phenomenon, including an assessment of the costs (in terms of the complexity of implementation and use, beyond the obvious hardware expense) and benefits of choosing 3D solutions to specific interface design problems. This panel discussion gives CHI'95 the opportunity to initiate informed discussion on a topic of growing significance for our community, as well as to help frame the issues in terms of implications for design.

Panelist's Statements

George Robertson is a Principal Scientist at Xerox PARC, in the User Interface Research area, working on 3D inter- active animation interfaces for intelligent information access. He is the architect of the Information Visualizer. In the past, he was a Senior Scientist at Thinking Machines, a Senior Scientist at Bolt Beranek and Newman, and a faculty member of the Computer Science Department at Carnegie-Mellon University. He has made contributions to machine learning, multimedia message systems, hypertext systems, operating systems, and programming languages.

Highly-coupled, Perceptually-Intensive User Interfaces.

The question of "3D or Not 3D" is not well-formed. Techno- logical advances in graphics and processor performance make new user interface paradigms practical. 3D is just one of the new capabilities. The characteristics are highly-cou- pled, perceptually-intensive interaction. The proper ques- tion is, how can these capabilities be used to enhance hu- man capabilities?

There are two broad classes of enhanced human capabili- ties. First is extension of perceptually-oriented capabilities. (1) The perceptual vocabulary available for representation and visualization is increased. (2) Animated transitions ex- tend representation from space into time, communicate transformation, and support object constancy.(3) Natural 3D entities can be naturally represented. And (4) better" impedance matching" to human perceptual operators is possible.

Second is enhanced support of human activity. (1) Larger workspaces that lower the cost structure for activity are possible. (2) People are made dramatically smarter by pro- viding external cognitive working memory. Finally, (3) more peripheral channel and back channel communication is possible. It is through their effect on specific human ca- pabilities that highly-coupled, perceptually-intensive inter- active systems can be understood. 3D contributes to these effects and hence is an advance, but it is not the most im- portant contributor.

Joel Tesler is a Member of the Technical Staff at Silicon Graphics Computer Systems. Along with Steven Strasnick, he invented the Information Landscape for the visualization of large amounts of data. He was also part of the design team for the Silicon Graphics Indigo Magic Desktop. He has a B.S. in Electrical Engineering and Computer Science from the University of California at Davis. Prior to joining Silicon Graphics, he held a position at Hewlett Packard.

The Information Landscape

The information landscape is a paradigm for displaying large amounts of multivariate data. It combines the simplicity of 2D layout and navigation with the power of 3D visualization. In our environment, the user is able to "fly" over a plain, on which objects are laid out representing the data. Our implementation places many 3D bar charts in the landscape. The bar charts are connected to one another, indicating relationships. Examples built with this technology include a filesystem viewer, retail sales analysis, state budgets, and an org chart.

This model nicely combines benefits of 2D and 3D. 3D al- lows display of additional information through the use of 3D attributes such as height. Also using 3D perspective al- lows viewing more data simultaneously than a 2D view would permit. For example, we can quickly search through thousands of bar charts, which would be much harder using a traditional 2D layout using scrollbars and zoom wheels. At the same time, by limiting the layout to a 2D plain, nav- igation through the objects is simpler than through a full 3D network. The number of degrees of freedom is lowered, and the model resembles our day-to-day navigation through the real world.

Barbara Tversky is a cognitive psychologist at Stanford University working on visual thinking and production and memory of graphic displays.

Perception & Cognition in 2D and 3D Graphics

Graphic displays serve many functions. A major one is to facilitate memory, partly by using space as a cue, partly by turning recall tasks into recognition ones. Another function of graphic displays is to convey meaning more directly and succinctly than through words. Meaning can be conveyed representationally as in pictographic languages and simple computer icons, or metaphorically, as in graphs and charts, that make abstract ideas more concrete and accessible. New technology has made the third dimension and animation available for graphic displays. Clearly, these improve rep- resentational displays, like CAD/CAM or navigation.

For conveying more abstract information, the third dimen- sion and animation can be used or abused. Each can present visual problems, for example, 3D graphics can be unstable and animation can be hard to perceive and remember. These devices, however, can be used metaphorically in meaning-ful ways. Depth conveys distance, so that things in the fore-ground are in the focus of attention. Animation can be used to convey change or transformation. But, these devices can also be misused, conveying information less successfully than simpler displays.

Kevin Mullet is a human interface engineer at Macromedia, where he is designing new tools for multimedia authoring. He has worked previously as an interface designer for Sun Microsystems, SunSoft, SunPro, and Aaron Marcus + Associates, and is an author of the recent book, Designing Visual Interfaces: Communication Oriented Techniques (SunSoft Press/Prentice-Hall, 1994).

3D Representations: Tools or Toys?

It has long been rec- ognized that the underlying structure of an argument can be obscured by the technical virtuosity of its presentation. The same risk is apparent in the trend toward increasingly richer interface representations - at least to the extent that this new realism does not serve legitimate communication goals. Consumers clearly enjoy the "3D button" or "virtual of- fice" metaphor, but are hardly in a position to assess their human performance costs.

Under what conditions does a true 3D representation add to the effectiveness of an information display? This is a cen- tral question that ought to addressed by the research com- munity. As advocates of user-centered design, we have a responsibility to evaluate ideas on the basis of their tangible benefits to the user rather than our own fascination with a hot new technology or our boredom with the familiar WIMP interface. 3D representations hold much promise, but only if used in ways that exploit their strengths while avoiding their weaknesses.

It seems reasonable to suppose that any visualization whose third view dimension does not carry a data dimension would be better served by an optimized 2D display. To test this hypothesis, a 2D analog of the Spiral Calendar (from the Xerox Information Visualizer toolset) was created in Macromedia Director to study the contributions of real- time interaction dynamics and 3D representational structure. The resulting DynaCalendar shows that an effective two-dimensional spatial representation can be as useful as a sophisticated geometrical model in navigating through large hierarchical data sets.