Graphical Encoding in Information Visualization

Lucy Terry Nowell
Virginia Tech Department of Computer Science
Blacksburg, VA 24061 USA
Email: nowell@vt.edu

ABSTRACT

In producing a design to visualize search results for a digital library called Envision [5, 7], we found that choosing graphical devices and document attributes to be encoded with each graphical device is a surprisingly difficult task. By graphical devices we mean those visual display elements (e.g., color, shape, size, position, etc.) used to convey encoded information. Research in several areas provides scientific guidance for design and evaluation of graphical encodings which might otherwise be reduced to opinion and personal taste. However, literature offers inconclusive and often conflicting viewpoints, leading us to further empirical research.

Keywords:

information visualization, iconic display, user interface design, graphical encoding.

PROBLEM - DESIGNING GRAPHICAL CODES

In producing a design to visualize search results for a digital library called Envision [5, 7], we found that choosing graphical devices and document attributes to be encoded with each graphical device is a surprisingly difficult task. By graphical devices we mean those visual display elements (e.g., color hue, color saturation, flash rate, shape, size, alphanumeric identifiers, position, etc.) used to convey encoded information. The challenge for a user interface designer is to choose graphical devices to support the range of tasks users are likely to perform with an application, while also supporting perceptual and individual differences of the user population.

Providing access to graphically encoded information requires attention to a range of human cognitive and perceptual activities, explored by researchers under at least three rubrics: psychophysics of visual search and identification tasks, graphical perception, and graphical language development. Research in these areas provides scientific guidance for design and evaluation of graphical encodings which might otherwise be reduced to opinion and personal taste. Especially useful are rankings of the effectiveness of various graphical devices in communicating different kinds of data (e.g., nominal, ordinal, or quantitative). Christ [2] provides such rankings in the context of visual search and identification tasks and provides some empirical evidence to support his findings. Mackinlay [6] suggests rankings of graphical devices for conveying nominal, ordinal, and quantitative data in the context of graphical language design, but these rankings have not been empirically validated [personal communication]. Cleveland and McGill [3, 4] have empirically validated their ranking of graphical devices for quantitative data. However, information visualization designs which resemble scatter plots (e.g., starfield displays [1], air traffic control displays, and other iconic representations of data) often support tasks quite different from those for statistical graphs. Rankings suggested by Christ, Mackinlay, and Cleveland and McGill are not the same, while other literature offers more conflicting viewpoints, suggesting that further research is needed.

RESEARCH IN PROGRESS

Test Bed:

Named after Tufte's book [8], Envision is a multimedia digital library of computer science literature, with full-text searching and full-content retrieval capabilities, serving computer science researchers, teachers, and students at all levels of expertise. A unique characteristic of Envision, the Graphic View Window (see figure), graphically presents each document in a search results set graph as an icon, while the Item Summary shows a textual listing of bibliographic information for documents whose Graphic View icons are selected by the user. The Graphic View supports users in making decisions about which works to examine in potentially large sets of documents. Since users' perceptual strengths vary and users' decision criteria reflect their current information needs, each graphical device in the Graphic View is user-controllable to represent different document attributes as a user desires. Thus the Envision user interface design suggests a number of experimental studies.

Experimental Design:

Using the Envision Graphic View, we are conducting a within-subjects empirical investigation of the effectiveness of three graphical devices - icon size, icon shape, and icon color - in communicating nominal (document type) and quantitative (document relevance) data. We have chosen these graphical devices because of their widespread use and expected power in communication, combined with the uncertainty of their actual impact. We use three levels (i.e., representing nominal or quantitative data, or uncoded) for each of the three factors (e.g., color, size, and shape).

Using SuperCard, each trial presents a search results display captured from Envision. Each subject is asked to count the icons representing documents that meet given conditions, where the information about those conditions is graphically encoded in the display. Trials are divided between training and measured trials. Since a given design point may present multiple options for information extraction (e.g., using a single code out of several presented, or some combination of the codes), tasks are balanced among the options, thus enabling us to study interaction of codes with one another. Objective measures are accuracy and time for task completion. Subjects are also asked to rate each design point for cognitive difficulty and for desirability as an information source.

EXPECTED RESULTS

Our studies are providing empirical evidence of the effectiveness of icon color, shape, and size in conveying both nominal and quantitative data. Because our quantitative measures are also widely used in formative usability evaluation while our qualitative measures are aimed solely at subjective measurement of usability, we expect to provide empirical evidence regarding the effectiveness of psychophysical foundations for designing information visualization displays. The resulting rankings of graphical devices will either confirm those suggested by other authors or support new rankings. Such rankings are fundamental to developing effective information visualization. We may also be able to indicate which combinations of graphical devices are "good" or "bad" for presenting information.

ACKNOWLEDGMENTS

We gratefully acknowledge the support of the Envision development team, especially its managers Dr. Deborah Hix, Dr. Lenwood S. Heath, and Dr. Edward A. Fox, and members Robert France, Eric Labow, Dennis Brueni, Kaushal Dalal, Scott Guyer, Stephen Moore, and William C. Wake. Envision was funded by National Science Foundation grant IRI-911699 (Maria Zemankova, monitor),Virginia Tech, and the ACM. This work is also supported by Lynchburg College in Virginia.

REFERENCES

1. Ahlberg, C. and Shneiderman, B. Visual information seeking: tight coupling of dynamic query filters with starfield displays, Proceedings of CHI 94 (Boston, MA, April 1994) ACM Press, 313-317 & 479-480.

2. Christ, R. E. (1984) Research for evaluating visual display codes: an emphasis on colour coding. In R. Easterby and H. Zwaga (Eds), Information Design: The Design and Evaluation of Signs and Printed Materials, New York: John Wiley and Sons, 209-228.

3. Cleveland, W. and McGill, R. (1984) Graphical perception: theory, experimentation, and application to the development of graphical methods. Journal of the American Statistical Association 79(387), 531-554.

4. Cleveland, W. and McGill, R. (1985) Graphical perception and graphical methods for analyzing scientific data. Science, 229(August), 828-833.

5. Heath, L.S. et al. Envision: A user-centered database of computer science literature. Communications of the ACM, 38, 4 (April 1995) 52-53.

6. Mackinlay, J. (1986) Automating the design of graphical presentations of relational information. Transactions on Graphics, 5(2), 110-141.

7. Nowell, L.T. et al. Visualizing search results: some alternatives to query-document similarity. In Proceedings of SIGIR 96 (Zurich, Switzerland, August 1996), ACM Press, 67-75.

8. Tufte, Edward R. (1990) Envisioning Information. Cheshire, CT: Graphics Press.