Presenting Visual Information Responsibly
Vol.33 No.3 August 1999
Knowing What We Don’t Know; How to Visualize an Imperfect World
Sources of Imperfection
Understanding reality or concepts by examining data and information depends on the quality of the data and information and the quality and effectiveness of the presentation (see Figure 1).
There are a number of factors that potentially could reduce the quality of data and information. These factors are schematically represented in Figure 2.
In more detail, these factors are:
To represent what we do not know and to be cognizant about the imperfection of the data and information is a challenge to both the visualization developer and to the user.
Representation of Degree of Imperfection of Data and Information
An object can be represented by point, line, surface, volume, etc. The degree of imperfection (e.g., error in location of an object, uncertainty of existence, etc.) of what we know about a displayed object is an associated piece of information about that object.
Associated information can be generally represented by variations of the object appearance (intrinsic representation). For example, the object is a point and the color (e.g., , ) represents different values of degree of imperfection).
Another way of representing associated information is by depicting associated objects in the proximity of the real object (extrinsic representation). For example, in the following graphic, the object is the plane and the symbol associated with it conveys a doubt about its existence.
Intrinsic Representations of Imperfection
Jacques Bretin  presented seven visual variables (position, size, brightness, texture, color, orientation and shape) to represent relationships, of resemblance, order, value and proportion. In addition, the associated information on imperfection could make use of variables such as boundary (e.g., thickness, texture and color), blur, transparency, animation and extra dimensionality (e.g., a point could be depicted in 3D).
Extrinsic Representations of Imperfection
These include objects such as question marks, arrows, bars, dials, thermometers, objects of different shapes, complex objects (e.g., pie chart, series of graphs or bars and complex error bars).
Examples of intuitive visual metaphors and cues for representation of imperfect information are given in the sidebar.
Presenting in an inappropriate manner can prevent the user from comprehending the information, or reduce the rate at which it can be absorbed and understood. This is a different issue than finding the best way to represent an acknowledged degree of imperfection as discussed above.
Some of the sources of imperfect presentation include:
This can be the result of a number of causes:
Too sophisticated. A presentation can be too sophisticated for a particular group of users, thus preventing them from understanding the information, and doing so promptly.
Inappropriate spatial metaphor. There are always a number of alternative choices in representing abstract information in a physical space (e.g., paper, computer screen). Finding the right spatial metaphor for abstract information is one of the challenges of information visualization . For example, a tree representation in two dimensions is suitable for displaying a small organizational hierarchy while for large hierarchies, a ConeTree 3D representation  might be more appropriate. Choosing the wrong metaphor (e.g., a 2D tree for very large hierarchies) might confuse the user and thus lead to imperfection in information representation.
Visualization. The visualization of information can be done in a way that misrepresents the information contained in the data. For example:
Clutter is another example of imperfection - too many details can overwhelm the users, preventing them from comprehending the information. Clutter can even produce information fatigue, as well-known to many users of email! Clutter can be both spatial and temporal (presenting the information too fast).
Representing all of the sources of imperfection can create an overload on the part of the user. One way to improve the situation is to represent imperfection sparingly (e.g., a few error bars at large intervals). Users then can interactively choose what details or sources of imperfection they would like to see at a particular time or at a particular region of the visualized scene. Examples of pieces of information that the users might like to selectively hide or represent include errors in object locations; their speed if they move; the age of the information; doubt about object existence and tracks of moving objects.
Another form of information overload can occur if the presentation includes too many pieces of irrelevant information (“eye candy”), distracting the viewer attention from the important pieces of information.
User diversity is another issue that can “make” the visualization more imperfect. No two users are alike. Yes, there is a statistical average, but individual frames of mind, capabilities, education, etc. differ from one user to another. We thus need to develop techniques for visualization management that allow for the tailoring of visualization to particular problems and users.
Bear in mind that a presentation designed on one device may not be appropriate for display on another. For example, a laptop presentation might not be fully effective if displayed on a device with a lower resolution, perhaps with no color – for example, a palmtop or a cell phone. The user or viewer of the palmtop may have trouble reading the information as effectively or accurately as designed on the original device.
The Last Word
Life is not perfect and will never be. So, it is unrealistic to expect that data and information as well as their representations will be perfect. We thus need to become accustomed to (or at least be at peace with!) this fact and not to expect decision making based on perfect data, information and presentation. In that regard, we need to develop principles of how to understand imperfect information using imperfect representations, and reach sound decisions in real-world conditions (imperfection or “uncertainty” management).
Alas, finding ways to do so is not always an easy challenge, since our perceptions, experiences, biases and feelings sometimes dominate our logical mind.
This paper is based on the work done at the MITRE Corporation and on discussions at the past seven annual sessions of “How to Lie and Confuse with Visualization” (VisLies) which took place at IEEE Visualization ‘92-’98 with many enthusiastic attendees. The helpful discussions with Pierre-Yves Bertholet, Jack Berkowitz, Bill Ide, Jim Garrova, Ari Pernick, Mark Pearson, Roger Braunstein and Daniel Haspel are also greatly appreciated.
Nahum Gershon is a Senior Principal Scientist at The MITRE Corp. His work is concerned with information and data visualization, network browsers, image processing, data organization and analysis of medical, environmental and other multidimensional data. He pursues research in the use of understanding of the perceptual system in improving the visualization process.
Gershon has published extensively in the area of visualization and has organized and chaired seven SIGGRAPH panels. He served as a Co-Chair of Visualization ‘94 and ‘95 conferences and co-organized the Information Visualization Symposia (1995-98).