Guest post from T. Alan Keahey, PhD
Dr. T. Alan Keahey has played a leading role in the research and development of highly innovative information visualization systems for close to 20 years. His experience spans a wide range of environments including national labs research scientist, research director at a Lucent Bell Labs spin off and founder of his own visualization R&D company. He thrives on anchoring connections between the capabilities created in research environments and the real-world needs of business customers. Alan is currently a Visualization Science and Systems Expert at the IBM Business Analytics Office of the CTO.
Information Visualization first became a defined field of study in the early '90s. The goal of the practitioners was to create effective mappings between abstract, non-physical data and visual properties. The subsequent research and development has generated many novel ways of viewing and interacting with data, yet for many mainstream users visualization is still thought of as just being a collection of traditional pie, line and bar charts, perhaps with some "exotic" scatter plots or Gantt charts on the high end.
In the latter, more traditional, view of visualization, the chart or visualization sits neatly inside a rectangle in the overall display, with other controls and information arrayed outside of the rectangle. I call this arrangement Visualization in a Box, and it has been by far the most common framework for embedding visualizations within applications over the last few decades.
A schematic of a "Visualization in a Box" in which all of the interaction controls are located externally to the actual visualization.
However, this Visualization in a Box paradigm is severely limiting in terms of how we allow the viewer to explore and interact with the information that is being presented. A recurring theme in my research and development over the years has been the push to get more of the interaction inside the box, thereby allowing the viewer to manipulate the information directly rather than through external controls. Many fellow researchers over the years have also been working on these issues, and several interaction paradigms have arisen as a result of this work.
Visualization with Basic Interaction and Drill Down
Basic direct interactivity includes items like tool tips that pop up when the user selects a single graphic element within the visualization to show the specific value or other related information associated with the element. Tool tips are great aids for providing quick additional detail; however, due to their one-at-a-time nature, they cannot be used to provide broad information .
Drill down is another type of basic selection and interaction that is best illustrated by example: clicking on a bar chart element from a chart of stated values might cause the chart to redraw itself, showing, for example, the values for major cities in the selected state. Drill downs are a powerful navigational metaphor, but their use needs to be limited, as sequences of drilling down and backing up can quickly cause an overwhelming loss of navigational context. Drill down is also frequently implemented via external multi-layer tabs similar to the one shown below.
Above, a generic example of a hierarchical tabbed navigation.
Visualization with Enhanced Brushing/Selection
Brushing is conceptually similar to tool tips in that it allows the viewer to directly select one or more data points and get quick additional details on those points. Typically, but not always, this is done via a rectangular selection tool. This selection might be used to drive the display of summary statistics for the selected items, or to highlight the related items in another synchronized visualization elsewhere on the screen. This latter usage provides a powerful tool for navigating highly dimensional data sets that cannot be shown directly in one view.
Maps are one of the oldest forms of visualization, and since ancient times explorers, generals and tourists have used them interactively, whether folding them to focus on a particular region or scribbling on them to trace a route. With modern mapping and mobile technologies this interaction has become even richer. We now pan and zoom through our maps directly with pinches and swipes of our fingers, selecting routes, restaurants and sharing locations with others. Interaction with mobile maps is now happening almost exclusively inside the box. In my view, we will be seeing even more of that type of interaction with other information visualization tasks in the future.
Visualization as Navigation
The interaction techniques discussed thus far have been used for many years, and are often applied to fairly standard charts such as bar charts and scatter plots. However, there are many much more sophisticated visualizations available that present richer opportunities for navigating and interacting with complex date sets.
Consider our earlier example of a set of bar charts showing information at several different levels such as state and city level, or organizational division, department and group level. Drill down provides a reasonable first step for allowing the viewer to make sense of the information; however it creates a navigational hierarchy that prevents the formation of a gestalt understanding of the data due to the required context switching between views and the limited capacity of human short term memory.
Using advanced visual metaphors, however, we can show the entire navigational hierarchy and thus allow the viewer to see the big picture and smaller details simultaneously. There are many visual metaphors that can do this: the tree map is a standard example; other tree diagrams are also possible.
A tree diagram can be used to show the same information found in a hierarchy of bar charts. In the example at left, the size of the bubble could be proportional to the height of the bars, and the color used to show some additional measure of the data.
This new approach allows us to substitute a single visual metaphor for extensive navigation -- which is a great Illustration of the theme of getting interaction and navigation inside the box directly with the visualization. Once we have the entire navigational hierarchy in that single visualization, we can also start to interact with the entire hierarchy of data via selection, brushing, filtering etc. Those operations would be much more difficult, if not impossible, with the old way of drilling down between bar charts.
Interacting Inside the Box
This example is just one illustration. There are a wide range of scenarios in which richer navigation and interaction are afforded by rich visual metaphors that capture not only the data itself, but also the relations between the data and the ways in which the viewer will want to interact with it. Once you get used to thinking about visualization in this interaction "inside the box" manner, it seems terribly old fashioned to go back to the simple visualization with external controls.
By analogy, imagine a mobile maps app that does not let you pinch and swipe to navigate, but instead requires you to navigate via a set of external buttons that force a smaller map display area and constant focus shifting between the map and the controls in order to navigate. The modern mobile user would not consider the latter option to be an acceptable alternative these days. In my opinion, visualization consumers over the coming few years will develop a similar set of expectations for having an ability to directly navigate and interact with their visual information displays inside the box.
Try out new visualization techniques
Create your own visualizations in three easy steps and learn more about visualization trends and best practices from IBM visualization luminaries on the IBM Many Eyes Community.
Visit the Analytics Zone to download new visualizations into your IBM Cognos Active Reports. With the extensible visualization capabilities in IBM Cognos Business Intelligence V10.2.1, new visualizations are only a download away.