Dynamics and interactions

Visual Information Seeking

The starfield display proposed by Ahlberg & Shneiderman provides a unique (although not new) method of visualizing data. In a sense, it takes Tufte's small multiples idea to an extreme--it portrays multiple data objects and their relation to one another in a visually compact way that seems somehow more informative than a plain graph (although in a sense it really is just this). The obvious extension of this idea to the domain of crowd visualization is to represent the people in the crowd as stars on the display, with coding (such as color, or perhaps size) to portray supplemental information about the people within the crowd.

Combining this data display with tight coupling of the interface would provide a supplemental way to portray information about the crowd as a whole. Ahlberg & Shneiderman do not deeply go into this aspect of tight coupling in their discussion, but especially for visualizing groups of data, this mode of information display is crucial. The idea is that if a particular option is unavailable because the data do not support it, than it is not presented to the user. However, for a user that knows the capabilities of the software in terms of the options that can be selected--for example showing only persons in the crowd over 40 years old--then the absence of this option tells the user information about the group--in this case that the group represented contains no one over the age of 40.

Toolglass and Magic Lenses

The key to the Toolglass interface is the use of lenses as interactive interface elements. This provides an interesting twist on the use of magic lenses as display devices, allowing the user to modify the behavior of a mouse click based on the lenses between the pointer and the data. This concept was not presented in a clear manner in the paper, but essentially the toolglass magic lenses work bi-directionally, modifying both the viewing and the interaction with the data.

Though the interface presented is different, there exist certain problems with the use of these bi-directional lenses. Most importantly, the lenses require the user to context shift--a lens can function either as a viewing device or a data manipulating device depending on the context of the user interaction, and these two states are mutually exclusive (as far as I was able to discern from the paper). The problem here (since context shifting isn't inherently a difficult tast) is that there is no real distinction made at the interface level to enable the user to tell when this context shift should take place.

A secondary problem is that many of the data manipution lenses seem to make editing data more tedious than current GUIs. When working with text, as in the example given in the paper, manipulating a lens to make a word bold seems more complex than just hitting a key on the keyboard. Additionally it requires synergy between both hands

The interface metaphor is confused by the changing use of the left hand trackball. Again, depending on the context, it functions either as a device to move and manipulate the lens or as a device to move and manipulate the display within the lens. For the same reason that moving menus--menus that when clicked cause the mouse to manipulate the position of menu items on screen to align them with the cursor, as opposed to aligning the cursor to the menu items--are found to be more difficult to use because of the added context switch from manipulating the cursor to manipulating the menu, the context switch involved in using the trackball will likely cause confusion, and ultimately be less efficient than other methods of manipulating the display of data within the lens.

Besides the use of the magic lenses as data display devices (which is discussed below using Pad), I can't see any real use of the data manipulation side of the magic lenses in toolglass to visualize crowd data. (This may of course be due to personal short-sightedness.) The real power of the toolglass architecture seems to be in data creation, such as graphic arts or writing.

Pad/Pad++

There are two key elements to the Pad/Pad++ interface that allow us to visualizw crowds in an intelligent manner: an infinitely addressable (or nearly so) 2D space with semantic zooming and a magic lens interface.

The infinitely adressable space with semantic zooming gives us a medium on which we can make extensive use of Tufte's small multiple concept.



This is a graphic that shows a series of street corners (ala Whyte) and the crowds of people populating them. Each person is represented by a dot on the stylized corner, and the forms of the different crowds can easily be compared.



Zooming in on a particular crowd, this graphic shows the semantic zooming feature of Pad. Now we can see differentiation between the people (which now look more like people): in this case, the saturation represents the speed of the people.

In this interface, magic lenses can be used to either selectively zoom in on only a part of the display, or can reform the data represented.



Here we see the same image as above, but with a magic lens that functions as a history display. Colors represent previous hours population. By using this lens, we can compare the current crowd with past crowds.

page last updated 3/27/2001