laser scanner. The scan served as input to a 3D landscape visualization, which was then projected back onto the clay surface. Illuminating Clay illustrates the blurring between input and output device: Users could, for example, alter the flow of a river by molding valleys into the clay. Here, function is also triggered by form, which literally follows the flow of interactions.

Gummi: Flexing Plastic Computers. Another early envisionment of an OUI was Gummi [ 8]—a pressure-sensitive PDA that simulated a flexible credit card displaying an interactive subway map. Bending the display would cause this subway map to zoom in or out, while a touch panel on the back would allow users to scroll. Again, function equals form: the shape of the display affords zooming, and the interplay between haptics and visuals reinforces this functionality. (For more information, see Carsten Schwesig’s article in this section).

Paper Computers. Books and paper also form powerful sources of inspiration for flexible computer design. Paper is particularly versatile as a medium of information. According to Sellen and Harper [ 9], users still prefer paper over current computer displays because it makes navigation more flexible. Paper input is direct, two-handed, and provides a rich synergetic set of haptic and visual cues. Paper supports easier transitions between activities by allowing users to pick up and organize multiple documents two-handedly. Paper is also extremely malleable: it can be folded—a primary source of input when constructing models—or bent, most often applied in navigation. Paper can be randomly arranged, or in stacks, and can even contain other objects. With the development of flexible E-Ink displays we can imagine that in the future our computers will be indistinguishable from a sheet of paper. One of the questions we have been trying to answer is how will we interact with such flexible computers?

We experimented with the use of Foldable Input Devices (FIDs) for GUI manipulations by tracking the shape of several cardboard sheets that featured retroreflective markings (see Figure 4). Behaving like real paper documents, 3D graphics windows follow the shape of associated FIDs. When FIDs are stacked,

Figure 6.

Interactive Blobjects: (a)

Dynacan computer with flash animation;

(b) iPod form factor on an

interactive cardboard design bench;

(c) Zoom gesture on spherical screen with Google Earth.

so are the window sheets in the GUI (Figure 4a). Stacks of window sheets are sorted with a shake and browsed by leafing action (Figure 4b). Using a special FID, window sheets can even be folded into threre-dimensional models, further blurring the distinction between a window sheet and its content (Figure 4c).

Inspired by Wellner’s DigitalDesk [ 11], Paperwindows [ 2] was the first computer made entirely out of three-dimensional sheets of paper (see Figure 5). It simulated a flexible, high-resolution, full-color,

 

and wireless E-Ink display of the future.

Paperwindows are regular sheets of paper, augmented with eight retroreflective markers. These markers allow a Vicon to capture the motion and interactive shape of the paper, which is modeled as a non-uniform rational B-spline (NURBS) surface textured with the real-time content of an application window. When projected back onto the paper, the 3D models correct any projection skew caused by paper folds, giving the illusion the paper is, in fact, an interactive print. We experimented with a Web browser of which most functions were accessible through changing the shape of the paper—the primary display of the computer. Bending the sheet around its horizontal axis would cause the Web browser to page down or up. Bending the document back around its vertical axis would cause the Web browser to go back or forward in its browsing history. Fingers were also tracked: a link was clicked by touching it. A paper window was activated by picking it up. Information could be copied from one document to the next by rubbing two windows onto each other. Documents could be enlarged through collation, and sorted by stacking. Such physical interaction techniques remove any distinction between input and output: in paper windows the shape, location, and