Another eye-gaze based application we expect will be developed in the future is what we term a cyberputer. The cyberputer is thought of as a media one uses on an everyday basis in one's home. It will be an integration of several current-day appliances (e.g. television, stereo CD-i, telephone), and will provide entertainment along with support for interaction with and manipulation of private or work-related data. It is even very likely that the cyberputer will provide IES films as an integral part of the interface.
We expect the cyberputer interface to be screen-based like current
computers, only the screen will be somewhat larger, perhaps 2m × 1m,
and naturally some unobtrusive eye-tracking device will be a part of it.
This device will be capable of detecting when some user approaches the
screen, and will be able to recognize said user if she is known to the
cyberputer. This information will be used to decide
how the screen is to jump into action when the cyberputer determines that
the user is interested in interaction. Heuristics or perhaps an aritifical neural network (ANN)
will, on the basis of stored information about indiviual, display a "home
position" that is the usual starting point for that individual; or display
the position where the user left off last time; or perhaps some time-of-day
or time-of-week determined position. A position is a spatial
position in a three-dimensional virtual world full of objects representing
users' data, communication channels, games and other recreational objects.
The realization of the three dimensions could be done by using
high-frequency shutter-glasses (cf. page 
) or
perhaps by using techniques that do not require the user to wear special
glasses, but instead sends two different light beams to each of the user's
eyes (cf. Arthur 1995).
The interior of the displayed user data objects are some visually encoded incarnation of the data, constructed in such a way that important abstract relations between aspects of the data are displayed figurally, for example as contact/noncontact, symmetry, size relationships, congruity etc. This way of communicating the data in configural displays will assist the user in grasping the meaning of the data, because
[a]pplying these figural principles as visual means by which system states are experienced should insure that they have a natural relation to human cognition and can be perceived with well established perceptual skills.(Hansen 1992)
The exterior of these objects are geometrical shapes and colours that the user will recognize as indicators of what type of object they are. Perhaps a "family economics" object is square and blue, whereas the favorite entertainment channel is rounded and orange. The net effect is that user will remember the approximate spatial location of the desired objects in the virtual world (humans typically employ a spatially oriented memory scheme (Eysenck & Keane 1990)), and recognize the exact object by its external geometrical attributes.
So far, we have described a stationary virtual world of objects with which the user can interact. The user must of course also be able to concentrate attention to individual objects or get an overview of collections of them; this is accomplished by zooming and 3-D navigation. Zooming is performed when the cyberputer detects that the user is interested in some individual object (cf. the "Little Prince" storyteller application described in section 4.3.1); this object is then "opened up" and occupies most of the display. User interest is detected with the aid of eye-gaze tracking and perhaps gesture tracking; the user will experience this as if the data discloses itself to her when she gazes at it with an interested look. 3-D navigation will be possible by gazing at distant objects, thus in effect "travelling" towards them (this can of course also be thought of as a kind of zooming), but whenever the line of sight to the desired object is blocked, the user will need some means of getting around the obscuring object. This can be done by the user shifting her position slightly, just as she would if she wanted to get around some real-world object-it must be noted that this is a very natural, "direct" way of navigating; the evidence can be seen by observing the body of any video-game player. The resulting interface will faciltitate rapid exploration of the virtual space and fast, "direct" interaction with the objects-with minimal requirements for the user to have syntactical knowledge about the interface itself. And the configurational displays of the data objects will enable the user to grasp much larger bodies of data than today.
The cyberputer will, of course, also react to utterances like "that looks interesting" while looking (or pointing, as in figure 11) at something by zooming in on the appropriate object, and the user can "warp" to entirely different positions by telling the cyberputer what she would like to see.
It is an open question how the Midas Touch Problem
(cf. section 5.4) is best solved; probably some latency
mechanism combined with the spoken utterances will prove to be a suffiently
usable solution. The One-Way Zooming Problem (cf.
page ) is probably best solved by making a
retreating user movement initiate zooming out.
The cyberputer described here is much like "the Media Room" supporting "dataland" described by Bolt (1984, pp. 9-21), but there are differences; we consider some objects as being new virtual spaces, so that zooming in on these objects would "warp" the user to another virtual space, in effect adding a fourth dimension of discrete values. Also, we have done away with two television monitors-but the most important difference lies in the navigating mechanisms: By replacing the joysticks of the Media Room with eye-gaze control, we expect to see a much more "direct," natural way of interacting with data; the close connection between eye-gaze and interest is the foundation on which this fast interface stands.