This report is also available as an Acrobat file.

Contents
The Design of Virtual Environments with particular reference to VRML

Virtuality

[The body] works in Euclidean space, but it only works there. It sees in a projective space; it touches, caresses and feels in a topological space; it suffers in another; hears and communicates in a third; and so forth...
Serres 1977

To an extent we can consider spatiality as something merely perceived, without any action by the user, but the same is not true of virtuality, which is essentially about the relationship between the user and the system.

Virtuality builds on the idea of Direct Manipulation popularised by Shneiderman (1992), discussed below. It aims to produce forms of human computer interaction which give users the feeling that they are engaging with data rather than with tools to manipulate data. Hutchins et al (1986) describe 'the feeling of involvement directly with the world of objects rather than of communicating with an intermediary.' Instead of turning on a light by touching a light-switch, we simply touch the light itself. Virtuality is opposed to the idea of the Separable Interface (Edmonds 1992) which is applied as a control surface after the functionality has been decided on.

Nelson (1990) gives as an example of effective virtuality a terrestrial globe: 'A globe does not say "Good Morning"; it does not bother you with menus, icons or prompts. You turn it and move your head to the most useful position for overview or detail, that's all.' A virtual globe is indeed an interesting case for any discussion of VR. It has the great advantage of eliminating one kind of representation - that of mapping the world onto a 2D surface with all the unhelpful distortion that implies. We might imagine a time when students never again see a 2D mapping of the world. Nevertheless it is obvious from what we know of maps that many issues raised by 2D map-making are NOT eliminated by making a 3D virtual globe. For example when the user first encounters the globe, which country is centred in their view? Which way is up? - there is no scientific reason to put North at the top of a globe. We must presumably have the freedom to change from physical to political and other models. And as with all direct translations of physical dimensions into computer dimensions, a virtual globe of course privileges a particular form of measuring, whereas 2D maps have been able to exploit the freedom of their greater separation from actuality by, for example, using the depicted size of countries to represent their GDP, rather than their physical size. So Nelson s example of virtuality is not quite as persuasive as he makes out.

Direct manipulation and virtuality

The classical definition of Direct manipulation requires:

continuous representation of the objects and action of interest
physical actions or presses of labelled buttons instead of complex syntax
rapid incremental reversible operations whose effect on the object of interest is immediately visible.

Shneiderman's thoughts on Direct Manipulation (1992 p204) are germane to any discussion of Virtual Reality. In particular he lists some difficulties with visual and spatial representations compared with text:

graphical representations may be more extravagant in their demand on screen-space than the corresponding text [and of course on computing power]
the low density of information in pictorial layouts may inhibit performance of tasks compared with, say well designed textual or numeric tables
if visuals are used iconically, they must be learned [this is another factor in the debate between mimicry in VR and the construction of semi- or wholly abstract 'information landscapes']
the visual representation may be misleading, causing users to predict behaviours in the system which it does not in fact permit
choosing the right objects and actions is not an easy task

While Shneiderman focuses on relatively simple graphical interfaces, these points serve yet again to emphasise that the construction of Virtual Environments is an intentional, design-based activity which must take into account many factors, not a mechanical process of developing electronic counterparts of slices of the world.

However, Shneiderman does not mention what is perhaps one of the most problematic aspects of both Direct Manipulation interfaces and VR, namely that in abjuring abstract, symbolic notations we may suffer an intellectual loss as well as some gain. In tying ourselves to a visual, physical paradigm we may limit what we can think. A simple example would be that of a child learning mathematics. First the child does simple arithmetic by manipulating beads; later the child learns that number can be represented by abstract symbols; later still, realisation dawns that mathematics which cannot be done with the physical beads can be done with the symbol system. For example, zero and negative numbers are both concepts which are inaccessible without a symbolic notation, and both are indispensable to anything beyond elementary mathematics.

This is a problem of concretisation which needs to be acknowledged in VR generally, but especially in relation to Education. It could be argued that some of our greatest intellectual achievements arise from the mental manipulation of symbol systems, most especially text and mathematics, and it would be an odd outcome of adopting concretising approaches, via virtual reality, to deny ourselves and our students the ability to think in terms other than the physical. After all, most of our current education system is arguably geared toward encouraging the facility of symbolic thought.

Virtuality and human error

At first sight, it might seem that in constructing virtual worlds we can escape from the problems of human error associated with more abstract computer interfaces. However, no system can be devised which is not susceptible to human error: errors of conception - 'mistakes' and errors of action - 'slips' will both occur (Norman 1988). Nevertheless, the kinds of errors are likely to be different. With command-line interfaces (CLIs) and with graphical user interfaces (GUIs), slips are all too easy for a user who is manipulating objects. For example, when deleting files, the user of a CLI may incorrectly type a command, irretrievably deleting the wrong files; in a GUI, where the user may accidentally drag the wrong objects into the trash, there will normally be some limited possibility of recovery (but only if the trash has not immediately been emptied by the user); in a virtual environment, we might hope that the greater variety of visual cues possible in a Virtual Environment will militate against this sort of slip. Colour, size, texture, transparency, position, orientation, and above all resemblance to some real-world referent (for example objects belonging to individuals might bear their signature, portrait or other characteristic marker), can all invoke our associative skills to correctly identify objects. However, any user of a 3D modeller will be familiar with problems of virtual objects and environments which are not so easily solved, for examples confusions about scale (VRML 1.0 used an arbitrary measurement unit which predisposed designers to give insufficient consideration to the 'actual' size of what they were building), and about whether the model or the user is moving (a problem of desktop VR rather than immersive VR).

Contents

Graphics Multimedia Virtual Environments Visualisation Contents