We contrast the traditional, cognitivist approach to interface design, with an alternative we term experiential interface design. The traditional approach sees interface metaphors (and other components of the system image) as a medium of communication between the designer and the users; the aim is for the user to develop a mental model that matches that of the designer as closely as possible. In contrast, experiential interface design sees the designer as a creator of possible user experiences. The approach is based on the linguistic-philosophical theory of Lakoff and Johnson, which sees meaning as metaphorical projection of bodily experiences. Finally, we present SchemaSpace, a virtual environment for information visualisation and exploration, in order to illustrate an experientialist interface.

Keywords: User interface design, metaphor, experientialism, theory of meaning

1 Traditional Interface Design

The problem of interface design has often been characterised as one of communication between the designer and the users. Norman's [12] well-known account of HCI design centres on three kinds of model: the design model (in the head of the designer), the user's model (in the head of the user) and the system image (as presented in the designed interface). The system image serves as the medium of communication between the designer and the user. In the ideal case, the user's model comes to match the design model closely. The common approach to facilitating this process has been to incorporate one or more metaphors in the system image. It then becomes of great importance that the designer chooses appropriate metaphors which convey relevant aspects of the functionality of the system in terms that are understandable to the user [5]. A good metaphor is supposed to permit the user to apply knowledge of the source domain of the metaphor to the unfamiliar target domain of the interface [6].

The traditional approach to interface design, briefly outlined above, views metaphor as a tool of communication between the designer and the user. Much debate has surrounded the usefulness, or otherwise, of interface metaphors (see [14], Ch. 5). Of course, the interface isn't exactly like a desktop, but sharing the idea that it is somewhat like a desktop is intended to bring the user's and the designer's conceptions of the interface closer together.

The rationale for traditional HCI design depends on some version of an encode-decode model of communication. There is knowledge in the designer's head which he communicates to the user, through the medium of the system image, often by using one or more metaphors. Similarly, an encode-decode model of language assumes that the speaker or writer has knowledge in his head which he communicates to the hearer or reader through the medium of speech or text by using sentences. This model of communication is unsatisfactory for a variety of reasons (see [13]). Most importantly for this discussion, it assumes that meaning is objective and can be captured in a fixed correspondence between aspects of the world and some system of representation. Sperber and Wilson [13] present an alternative view of communication which, while avoiding the relativistic problems of a totally subjective account, stresses that communication depends on a consideration of what is relevant in a particular situation, not just on what is passed along the "channel of communication". It also depends on "ostension", on acts that produce particular experiences in the audience by virtue of being performed in the current situation.

2 The Experiential Alternative: Meaning as Metaphor Reflecting Embodiment

The traditional approaches to both HCI design and understanding communication in general suggest that metaphor is some kind of specialised device for conveying a complex of concepts, based on speaking of, or presenting, one thing as if it were another. In several books published over the last two decades, George Lakoff and Mark Johnson have presented an alternative view of meaning, one that casts a completely different light on the role and importance of metaphor, while at the same time avoiding the problems of both objectivism and pure subjectivism [7, 8, 9, 10].

According to Lakoff and Johnson, metaphor is much more than a linguistic and rhetorical device. They argue that we always think metaphorically, that our everyday experiences are shaped by three kinds of metaphor: structural, orientational and ontological. Structural metaphors are found when one concept is structured in terms of another, for example that argument is war: "Your claims are indefensible", "He attacked every weak point in my argument", "His criticism was right on target", "He shot down all of my arguments", etc.. We not only speak of argument as if it were war (and very pervasively, so that many statements about argument reflect this underlying structuring - although we don't actually think of them as metaphorical), we think about argument as if it were war, and often act according to the same, unconscious, assumption.

Orientational metaphors structure experience in terms of spatial orientation. For example, down is negative, up is positive: "I am depressed", "I am really down", "I feel low", "Things are looking up", etc.. Ontological metaphors structure our experiences of abstract phenomena in terms of concrete objects and forces (see also [15]). Efforts to visualise information as shapes, colours, and textures can be seen as reflecting the operation of underlying ontological metaphors, as, arguably, can the application of general interface metaphors such as direct manipulation.

If every concept is metaphorically structured, are we not stuck in some kind of infinite regress? If every concept is structured in terms of another, we are indeed (which is, of course, a fundamental problem with objectivist accounts of meaning). Lakoff and Johnson avoid this infinite regression by suggesting that, at bottom, meaning is rooted in basic, bodily, experiences of life as animals with a certain physical configuration residing on a planet with certain characteristics (notably, gravity). When we use expressions like, "I fell asleep" or "Wake up!" we use metaphor in a way that reflects the physical nature of life on earth. Our body configuration, combined with gravity, makes it necessary for us to sleep in a horizontal position.

Johnson [7] provides more detail on the grounding of the (fundamentally metaphorical) conceptual system in corporeal, earthly existence. He proposes the existence of image schemata, which are basic structures of experience. These structures are then projected metaphorically onto more complex experiences. Lakoff [9, pp. 271-278] suggests that image schemata i) are based on bodily experience, (ii) have structural elements, (iii) have a basic logic and (iv) are manifested in actions and expressions. He gives many examples of image schemata, including the container, the centre-periphery, and the verticality schema.

3 The Experiential Approach to HCI Design

An approach to HCI design based on this experientialist account of meaning, rather than the usual objectivist cognitivism of the traditional "mental model" approach, is based on the fundamental premise that to design HCI is to design the conditions for possible users' experiences.

In the traditional approach, the metaphor is part of the interface. This need not be the case with experientialism since, by this account, metaphor is everywhere. Taking an experientialist view of interface design suggests that a meaningful interface is one that is experienced in a way that supports the metaphoric projection of image schemata. This is done by the user in the same way that he makes sense of all the other experiences of his daily life, by unconscious projection of bodily image schemata. If the experientialist designer is primarily a creator of user experiences, the traditional interface designer is primarily a communicator of mental models, using metaphor as a useful device.

While we are not arguing that all traditional interface metaphors should be replaced, we do suggest that for several application areas - and these are areas that are at the forefront of current HCI research and development - an experiential approach to HCI design may be more appropriate. A notable example is that of information visualisation and exploration.

Conklin [2] argues that "there is no natural topology for an information space". However, an experientialist designer would argue the opposite; that there are, in fact, not one but many natural topologies for such a space, topologies ultimately grounded in human bodily experiences, and projected as image schemata. Waterworth [16] outlines a design for a World Wide Web browsing environment - a Personal Space - that was informally based around considerations of human bodily experiences in real, physical spaces.

In adopting the experiential approach, a valuable source of design insights is that of language. How do users talk about their experiences? Utterances can be gathered at two stages of the design process: user requirements analysis early on and, later, as corroboration that a particular design is producing the kind of experiences the designer intended.

The traditional approach to HCI design uses metaphor to communicate the functionality of the system to the user. The designer draws on users' experiences in another domain to assist their understanding of the system. As Erickson [5] has pointed out, this implies that designers know what the system really is. Despite its problems this approach has been successful in encouraging the widespread use of computers, at least for certain classes of application.

The experientialist approach to design also draws on usersÕ prior experiences, but there are several fundamental differences. Firstly, from the traditional perspective, metaphors are useful (usually) but not essential. A traditional user interface metaphor can always be paraphrased into a literal interface. From the experientialist perspective, however, metaphoric projection is essential to the way people make sense of the world, including a user interface. Secondly, that metaphoric projection is essential to sense-making does not mean that we live in a world of metaphors. If we design from an experiential perspective, this does not mean that the interface need be a virtual world of metaphoric objects. Such a world is more likely to be the outcome of the traditional approach.

From the experientialist view, what is needed in HCI design is for the interface to be a source of experiences, designed in such a way that the experiences generated may be structured by the projection of image schemata. What the resultant interface means, what it is for a given user, depends on his unconscious reactions to the structures provided. If the interface feels right for its purpose, it is successful. No designer can know what the system really is, in general. It is what it means to individual users and, like life, it means what it is experienced to be.

Even though we consider the experientialist approach to user interface design as described in this paper to be a new approach, the experientialist theory of meaning has already attracted in fields related to user interface design. For instance, Clay and Wilhelms [1] present a linguistic interface for placement of 3D objects which focuses heavily on spatial relationships as discussed by Lakoff and Johnson. There has also been criticism directed towards the experientialist view of metaphor. Although he recognises some merits of experientialism, Coyne [3] claims that Lakoff and Johnson put too strong an emphasis on the primacy of bodily experience and that there are non-embodied and non-spatial uses of concepts like containment and balance. However, Coyne's criticism seems to illustrate Lakoff and Johnson's main point, that is, that we project our spatial experiences (embodied as image schemata) to abstract, non-spatial domains of experience.

4 SchemaSpace: an experientialist environment

We are currently engaged in a more thoroughgoing attempt at the experiential design of an environment called SchemaSpace [11]. The approach is a development of the idea that HCI design is mostly a matter of sensual or perceptual ergonomics rather than the 'cognitive ergonomics' that follow from the traditional, cognitivist approach (see [15] and [17], for more detailed discussion of these two perspectives).

SchemaSpace is a three-dimensional virtual environment in which a potential user may organise and browse a collection of references to information sources, located on the Web or elsewhere. As such SchemaSpace is a personal information space (see [16] for an elaboration on personal information spaces).

How should an information space like this be designed? We claim that an answer to this question - from a traditional point of view - would in part be formulated in terms of functionality and ways to convey that functionality to the user through the system image. If we instead try to answer the question from an experientialist point of view we first have to reformulate the question: what kind of experiences does the user want to experience from the interface? By posing the question this way we put emphasis both on the designerÕs role as a designer of meaningful experiences and on the role of the user interface as a source of meaningful experiences.

Our intention when designing SchemaSpace has been to design the interface in such a way that it allows the user to experience four different kinds of experiences that each informs the user about different qualitative aspects of the information space:

Obviously, these qualitative aspects are by no means all encompassing and we can see a whole range of other aspects that a user might want to experience from a personal information space. However, our purpose is not primarily to design an ultimate application, but rather do illustrate what we understand to be features of practical experientialist design.

As already mentioned, a meaningful experience is an experience which allows for structuring by means of metaphoric projection of image schemata. Thus, one important step in the design process is to identify image schemata that are associated with the qualitative aspects of the information space we want the user to experience. This identification is by no means arbitrary, on the contrary it ought to be informed by empirical inquiries like the ones conducted by Lakoff and Johnson [7, 8, 9, 10].

4.1 Distinctiveness through containment

In the particular instance of SchemaSpace we describe here we have about three hundred different Web-references to information on very disparate subjects, ranging from modern literature, via architecture, to computer graphics. Such a large collection calls for some kind of categorisation, a way to organise and order the information in sub-collections consisting of references belonging to the same category. Put differently, we have to provide for the possibilities of experiencing distinctiveness, that is, an experience which informs the user that some information references in some respect are different from other references. In order to provide such an experience we have to identify an image schema which is involved in our general understanding of ordering objects and activities in our everyday life.

According to Johnson [7, p. 21] human experiences can in many cases be described as experiences of containment:

Our encounter with containment and boundedness is one of the most pervasive features of our bodily experience. We are intimately aware of our bodies as three-dimensional containers into which we put certain things (food, water, air) and out of which other things emerge (food and water wastes, air, blood, etc.)

Not only are we containers ourselves, but our everyday activities in general - and ordering activities specifically - often involve containment in some respect: we live in containers (houses, shelters, etc.), we organise objects by putting them in different containers. Our frequent bodily experiences of physical boundedness constitute an experiential basis for a container schema ([9], p. 272 and [7], p. 21-23), a schema which structures some of our experiences by means of metaphorical projection.

A plausible way of providing for the experience of distinctiveness is to present the information references that belong together in a way that allows for a projection of a container schema. There are countless ways of visualising containment and folders and rooms are probably the most familiar user interface containers. However, in our design of SchemaSpace we have as much as possible avoided elements which are - like folders and rooms - heavily metaphorically laden, in order to stress the experientialist features of SchemaSpace (although it is our strong belief that experientialist design need not by necessity exclude "ordinary" user interface metaphors). Instead, the elements of SchemaSpace consist largely of simple geometric shapes which are not closely associated with a specific source domain. We have chosen to visualise containment by means of semi-transparent cones (see figure 1). A cone contains information references visualised by stacks of slices, each with a descriptive textual label. By using semi-transparency it is possible to see that a cone actually contains information references, at the same time as it is apparent that they are bounded by the cone and by virtue of being bounded they are distinct from other references.

Figure 1

4.2 Quantity through verticality

Each cone contains a sub-collection of the totality of information references in SchemaSpace. Some of the sub-collections will contain more or less references in comparison to other sub-collections. Even though the cones are semi-transparent, viewed from a distance in the three-dimensional environment it will be difficult to judge the quantity of each cone. In order to provide for a meaningful experience of the quantity of each cone's contents we have to identify an image schema which is involved in our general understanding of quantity.

According to Johnson our basic experiences of quantity are closely associated with verticality [7, p. 277]:

Whenever we add more of a substance - say, water to a glass - the level goes up. When we add more objects to a pile, the level rises. Remove objects from the pile or water from the glass, and the level goes down.

Spatial experiences of the this kind constitute an experiential basis for a verticality schema, a schema which by means of metaphoric projection plays an important role in our understanding of non-spatial quantity. Our tendency to conceptualise quantity in terms of verticality reveals itself in everyday language used to talk about quantity [7, p. 121]:

The crime rate kept rising. The number of books published goes up and up each year. The stock has fallen again. YouÕll get a higher interest rate with them. [...]

In our design we have tried to exploit this verticality aspect of quantity. As seen in figure 1 the cones in SchemaSpace vary in height. The larger cones have a larger quantity of references inside compared to the shorter cones. Our intention have been to combine the container and the verticality schema in order not only to express quantity, but also to strengthen the experience of cones as containers of information references.

4.3 Degree of relevance through centrality

As already mentioned, one of our goals has been to provide for the experience of distinctiveness. Even if a sub-collection constitutes a unity by virtue of belonging to the same category or subject, different references within a sub-collection may be of different importance or relevance in relation to that particular subject.

As pointed out by Johnson [7] "our world radiates out from our bodies as perceptual centers from which we see, hear, touch, taste and smell our world". We also have very basic spatial and physical experiences of centrality as a measure of importance and relevance. Not only is that which is near the centre (the body) within our perceptual reach, but we also experience our bodies as having a centre and periphery where the central parts (trunk, heart, etc.) are of greatest importance to our well-being and identity [9].

In order for a potential user of SchemaSpace to experience some references as more important and relevant in relation to other references within a cone, we exploit a center-periphery schema, which has its experiential grounding in perceptual experiences of centrality mentioned above [7, 9].

Figure 2

As seen in figure 2, stacks of information references are organised along an arc. In those cases where there are a lot of references within a cone, the arc will eventually be closed and form a circle centred around the vertical axis of the cone. Information references can, however, be placed at varying distances from the centre, that is, some references will perceptually be closer to the centre and some will be more peripheral (see stack to the right in figure 2). Our goal with this arrangement is to invoke a metaphoric projection - on the part of the user - of the center-periphery schema in order to experience those references which are perceptually central as conceptually central.

4.4 Connectedness through linkage

Finally, we want the user to experience that some sub-collections of references are related to each other, even though they are distinct from each other. In SchemaSpace we have a collection of references on the subject Virtual Reality Modelling Language (VRML). But we also have two categories with references to information on VRML-browsers and VRML-worlds. These two categories may be considered as distinct from VRML information in general, but not in the same sense as information on the writer Paul Auster is distinct from information on architectural magazines. There is a connection between general information on VRML and VRML-worlds and browsers, that does not exist in any obvious way between Paul Auster and architecture. In order to provide for an experience of this kind of connectedness we exploit a link schema. The link schema is often involved in our understanding of relations and connections of different kinds, not only physical connections, but also more abstract, non-physical connections like interpersonal relationships [7, 9].

Figure 3

In SchemaSpace cones are connected with a perceptual link if the sub-collections contained in the cone are considered to be connected, as in the case with VRML in general and VRML-browsers and worlds (see Figure 3). As with the center-periphery example above, our goal with this arrangement is to provide the user with perceptual cues that allow for structuring by means of projection of a certain schema, in this case the link schema.

5 Concluding remarks

In the preceding section we presented SchemaSpace to illustrate an example of what experiential design could be in practice. Obviously, the illustration has its shortcomings. For instance, how do we know that a potential user actually understands the interface by means of metaphoric projection of the schemata we exploit in SchemaSpace? The answer is that we do not and this stresses the empirical nature of experientialist design. In order to evaluate the success or failure of an interface designed from an experientialist point of view language analysis is crucial. How do users talk about their interfaces? What expressions do they choose to describe their actions? Only by answering questions like these can we judge whether the interface is reflecting embodied schemata or not. This is also the next step for us to take.

Although the main guiding concept of the experientialist approach is that of meaningful experiences, there are reasons to believe that the same approach may be used for the opposite purpose. OSMOSE - the brainchild of Canadian artist Char Davies - is an immersive virtual environment which exhibits some truly experiential features. Instead of providing meaningful experiences, DaviesÕ intention seems to be the opposite [4]:

Rather than representing the world as we usually perceive it, i.e. as empty space containing solid, static, hard-edged and separate objects rendered in photo-realistic style, our approach was to de-solidify things, dissolving distinctions between them, creating a slippage of boundaries or osmosis of inside and out, figure and ground, self and world.

It is striking how DaviesÕ approach violates what the experientialist approach suggests. Still it is very experientialist in nature; the fundamental difference is to be found in the intentions of the two approaches. Although Davies never mentions image schemata, she recognizes similar recurrent structures of experience (boundaries, inside/outside, etc) as an important aspect of sensemaking and by delibaretly designing the interface in a way that is not in accordance with those structures, she provides for the experience of ambiguity and paradox.

Even though SchemaSpace to a great extent is an information visualisation and exploration interface, we do not necessarily exclude more ordinary and traditional applications from experientialist design. However, applications like SchemaSpace are not merely tools, but virtual environments. We suggest that a key difference between more tool-like applications (word processors, spreadsheets, etc.) and virtual environments is that users will need more than a functional understanding to interact with virtual environments. Thus, design of virtual environments involves more than conveyance of functionality. Just as we need an understanding of our everyday physical environment that goes beyond that of merely knowing how to "operate" it, virtual environments ought to be designed in ways that make sense to those who inhabit them. We believe that the experientialist approach may contribute to design of virtual environments which provide meaningful experiences for their inhabitants.

6 References

1. Clay, S R. and Wilhelms, J. (1996). Put: Language-Based Interactive Manipulation of Objects. In IEEE Computer Graphics and Applications, Vol. 16, No. 3, May 1996.

2. Conklin, J. (1987). Hypertext: An Introduction and Survey. IEEE Computer 20 (9), 17-41.

3. Coyne, R. (1995). Designing information technology in the postmodern age: From method to metaphor. Cambridge MA,:MIT Press.

4. Davies, C. (1997). Changing Space: VR as a Philosophical Arena of Being. In Proceedings of 1st Internationa CAiiA Research Conference: Conciousness ReframedÕ97. Centre for Advanced Inquiry in the Interactive Arts, University of Wales College, Newport, UK.

5. Erickson, T. D. (1990). Working with Interface Metaphors. In Laurel, B (ed.), The Art of HCI Design. Menlo Park, USA: Addison-Wesley.

6. Gentner, D., Falkenhainer, B. and Skorstad, J. (1988). Viewing metaphor as analogy. In Analogical Reasoning: Perspectives of Artificial Intelligence, Cognitive Science and Philosophy. D. H. Helma (ed.), Kluwer.

7. Johnson, M. (1987). The Body in the Mind. Chicago: Chicago University Press.

8. Johnson, M. (1993). Moral Imagination. Chicago: Chicago University Press.

9. Lakoff, G. (1987). Women, Fire and Dangerous Things. Chicago: Chicago University Press.

10. Lakoff, G. and Johnson, M. (1980). Metaphors We Live By. Chicago: Chicago University Press.

11. Lund, A (1997). Embodied Interfaces. In K. Braa and E. Monteiro (eds.): Proceedings of IRIS 20: "Social Informatics". Oslo: Department of Informatics, University of Oslo.

12. Norman, D. (1986). Cognitive Engineering. In Norman, D and Draper, S., User Centered System Design. Hillsdale: Lawrence Erlbaum Associates.

13. Sperber, D. and Wilson, D. (1995). Relevance, 2nd edition. Oxford, UK: Blackwell.

14. Waterworth, J. A. (1992). Multimedia Interaction: human factors aspects. Chichester, UK: Ellis Horwood (Simon and Schuster International).

15. Waterworth, J. A. (1996). Virtual Reality for Animals. Proceedings of Ciber@RT'96, First International Conference on Virtual Reality. Valencia, Spain, November 1996.

16. Waterworth, J A (1997a). Personal Spaces: 3D Spatial Worlds for Information Exploration, Organisation and Communication'. In R. Earnshaw and J. Vince (eds.): The Internet in 3D. New York: Academic Press.

17. Waterworth, J A (1997b). Creativity and Sensation: The Case for Synaesthetic Media. Leonardo, 30, (4), 327-330.