Position paper for the CHI 98 workshop Hyped-Media to Hyper-Media

Towards a Theoretical Space for Hypermedia Systems
Kinds of Systems, Kinds of Users (and a Look at Design)
Nils Dahlbäck
SICS, Box 1263, S-164 28 Kista, Sweden
and
Department of Computer and Information Science
Linköping University, S-581 83 Linköping, Sweden
 
 

The Context: The PERSONA project

The PERSONA project is a two year European project working on PERsonal and SOcial aspects of NAvigation (in hypermedia and other electronic spaces). The partners in the project are SICS and Napier University.

PERSONA is an acronym for PERsonal and SOcial NAvigation. The name of the project illustrates its two-fold approach. We are both working within the current mainstream approach to the problem area, especially by studying the individual differences, social, cultural, etc., in navigation ability, and based on this developing guidelines for system design. What characterizes our work here, is our emphasis on adapting the system to the users' abilities and needs, including cultural, cognitive and other aspects, in other words adapting to the individual person's needs. The adaptations can be both dynamic, as in adaptive interfaces, and static, i.e. designing systems or interfaces for clearly defined user groups, or making it possible in for the user to adapt the system to her needs.

At the same time we are trying to develop alternative approaches to system design, breaking away from the lonely 'walker in the woods' picture of the information system user, to a social being in constant interaction with other fellow beings through all steps in the tasks and activities pursued.

In our design work, we are seeking inspiration also from less traditional knowledge sources for interface and system design, e.g. cinema studies. (We have recently compiled a large literature review on a number of areas of relevance for this project. It can be found here.

There are a number of issues mentioned in the workshop call for papers that relate closely to our work in the PERSONA project, and which I would enjoy discussing with fellow workers during the workshop, should the possibility arise. Here I will only give a short description of my work on some of these areas. Should there be an interest in it, I could expand any of these into longer presentations for the workshop.

Dimensions in the hypermedia space 
(Evaluation of hypermedia systems.)

It is my belief that one important pre-requisite for evaluating hypermedia systems, or rather for making it possible to transfer results obtained in evaluation studies to other systems and domains, is a set of descriptive dimensions of different hypermedia systems. Without such a taxonomy, it is difficult to know to which extent the system I am currently working on is similar enough in relevant aspects to make it possible for me to make use of evaluation results on another system.

It is of course trivially true that we have different kinds of hypermedia systems. We all know about hypertext, hypermedia, Virtual Reality, Immersive Virtual Reality, MUD:s, etc. But this is basically a technological classification, that does not consider the contents of the information space, nor the tasks performed in it. Leaving the task aspect aside here, it seems clear that there are different levels of structure in most systems of this kind, and this is especially true for hypertext and hypermedia systems. I will here mention some of the dimensions that I currently find it important to consider.

When developing designs or evaluating or comparing existing designs, the kind of mapping between the structure of the information presented and the structure of the information presentation need to be considered. I currently find it useful to distinguish between three kinds of hypermedia information content-structure mappings
 

It should be noted that for the last category there might exist consistent conceptual structures for sub-groups of users. And many competing such structures or world views can co-exist and even compete at the same time. In fact, the difference between the second and third category is very relative to a particular cultural perspective. All people that have received traditional Western schooling would probably agree upon at least the crude outlines of a biological classificatory system, but the Kaluli people on Papua New Guinea would probably not share this view. So in some sense we have not three but two distinct types. But the distinction between the latter two has an heuristic value when designing hypermedia systems for specific purposes, in making the designer forced to consider whether there exist a structure common to all the intended users, or if many different such structures need to be catered for in the design.

Another interesting classification is the one by Dourish and Chalmers' (1994) of three major modes of navigation, namely spatial, social, and semantic. Leaving the social navigation aside here, spatial organization, the prime example being immersive VR systems, is obviously closely related to our ability to navigate in geographic space. The class called 'semantic' (CD-ROMs, help systems, etc.) has a structure organized by semantic connections.

The distinction is obviously important, but I suspect that further clarification of what characterizes the different classes might be needed here. It has been shown that hypermedia, data bases, and hierarchical file systems are of a spatial character (Dahlbäck et al, 1996, Benyon & Murray, 1993, Vicente & Villeges, 1988). But these systems are not spatial in the same sense as an immersive VR system. They seem rather to belong to the non-spatial or non-geographical system in the tripartite classification presented above. They have, in a sense, a spatial structure but not spatial content. The concept of time seems to be a prime example of this (e.g. Clark,1973). There are a number of different conceptual structures of time, in different cultures or for different purposes (linear, circular, and sub-types of these), but they all seem to have in common the use of a spatial structure to conceptualize something inherently non-spatial.

Some observations by Dahlbck et al (1996) in fact suggest that different human spatial abilities are correlated with navigation in geographic space and navigation in a large help system, but more work is clearly needed here. It is important in further work to clarify whether these two kinds of spatiality share enough properties to make similar solutions work in both cases, or whether they should be treated differently, despite both being of spatial nature. This leads over to my next topic, on individual differences.

But before leaving the topic of descriptive dimensions, let me just say a few words on the the other major category mentioned above, i.e. the different kinds of user tasks that might influence the design, and which, in my opinion, need to be considered when evaluating hypermedia systems. Space does not permit any longer discussion on this, so I will here just list a few of those that seem important to us.(The literature review from the PERSONA project gives more details on this. A link to the review will be incluced here as soon as we have it on the web.)

Hypermedia tasks

Individual differences and hypermedia navigation
(Cognitive models of interacting with hypermedia)

Cognitive psychology and cognitive science have never taken any strong interest in issues of individual differences in cognitive abilities. It is my belief that applied cognitive psychology to a much too large extent has inherited this stance, and this with unfortunate consequences. In our group we are currently investigating the impact of cognitive and cultural differences on the ability to navigate in hypermedia systems. In a previous study (Dahlbäck et al 1996) the ratio between the fastest and the slowest subject was 19:1). Furthermore we found that this result correlated .56 with the results on a subset of tests of spatial ability.

There is one interesting, but worrying, aspect of this result. The low spatial users were much worse of using this system. But we also know that low spatial users are less comfortable with using maps. So perhaps designing maps and other navigational aids will not be a benefit to those users that are in most need of help? Perhaps we need to find different remedies (this is where the `social' part of the project comes into play, since what people that are not comfortable with maps do when trying to find the way to their destination, is of course to ask people. so perhaps we can find or develop some equivalent of this also in hypermedia systems?)

Or is the 'navigation' in geographic and electronic space perhaps not the same thing, or at least not dependent on the same basic cognitive abilities. Some results from this study suggests that this might be the case.

We are currently planning for a series of follow-up studies on these findings. We are currently on the look out for a number of different systems, varying on the dimensions mentioned above, and possibly others. The task dimension is obviously also important here. We will then study users' ability to find navigate and find information in these spaces. We will also assess their cognitive abilities, general wayfinding abilities, and possibly other relevant dimensions. Our aim is to take a step towards a clarification of the relationship between users' individual abilities, strategies etc., and their preferred hypermedia design. One reason for this, is that we believe that only emiprical studies can tell us which of the large set of possible descriptive dimensions for hypermedia systems that are the important ones.
 

Geographic, electronic and cinematic space 
(Design guidelines for hypermedia)

Electronic spaces usually lack one important feature of geographic space, namely the explicit or implicit information that we are progressing in the right direction. This is not only given us by route signs and other artefacts designed for this. When walking in the forest in search for a good place to stay the night, we are presented with an abundance of cues possible for use for monitoring the task; the slope of the hill, the changes in the vegetation, the kind of soil we are walking on, all helps the experiences hiker to find a place for his tent.

Much of present day research is devoted to designing tools that will help the user navigate the space, by e.g. giving her a map of the domain. But it is possible to entertain the hypothesis that we are overusing the `navigation metaphor' here. Is it really these kinds of tools that we need to develop? Designing maps and similar devices to help the user see where she is in the space could be likened to having a person drive a window-less car through the help of a sophisticated map and navigation system, where a simpler and more efficient solution could be to give her a window to look through?

But wait. What we see when we look at (or through?) the windows on our screens today does not give us much help in knowing where we are. So perhaps the `look through the window'- solution is no solution, after all? I believe it is, when we take into consideration the fact that we are not only designing the window, but also the world viewed through it.

Current design guidelines creates systems that are difficult to navigate, since the uniform design of all windows, gives no cues in peripheral vision supporting orientation. We know that in geographic space we use landmarks for orienting ourselves in the space. But current design practice dissuades us from creating hypermedia equivalents of landmarks. All windows should have a uniform design, to make the user easily find the relevant buttons and menus. But this is only OK when it is not important that the user is aware of how the current window is related to other windows. In a hypermedia system this is often not true.

One way of helping the user is to create landmarks and maps. But this is perhaps taking a path that actually is a detour? Cinema has techniques for making the viewer experience many cuts from a scene as watching not a set of isolated screens but different views on the same scene. And this without providing a map in the top right corner telling the viewer how the different camera positions are related to each other. We are currently exploring the use of cinematic techniques in the interface design of hypermedia systems.

The theoretical issue here that I would be interested in hearing other participants opinion on, is to what extent we for hypermedia systems need to develop new principles of interface design to support the users' navigation in space. And where are the theoretical foundations for these design guidelines to be found? My own guess, is that we need to shift perspective on the user. Hitherto, the user has been seen as some kind of a reader of texts. I am not claiming that this is not valuable. On the contrary, I think we need more of this. But if we also think of users as actors in the world, and start looking at what users are doing, not as viewers or readers of something external to them, but as actors in an information space, new areas of psychology and related areas will take on an increased importance.

In this section I have suggested that methods from cinema could be useful for giving the user locational information. But there is a theoretical side of this coin too; what psychological mechanisms, perceptual and cogntive, make these methods work for cinema?

My own guess is that in psychology much can be learned from the so-called ecological school, and in general theories of action. But also architecture becomes important. Designing a large office building so that people can find their way around, and even more important, get out in a hurry in the case of emergency, seems rather similar to designing a large information system. Making this shift in focus when studying hypermedia system probably also entails taking a fresh look at interaction and dialogue. When we are out walking the woods, we are not really in a dialogue with the forest, are we? But is what we are doing when using a hypermedia system then best seen as a dialogue? Is interaction and dialogue always social, and with another agent, to whom we can ascribe intentional states? The basic argument for distinguishing between these is that we as humans carry with us different expectations concerning action and interaction. Most of us would be as surprised if our car started arguing against our turning right, as we would be if one of our friends obeyed our every wish without showing any personal wants or needs. The hypothesis I am entertaining here, is that since we can both act and interact with computers, it becomes important to make clear to the user which of these states the system is at a particular time.
 

References

Benyon, D, and Murray, (1993) Adaptive Systems: From Intelligent Tutoring to Autonomous Agents.  Knowledge Based Systems, 6(3).

Clark, Herbert H. (1973). Space, Time, Semantics, and the Child. In T. Moore (ed.) Cognitive Development and the Acquisition of Language. New York: Academic Press.

Dahlbäck, Nils, Höök, Kristina, and Sjölinder, Marie (1996). Spatial Cognition in the Mind and in the World - The Case of Hypermedia Navigation. In Proceedings from the 18th Annual Conference of the Cognitive Science Society, CogSci'96, San Diego, Ca, 1996, pp 195-200.

Dourish, Paul and Chalmers, Matthew (1994). Running out of Space: Models of Information Support. In Proceedings from HCI'94, Glasgow, August 1994.

Vicente, and Villeges (1988) Visual Momentum as a Means of Accomodating Individual Differences Among Users of a Hierarchical File System. In J. Rasmussen and P. Sunde (Eds.) Proceedings from the 5th Symposium of EFISS, Riso National Laboratory, Denmark, November 1987. New York: Plenum Press.