Andreas Dieberger
Dept. for Design and Assessment of Technology
Vienna University of Technology
A-1040 Vienna, Moellwaldplatz 5/187
Tel: (+43-1) 504-11-86 Fax: (+43-1) 504-11-88
andreas.dieberger@acm.org
Jolanda G. Tromp
Faculty of Psychology
University of Amsterdam
Roetersstraat 15, 1018 WB Amsterdam
The Netherlands
tromp@uvapsy.psy.uva.nl
The information city is a spatial metaphor for hypertext browsing in a virtual environment. In hypertextual information, pieces of information are interlinked using hypertext links. One of the problems of hypertext is that users (or readers) can get disoriented after some time. Conklin [CONC87] points out that the ability to organize information in more complex ways gives rise to the problems of not knowing where you are in the information space or not knowing how to access something you believe exists there. Disorientation problems may also occur to some extent in traditional linear text documents, but in a book the required information can only be further forward or further back, whereas hypertext provides more locations in which to store information and more dimensions in which to travel, thus creating a bigger potential for becoming disoriented.
While recent research hints that the disorientation problem, (often coined as the "lost in hyperspace problem"), is a non-issue, there are still problems that users may experience sometimes, such as neglecting to return from a digression or not knowing if there are any other relevant frames in the document or forgetting which sections have been visited or altered. Elm and Woods, cited in [SmWi93], describe disorientation in terms of degradation of user performance rather than subjective feelings of being lost. They define getting lost as users not having a clear conception of the relationships within the system, or knowing their present location in the system relevant to the display structure and finding it difficult to decide where to look next within the system.
In a real city people seldom really get lost, first beause there are always other people they can ask for help and second because a real city is an environment filled with various sorts of information that help in navigating. There are street signs, traffic signs, posters, landmarks, buildings that have a certain meaning for the user. Thus it seems intuitive to make use of the everyday navigation skills we use in real cities to navigate a complex computer generated information landscape.
Using the city metaphor also allows us to add extra cues to the information because we can make use of much other real world knowledge (figure 1). At the ECHT92 Jay Bolter introduced his concept of "writing on the world" to "textualize space" [BOLT92]. Most real world objects give a lot of information to the viewer, whereas objects in virtual worlds seldom do that. Just take a look at the desktop in a graphical user interface (a very simple virtual environment where the user is "present" by virtue of the mouse-pointer [TOGN92]): Most objects look quite the same and can only be distinguished by looking at the file names. It is in fact a step back to the textual user interface. The Information City will be a rich environment by providing much information about the objects in the virtual environment.
A look at a house in real life reveals an astounding amount of information about the house itself and about its contents (see Figure 1). A church looks very different from a factory, an office building, or a house to live in. Business information would probably not be visualized as a church for instance, but most likely as an office building.
Figure 1 - examples of writing on the world on an amsterdam house
Writing on the world can also be used to visualize read wear [HiHo92]: data items grow older from usage and so a house containing older information should look older than a house containing new information. When looking at a bookshelf it is also quite evident which books are new, and which ones are old just by looking at the covers. A torn cover is not only a cue for an old book, it can also signify a book that is so interesting that it has been read very often. Data should wear out. The number of times data has been accessed is often an ubiquitious item of information in real life, but we lose this information in our digital worlds.
The information city is also a software agent in the sense that it groups houses into districts of interest, and helps users to navigate by using changes in light, color and sound as pointers to requested information [DIEB93b] [DiTr93]. By entering a house the user enters the hypertext itself. The principles of restructuring according to similarity, and giving as much information to the user as possible, are applied to the inside of the houses as well as to the outside.
Exploitation of architectural knowledge from city planning shall be used to build an information environment that helps in navigating hypertexts. For instance: Lynch has researched the kind of cognitive maps people make of a city and found many elements (landmarks, paths, nodes, districts, and edges) that seem to be of general application. "[I]f the environment is visibly organized and sharply identified, then the citizen can inform it with his own meanings and connections." [LYNC60]. These elements will be used in the information city.
The information city consists of more than simply hypertext houses. The information city can be used as a front-end for the Internet. It can incorporate the functionality of Gopher, WWW, NetNews, and other similar internet information browsers. It will contain electronic mail, bulletin boards, international meeting rooms where people from physically separate locations can interact virtually at (almost) real-time speed. There are information rooms that represent links to other cities (data-bases at remote internet sites), people can automatically ftp data by requesting a copy of something they saw in the city. There will be a real-estate office where people can request to have their own information linked to the city. There will also be information objects that act similar to information walls but are general objects that people can personalize, carry around, and give to each other. Examples are newspapers or business cards. These are text objects that may also contain hypertext links but they are not statically bound to a location and are not visualized as houses [TROM93b] (see also: [CuNi93] and [WATE92] for very similar visions).
We use building blocks of several types like linear structures, hierarchical structures and catalogs. Those are just a few examples of building blocks we found useful as a start. It is possible to define virtually every construct as a building block. Our concept of building blocks is similar to the template constructs in the intermedia system [CaGa91]. The simplest example of a building block is LIN (linear structure). It consists of a set of linearly linked rooms. The rooms have a linear ordering. In each room there are links to the next and previous room and to the first and last room in the LIN block. Those links are called implicit links since they are provided by the system. When inserting a new room into this construct those links are updated automatically.
A LIN can be visualized as a hallway connecting several rooms. Additional links are located directly on the information wall. The information wall contains the text and pictures that make up the information contents of the room. (figure 2). The look through the door gives a preview onto the next page of information in the LIN. Another possible visualization of a LIN is figure 3. It shows an information room where the user stands in the middle and can turn around. As the walls to the sides are smaller than the wall in the middle this visualization gives a sort of fisheye view on the text.
Figure 2 - a possible visualization of a LINear structure
Similarly a CAT structure is always a sorted catalog where it is possible to go to the next and previous item; a CAT also has a contents node that is updated automatically. It is even possible to have several orderings in a CAT at the same time (alphabetical ordering and ordering according to keyword for instance). The main idea of those building blocks is to provide a means to construct hypertexts that can be visualized as parts of houses.
The information walls in our information room use read wear to give the user a feeling of their relevance. So an information wall containing old information would change its color for instance (see also [LBSS85]). It is possible to stick annotations onto an information wall. They look like post-it notes. The post-its also wear out after some time. We think of making such post-it annotations only temporary. After some time they fall off the wall and then lie on the floor. When a user thinks the annotation is relevant s/he may stick it to the wall again. If that happens several times the annotation is converted to a permanent annotation. If it is not stuck back on the wall it is discarded after some time.
Figure 3 - A LINear structure with "fisheye" view.
A difficult problem in an environment like the information city is enactment of hypertext linking. While traversing an implicit link in a building block is straightforward (just strolling down the hallway), going through a link in an information wall is not. The whole city structure is built in order to give the user a feeling of place and to help him or her to build a cognitive map of the area. While building blocks connect rooms that are directly adjacent, hyperlinks on a information wall may lead to rooms far off in one step. Going through such a link can confuse the user and shatter every feeling of orientation in the city. The key to creating euclidian space which is coherent and therefore navigable, is through giving meaning to position [SHUM90].
One of the big problems of hypertexts is that they are largely non-hierarchical. It is probably possible to build "houses" of a web-structured hypertext directly into virtual space, but the resulting spatial structures would not be euclidean. While a non-euclidean geometry allows new concepts of neighborhood in the information landscape, it is often feared that it can lead to a problem when users try to build a cognitive map of the environment. A possibility to create a cognitive map is essential for purposeful navigation of an environment.
Several solutions for this problem have been proposed, each with different merits and drawbacks. At first sight link enactment using flying would seem a good solution. On second thought we believe that flying is too distracting and visually intense to be useful for general hypertext linking. The user probably has forgotten why he traversed a link when he finally reaches the end of the link. We therefore enact linking using a subway metaphor. Subways are visually less distracting (albeit boring) and always deliver the user at known landmarks from where it is possible to navigate further to the destination of a link. Navigation starting at a landmark can be supported using light cues, signposts or guides that lead the user to the destination. Another possibility for linking would be to zoom out to a map view of the city and then to zoom in again to the destination. Again this solution is visually very distracting. For quick scanning of the neighborhood of information houses a surface tramway service can be used [TOGN93].
Our city metaphor should not be carried so far that it limits our purpose; it should improve upon it. A single metaphor cannot cope with everything and we could make use of the magical properties of our virtual world where it suits best. Such deviations should, however, be very well-chosen [WATE92]. From research done by Malone (cited in [SHUM90]) it has been shown that the use of an euclidian spatial interface proved not only unnecessary but also intrusive. Virtual space was considered more appropriate as a medium. Local euclidian organization however was seen as ideal, offering useful properties inherent in the physical world. A questionnaire distributed on a MUD (a text-based, 3D virtual world) revealed that teleporting (jumping back) to well-known landmarks is considered a very useful method to reorient. [TROM93].
MUDs are made up of objects. These objects can be rooms, players, things the players can carry and/or manipulate and robots controlled by the MUD library. Rooms are linked in a network structure by means of exits. This network structure has many similarities to hypertext networks. A big difference between a adventure game (or MUD) and hypertext is that hypertexts try to help the users to find the data looked for, whereas adventure games often hide the solution in order to make the game more interesting [NIEL90]. Figure 4 shows parts of a log of a Information City MUD session.
>look
This is an instance of the Hypertext Information Room Vs. 0.1.
There is an information wall with a medium size text that contains
a few links. There are no annotations.
There is one obvious exit: out
> look at wall
An instance of the Hypertext Information Wall Vs. 0.1.
It has some text on it containing three new links.
It has been looked at 3 times.
It has been read 2 times.
It looks very new.
> read wall
This room contains a very early example of a [1:Hypertext] Information
wall. The wall text contains a few [2:Links] that can be
traversed by typing link n. There will be other commands too, when
the basic functionality is debugged. A Link transports the
reader to another room. It is possible to peep through the link to
the link destination to get a preview of what is to be expected
from this link. This is done with the command preview n.
There is another link back to the entrance of the [3:Infocity].
Enjoy.... Juggler.
The text seems to wear out when you read it.
> preview 1
The preview of link [1:Hypertext] is:
The concept of Hypertext.
It looks very new.
> preview 2
The text of link destination [2:Links] starts with:
Linking is often regarded as the essence of [1:Hypertext]. This...
It looks rather new.
> link 1
As you touch the link marked [1:Hypertext] your
finger slides partly into the information wall. The link marker
lights up and the light seems to creep on your arm and your
whole body. You are sucked into the wall and have the feeling of
weightlessness. Hovering for a moment high over a schematic diagram
of the information city you notice the two points in the city that are
connected by this link. You zoom towards the link destination.
A moment later you materalize in another information room.
-------------
This is an instance of the Hypertext Information Room Vs. 0.1.
There is an information wall with a long text that contains
a few links. There are three annotations.
There is one obvious exit: out
Juggler is here.
Juggler smiles at you.
Juggler studies the writing on the wall.
Juggler reads the wall inscription.
As Juggler reads the writing on the wall the text seems to wear out a bit.
>smile at juggler
You smile at Juggler.
Juggler previews link # 1
Juggler says: Follow me, I'll show you around!
Juggler touches link marker [1:Hypertext]. It lights up and
an eerie light engulfs Juggler. Juggler is sucked into the information wall.
You blink as the light fades. It smells a bit from ozone.
As you touch the linkmarker #1 it still feels a bit warmer than
the rest of the wall.
Juggler leaves into the wall.
Figure 4 - Log of a session in the Information City MUD
In our first questionnaire we set out to gather some answers about the kind of information people use to find relative locations in a MUD. Navigation in MUDs can be very confusing - like in hypertext. Still the experienced MUD players seem to have few problems finding their way through a MUD. We wanted to find out what the basic cues for their navigation behaviour were so that we could provide them to the users of the information city.
When entering a new area for the first time most players say they locate a landmark and explore the new area around it, by paying close attention to the descriptions of the rooms and by methodically checking every exit. This creates a mental map of the area for them. In MUDs roads are normally laid out in a rectangular grid. Roads forking in nonretangular angles are therefore also a sort of landmark that helps users to navigate. If the area is very large, or they need to remember things in minute detail, most players revert to creating maps on paper. A small but distinctive group of players says they always make maps on paper if they walk around in a new area.
When purposefully going from one place to another, players rely on distinctive words in the description of the rooms, or objects in the room, and count the number of times they have to go in a certain direction from there. They know this from the mental or paper map they made while exploring the area. A small group of players rely on their friends' memory to get their bearings. They describe their surroundings to a fellow player who may be in a totally different area. This fellow player then directs them to where they want to go (see our assumptions about collaborative navigation).
When totally lost most players have a threefold strategy. First they try to walk in a single direction until they come upon something familiar; if that doesn't work they will ask fellow players; and if that doesn't work they will either teleport, back to some standard place or quit the game and reenter it, which also brings them back to a standard place. Not all players can teleport and either strategy involves a considerable sacrifice of valuables.
Most players say they don't like to get lost, and they try to avoid it. This is largely the same group that always makes maps. Quite a few players say they get lost on purpose sometimes, when they are in the mood for surprises and have time and little to lose at that moment. Quite a few players think that getting lost is an essential element of adventuring, but these players also state they never really get completely lost. The questionnaire was answered by 60 players of a group of 400 regular players, it was a voluntary action and the questionnaire was available for a period of 5 days [TROM93]. It is very interesting to notice that rather similar arguments arise in hypertext navigation discussions: some authors admit there are positive aspects in the disorientation problem, like motivating the user to find their way out, thereby discovering information that would go unnoticed otherwise.
In a second questionnaire we are addressing the attributes people create about the virtual places they encounter in a MUD. An important aspect of a system like ours is how users perceive the spatial structure of the information environment. It is very difficult to assess the impact of room descriptions on the formation of cognitive maps. A room that contains quite a lot of data or has a very detailed description could be perceived as being bigger than another room that is described with less detail. A room described as a hall might be perceived as being bigger than a "closet". How does all this influence the perception of the whole information environment? Does it disturb the user when there are evident discrepancies in the euclidean geometry of the room? The results of these and other questions have not been collected yet, but the results will be presented at the symposium.
"[..] we are likely to see virtual worlds portrayed as the answer to all issues in multimedia interaction, carried forward, often unconsciously, by the unquestioned dogmas of the information age. In the longer term, a more thoughtful approach to harnessing the great potential of multimedia might emerge, though the marketplace is likely to determine developments rather than any scientific agenda."
With the information city and the other metaphors we have proposed here, we hope to have shown that much everyday knowledge can be applied to the user interface to improve navigation of information spaces. We maintain that the use of spatial metaphors will enhance the usability of the information retrieval system. The city metaphor provides it with structure that is flexible and at the same time an organizing principle.
It is clear that there are many unanswered questions about how to utilize the city metaphor and structure most effectively. We hope that this discussion will generate further research into the relationship between the city metaphor and the navigation and the representation of hypertextual information.
[BrRe93] Bruckman A., Resnick M.: "Virtual professional Community. Results from the MediaMOO project", presented at the 3rd international Cyberspace conference in Austin, Texas, 1993
[CaGa91] Catlin K.S., Garrett L.N.: "Hypermedia Templates: An Author's Tool", Proc. ACM Hypertext '91, pp. 147-160
[CONK87] Conklin J., "Hypertext: An Introduction and Survey", IEEE Computer, 20 (9), pp. 17-41.
[CURT92] Curtis P.: "Mudding: Social Phenomena in Text-Based Virtual Realities", FTP-able from parcftp.xerox.com in /pub/MOO/papers/DIAC92.*
[CuNi93] Curtis P., Nichols D.A.: "MUDs Grow Up: Social Virtual Reality in the Real World", FTP-able from parcftp.xerox.com in /pub/MOO/papers
[DIEB93a] Dieberger A.: "Summary Spatial Metaphors", posted to sci.virtual-worlds in may 1993
[DIEB93b] Dieberger A.: "The Information City - a Step towards Merging of Hypertext and Virtual Reality", Poster at Hypertext '93
[DiTr93] Dieberger A., Tromp J.G.: "The Information City - A Metaphor for Navigating Hypertexts", recent research paper at the HCI93, Loughborough, Sept 1993, posted to sci.virtual-worlds in October 1993
[ERIC93] Erickson T.: "From Interface to Interplace: The Spatial Environment as a Medium for Interaction", Proc. COSIT'93, Elba, Sept. 1993, pp. 391-405
[HiHo92] Hill W.C., Hollan J.D.: "Edit Wear and Read Wear", Proc. CHI'92, pp. 3-9
[LBSS85] Lippman A., Bender W., Salomon G., Saito M.: "Color Word Processing", IEEE Computer Graphics & Applications, June 1985, pp. 41-46
[LYNC60] Lynch K.: "The Image of the City", MIT Press & Harvard University Press, 1960
[NIEL90] Nielsen J.: "Hypertext and Hypermedia", Academic Press 1990
[MaOs93] Masinter L., Ostrom E.: "Collaborative Information retrieval: Gopher from MOO", Proc. of the INET93
[SHUM90] Shum S.: "Real and virtual spaces: mapping from spatial cognition to hypertext", Hypermedia, Vol.2, No.2, 1990, pp. 133-158
[SmWi93] Smith P.A., Wilson J.R.: "Navigation in Hypertext through Virtual Environments", Applied Ergonomics 1993, 24(4), pp. 271-278
[TOGN92] Tognazzini B.: "Tog on Interface", Addison Wesley, 1992
[TOGN93] Tognazzini B., personal communication, july 1993
[TROM93a] Tromp J.G.: "An electronic questionaire into navigational behaviour in a text-based virtual reality", posted to netnews group sci.virtual-worlds, July, 1993.
[TROM93b] Tromp J.G.: "Thesis proposal", internal memo, August, 1993.
[WATE92] Waterworth J.: "Multimedia Interaction with Computers: human factors aspects", Chichester, UK: Ellis Horwood (Simon and Schuster), 1992.