In review, I read the following books for this class:
I also read chapters or sections of the following:
Some interesting issues that arose during discussion over the term were the nature and practical application of categories, getting a grip on dynamic content, and defining IA.
Category boundaries are nearly always vague, and group memebership is graded and based on prototypes. Though this has been recognized in searching, and solutions such as neural networks and fuzzy sets have been proposed, the ramifications for browsing are not yet fully understood. "Fuzzy browsing" may be a new field of interest. Considering site structure and labels (and the navigation stemming from these two) are all usually created based on categories and classification, this nature of categories should have a big impact on site design. This is even more true when we realize that people have such an abhorence to placing the same item in more than one category--there is less chance of overcoming poor fits or multiple interpretations of category membership by simply putting a member in a second category.
Reading about animal taxonomy was also interesting. It was an example of practical application of the conceptual readings. Some points to note were the resulting Linnaean hierarchies were not balanced across branches. Membership was often prototypical and fuzzy and underwent revision as more information came to light. Though there are many possible aspects on which to categorize, there is a desire to find some "fundamental" system. Consistent and unique labels are vital. Categories are impacted by politics.
A recuring conversation topic was IA for growing, dynamic content, especially in online courses. The trouble seems to lie in the fact that the structure must be generated on the fly. In descriptions of the IA process, organization is usually accomplished by gathering all the content in one place and chunking it into groups and divisions. But how to do this before the content exists? A related complication evident in online courses, large intranets, and in discussion of my rather defunct animal classification project, is how to control user-generated content? In this case, not only is the content generated before the structure, but the structure is also determined by many different people at the same time. An IA nightmare! Yet there seems little an IA can do. The only help is to attempt to predict content as much as possible and base a structure on these predictions. An IA style guide or a moderator of some kind would help. And of course, updates and redesigns once the nature of the content is understood will always be necessary.
We also wondered, as all IAs do these days, just what is Information Architecture? I think, in our discussion towards the end of the course, we struck on a good definition: creating an organized structure of information in order to aid human users. Yet this process involves gathering and filtering the information, chunking and organzing, building and communicating the structure through a navigable interface, and the usability and aesthetics necessary for a human user. Each task in this process has attracted adherents from different disciplines: LIS, ICS, HCI, graphic artists, etc.
The investigation of wayfinding in real-world architecture, and then wondering how this could be translated to wayfinding in information architecture, proved to be one of the more interesting areas we discovered during this course. I'd like to take this opportunity to sum up some of the similarities and parallels here by comparing Rosenfeld's description of IA and Passini's description of wayfinding.
Once the initial planning and needs assessment is completed, Rosenfeld begins the process of site construction with the organization of content into a structure. He admits many of the problems discovered in our other readings: ambiguity of category membership, heterogenity of members, differences in personal perspective, and the effect of politics on classification. But, after overcoming these handicaps as well as possible, we must determine some scheme or principle on which to group and organize. This scheme can be exact (as in alphabetical, chronological, or geographical) or ambiguous (as in topical, task or audience oriented, or metaphor-driven) or some combination of these. This scheme then leads to a hierarchy, hypertext/web, database/tabular, or linear structure.
Passini, in his discussion of structure, refers to Kevin Lynch's study of city structure. Lynch identifies 5 elements that people use when navigating:
|Element||Architectural Description||WWW Element Translation|
|paths||the channels of transport, such as roads, elevators, paths, etc.||links, as they are how we travel between points on the Web|
|nodes||entry and focus points, such as squares, intersections, circles, and other open places.||since link paths lead to and from pages, webpages seem to correspond to nodes|
|landmarks||physical reference points, objects of note, buildings, mountains, towers, stores||page content, images, labels.|
|edges||linear elements not used as paths; especially important as borders, such as rivers, coastlines, walls, etc.||hardest element to find on the web; a strong site image provides a external border, but only noticable once crossed|
|districts||medium to large areas with recognizable flavor or distinction||subsites and differentiated areas of a site|
Though these elements were orginally relevant to wayfinding in a city, Passini believes they apply to internal sturctures as well. Their translation to the Web is not perfect, however. For example, I'm not sure if pages are landmarks themselves or simply nodes containing landmarks. I favor the latter definition, since a webpage feels more like a space: the content that can be updated or changed, yet the address/destination of connecting paths stays the same.
Another complication is that paths (links) have no length. In this way, webpages are like indoor rooms, connected by doors and elevators. There is no chance to look around or to orient oneself while traveling along a path. A user is either on one page or another. (Indeed, attempts to build longer paths on the web--a string of nodes, in a way; specifically, pages with only one exit, such as entry forums and splash screens that are meant to orient but do not provide navigation decisions--have not met with much user success. In fact, most are disliked and clicked-thru as quickly as possible.)
As I mentioned, webpages seem more like rooms. This makes the use of landmarks difficult. Landmarks are largely useful because they can be seen from a distance and provide positional clues. However, a landmark on one page cannot be seen from another page. Yet landmarks also serve to identify a certain place, even if they cannot be seen from the surrounding area. For this reason alone, we should perhaps try to use more landmarks on the web. Indeed, Passini claims that landmarks are the most important of these elements when it comes to wayfinding. The direction-finding aspects of landmarks could be simulated by using identifying icons related to certain districts or pages with links to those pages--sort of a "This way to [landmark]" sign. (An example of landmarks is the use of colored folders at vsmith.ws.) Because webpages are more like rooms, I think we should look more towards studies of wayfinding in large internal structures, like airports or museums, rather than only at city wayfinding.
Though he admits the use of Lynch's elements in wayfinding, Passini considers three other aspects of structure to be even more important. These are spatial organization, spatial enclosure, and spatial correspondence. Spatial organization is simply being able to grasp the principles on which the spaces are contrued and related. This is basically Rosenfeld's organizational scheme. If a scheme has been closely followed, people will be able to predict the structure once they recognize the scheme.
Spatial enclosure is the way that external boundries give a clue to internal structure. For example, when a building has a rectangular perimeter, people assume that the inside structure is also rectangular. In cases where rectangular buildings have a triangular internal circulation, major wayfinding difficulties are reported. Here, websites are at a major disadvantage because there are so few edges online. Certainly, a website's boundaries cannot be seen from outside the site. However, a strong site image can make the external boundry recognizable when a user crosses it. Also, some way of denoting links to external sites would likely be useful. This would give a clue that the user will not only be leaving this room (page), but leaving the whole building (site).
Finally, spatial correspondence is how one area lies in relation to another. Again, there is very little help relating the position and structure of one site with respect to another. Indeed, such a relation currently has no meaning on the Web. However, we can do more to show correspondence between parts of the site within the entire structure, much as rooms are related to each other within a building.
These elements--Lynch's five and Passini's three--all impact the imageability of a structure. If they are present, a user has an easier time forming a conceptual model of the space. But imageability is filtered by legibility. This means that if the environment is too cluttered or confusing, if there is too much information or contradictary information, then the user will not have access to the information needed to from a mental image. This is an argument in favor of a simple, usable design that emphasises the structure of a site.
(Possibly related to this examination of structure is our brief discussion of how people use books. They look at the cover, read the back, flip through the pages, look at the table of contents. Once they have established the context, they can begin reading. To do the same overall examination on a website takes much longer and is a more daunting task. I believe this is again related to lack of boundries and enclosure. The benefit of a book is that, even though it is information and can have external links in the form of citations, it is still obviously discrete and understandable. Though the cover art and chapter headings may serve as landmarks, the edges may be even more important here. This example makes me want to examine Lynch's elements in more detail in different evironments. For example, on the web, paths (links) are key. Yet on the open sea, there are no paths; or rather, there are an infinite number of paths. Passini's conclusion that landmarks are the most important element in the built environment may be because they are what is most commonly lacking, misleading, or useful there; other environments may make other elements more important. In particular, I find it difficult to image another environment that is as "edgeless" as the web. Though there are dead end pages, there is no real perimeters to the web that one can cross and exit or follow around the outer edge as a navigational aid.)
Some people navigate linearly and some navigate spatially. When they understand that they should use one method or the other (perhaps because there is not enough environmental information available for one method), they can switch modes. It seems that, though there is a spatial structure to websites, it is not really used. Jared Spool's usability tests claim that users do not form cognitive models of a site, but instead decide which links to follow at each decision point. Perhaps this is because site structure is an unreliable guide on the Web. Also, it takes time and exploration to produce a cognitive model of a site, whereas linear navigation can be started right away. And because links are also used as signs (discussed below), it can hard to determine whether people are following a sign or a cognitive model. I would like to examine the depiction of spatial orientation of information, as in VRML, 3D graphs, clustering, etc., in more detail, particularly how people wayfind through these alternate depictions. Also, it would be fun to test web designs where the structure is emphasised. For example, if a site is structured hierarchically, links to pages further down the hierarchy could be placed at bottom of the page, and links up placed at the top.
Yet, for all of this, I think the lack of certain aspects of the Web will alway make spatial wayfinding difficult. This does not mean that there should be no structure, however. A structure provides a framework for a navigation system, and, though perhaps not the primary mode of navigation, it does provide helpful clues. I suspect that the lesson here is not that structure in unimportant on the Web, but only that users rarely use a spatial wayfinding technique. Instead, they travel linearly, based on navigation aids and signs.
Rosenfeld warns first to produce a consistent site image and not to disable the few navigational aids provided by browsers. The actual navigation within a site can be hierarchical (as in a table of contents/site map), global (the few links that are always available), local (sub-site navigation), and embedded or ad hoc (links within the content or text). The book also discusses the advantages and disadvantages of nav bars, pull-down menus, frames, and remote navigational elements (like tables of content). In conclusion, it points out that there is no one answer, and that the navigation must arise from the site's overall structure and purpose.
This discussion seems to mirror Passini's description of signs and maps, which should also be determined by a space's structure. Passini claims there are three types of sign--directional, identifying, and reassuring. Links seem mostly directional, labels and headings are identifying, and, since paths are of no length on the web, there is little chance for reassuring signs, which occur at points where decisions cannot be made. (However, as shown in usability studies, users often do prefer more information about links, such as files size of multimedia or more details concerning the destination, before they click. Roll-overs, pop-ups, and title attributes can serve as reassuring signs on the web.) Navigation--whether signs or links--should be consistent in form (color, typeface, etc) and placement (certain eyelevel or on certain part of the page). They should also be complete; if you have signs pointing to a certain destination, you must continue to provide such signs until that destination is reached. Their content should be brief and chunked in such a way that they can be skimmed by users in a hurry. The intended audience should be obvious.
Rosenfeld's remote navigational elements--site maps, tables of content, and indexes--seem to correspond to Passini's [stationary] maps. Maps help both in determining a route from here to there and in building a conceptual image of a space. Maps should also be easy to read quickly. They should abstractly show those elements important to wayfinding--basically the eight elements discussed above in Organizing and Structure--but they must still provide enough information to support decision-making. Maps should be easy to relate to their surroundings; there should be a definite "You are here" gestalt. This is one aspect lacking in remote navigational elements (by definition, they are remote!). Indeed, most site maps are separate from the rest of the navigation, and there is certainly no "You are here." Using maps in frames or in a pop-up window might overcome this limitation.
Passini give a very nice 7-step "Ariadne's Thread" design guide, summarized in Week 14 comments. By identifying the users, tasks, and conditions common to the structure, one can determine major traffic routes and decisions that need to made along those routes. Then environmental information, such a signs, can be placed at those key decision points. I would recommend using this process when designing navigation for a web site; as Rosenfeld recommends, it bases navigation on the structure and users.
One complication of this analogy between signs and navigational links is that links are also paths. Paths of no length, no less. So we must think of links as two-dimensional teleportation gates or doors in a three dimensional architectural space. Yet when these two-dimensional gates also serve as signs, it becomes confusing. Signs are merely informational, while paths are structural. Thus, in a way, creating navigation also changes the structure by opening paths between different parts of the site. Yet these paths still seem somehow secondary. Or at least they do in a good design. (See OldGlory15s.com [has been redesigned since] for an example of different navigation imposed over a site structure.) As mentioned before, I think more research needs to be done on just what does define site structure and how it is different from navigation. Though the boundry between the two is definitely blurred, there does seem to be a difference.
Overall, I think there are strong parallels between wayfinding in real world architecture and wayfinding in information architecture. Based on the preliminary comparisons and translations discussed above, I think more reading and more research needs to be done find where the analogy is weak and where it can be used as a guide to better IA design. Indeed, not all IR is wayfinding, and so other analogies must also be used depending on the situation. The fuzzy nature of the organizing categories and dynamics of changing information provides further complications.
I have enjoyed this class very much. I think we have looked at IA from a number of different angles. I look forward to more study in this area.
|Last Edited: 20 Aug 2003|
©2001 by Z. Tomaszewski.