A Swiki or CoWeb is a collaborative Web space that allows people to modify Web content and create new pages in a very easy fashion. Originally developed as a tool for anchored discussions in an educational setting it is a flexible discussion space. Yet like in most of the Web there is little support for showing what people did in this space, where they went etc. We extended the CoWeb to show more of the interaction history for its pages right where people need it: next to links leading to pages. For example the CoWeb generates footprint markers that indicate the amount of recent traffic and when pages were last modified. People typically describe the social navigation CoWeb as being more "alive". The markers provide an additional sense of awareness of activities in the information space and thus increase the sense of the CoWeb as a social "place".
One of the major differences between navigation in the physical and the virtual world of information spaces is that the physical world is a rich and a much more social space. Humans are very social beings and it is common behavior to decide on, for example, a restaurant from recommendations, or simply from "following the crowd" (or avoiding the crowd). Social information is everywhere in the physical world, be that in the form of footprints in the sand or snow, to party noise we hear from an adjacent room, to recommendations from friends and colleagues. A lot of this information may not directly influence our behavior, but it increases our awareness of whatās going on or it is stored away as "something Iāll have to remember for later". Social cues and recommendations of this kind are examples for "social navigation" [1].
Information spaces like the Web are still poor conduits for social navigation behavior. We might be the only person looking at a specific Web page at a certain time, but we cannot know for sure because we cannot see other people and -- maybe more importantly -- we are not aware of them. A slower than usual response time may be the only indicator that a large number of people access a Web site at the same time.
There is a growing trend to provide more social navigation in shared information spaces like the Web. This trend started modestly with the inclusion of hit counters and "personal favorites pages". Lately social navigation seems to become almost synonymous with recommendations and ratings, like on epinions.com, the purchase circles on Amazon.com, or seller ratings on eBay.com.
However, social navigation is a more general concept than recommender systems or collaborative filtering. Social navigation rather is a new way of looking at the design of information spaces where we consciously make these spaces more social "places". Therefore social navigation is a "fundamental model for collaboration in information seeking" [2]. Erickson and Kellogg introduce the concept of "social translucence", which is an important ingredient for creating more social systems and therefore for social navigation [3].
There are two main types of social navigation. Direct social navigation takes the form of recommendations or guiding people to information and thus tends to be a one-on-one interaction. Indirect social navigation instead focuses on aggregating information of an entire community of users. This has the advantage of scaling well and typically avoiding privacy issues. Indirect social navigation thus can be a useful tool to provide awareness of the overall activity in a community. As the goal of our system is on increasing awareness of interactions with the information space, our system focuses on indirect social navigation at this time.
Being made aware of other peopleās activities around us can be more of a nuisance than useful information. For example a loud conversation in a neighboring office or the christmas party in the department across the hall can all be very disruptive to our work. However, it is a well-known fact that awareness of other peopleās activities can be an essential ingredient for collaborative work, see for example [4-6]. I might not be that interested in much of the awareness information, but an overheard conversation and the awareness of what other people are working on can trigger chance conversations in hallways or informal talk before and after meetings. These informal conversations then often prove more important for a project then the meeting itself.
When we try to provide awareness information in an information system it is a difficult design challenge to keep this information from being too intrusive. ICQ is an example for quite unintrusive awareness support. When the name of a member in my group turns blue, I am made aware that this person is now online (this does not necessarily mean that person would welcome interruptions at this time, but this is yet another issue). Coupling awareness information with a persistent discussion space can be a useful tool for supporting distributed group work. Excellent examples of such systems are Babble [7, 8] or Webplaces [9]. Both of these systems use a social proxy to visually show the activity in a group of users.
Another approach for awareness support employs non-speech audio cues. The advantage of using the sound channel is that sounds can inform us of activities even while we are looking elsewhere or when working in a different application. An important issue with sound is that it has to be noticeable enough to be perceived, but should be subtle enough to not disrupt work. For interesting work in this area see the Nomadic Radio system which uses a contextual notification model to avoid interrupting conversations [10]. Examples of systems using sound to support collaboration and enhance the awareness of background activities are [11] or [12].
Our work is based on a collaborative Web space named Swiki or CoWeb. CoWebs were developed by Mark Guzdial to support anchored discussions in an educational setting [13, 14]. CoWebs are persistent discussion spaces that allow any user to modify any page. A CoWeb page doesnāt look too different from any other Web page, except for a link or button reading "Modify this page". When clicked this presents the user with a form containing the current content of the page. Users can use HTML in their pages, but they donāt have to.
Links to other pages on the CoWeb are created by enclosing a page name in asterisks, like *my page*. If that page doesn't exist yet, a new page is created on the CoWeb. Interaction with the CoWeb thus is extremely simple. This simplicity and the possibility to create new pages almost without effort wherever people like makes a big difference in the type of discussions occurring on the CoWeb [Mark Guzdial, pers. comm.]. It should be pointed out though, that people tend to be shy about creating pages and modifying other people's pages, because this is not "how the Web typically works". Making people feel comfortable with modifying pages often has to do more with learnt conventions on accepted Web use and other social connotations of a space than with the tools we provide. These issues are discussed in more detail in [15].
CoWebs often are left unprotected so that even people from outside the typical user group can modify pages. One might think this is a receipe for disaster, but it turns out this openness is rarely abused, maybe because people perceive no challenge in breaking a CoWeb. Furthermore, as everybody can change anything anyway, another user simply takes offending material off the CoWeb again [Mark Guzdial, Lex Spoon, pers. comm.]. Such behavior indicates that CoWebs are social places where people sometimes feel "ownership" of the space. Therefore they are motivated to care for the shared space in some form or the other.
The original CoWeb provides a list of recently modified pages to alert people of new contributions. However, this list tells only part of the story. To be aware of what is going on in a CoWeb's social space I need to also know what information was accessed and what information is requested especially often. In addition this information needs to be provided in context. In an anchored discussion, where users can create a discussion about a topic at any place, this information thus needs to be available right at this anchor point: at the link leading to the discussion.
We modified the CoWeb to indicate both modification and access history for pages. This information is made available both on global lists as well as in context by placing social navigation markers next to all links within the CoWeb. See Figure 1 for a typical page in the social navigation CoWeb.
For access information we use the metaphor of footprints. Small footprint symbols in three different colors (gray, orange, red) are placed right next to links to indicated the amount of traffic the page behind that link received in the past 24 hours. A second marker, again in three different levels, indicates how long ago that page was last modified.
We further provide a "lack of activity" marker in the form of a little dinosaur. It indicates that the corresponding page has not been accessed for a longer period of time. This dinosaur marker proved to be especially useful for getting a quick overview of what discussions have "died out" in a CoWeb: one look at a page full of dinosaurs shows that this is probably a "dead community" (however see the caveat in the section on "What's next"). For an example of how dinosaur links stick out of a page, especially when compared to a very active CoWeb, see Figure 2
Because we provide social navigation information in context of the page, we make people aware of changes and accesses to the information they are currently working with. While a global overview of changes and activities is important, this contextualized information is essential both for focused work and for supporting awareness of active discussions linked to the current page.
We should point out that also the global lists of recently modified pages (see Figure 3) and recently accessed pages (see Figure 4) contain the social navigation markers. These lists provide a global snapshot of the recent activity on the CoWeb.
A recent addition to our server is access to the last few lines of the server log file. Originally we provided this function as a tool to administrate the server, but the log access soon became one of the most popular tools. The reason is that the log files give detailed information on the last few activities on the server without modifying the interaction history. An author can check who interacted with his or her page without causing misleading footprints. Meta behavior of this kind indicates that people consider the social navigation markers indeed as relevant information.
At the time of this writing we work on a detailed study to assess inhowfar our social navigation markers influence people's navigation behavior in the CoWeb. We also observe people's behavior in the community of our social navigation CoWeb and compare it with experiences from standard CoWebs. On standard CoWebs people tend to first check the "recently modified" pages list on every visit, because this is the only history information they can access. On the social navigation CoWeb this and more interaction history information is provided in several ways. Initial results seem to indicate that people rely much less on the recently modified list and still are well aware of recent activity on the CoWeb.
While the original CoWeb was built for educational purposes, anchored discussions are useful in a variety of settings. We found the CoWeb to be very helpful as a brainstorming device and idea and note repository for our projects. In this particular application the social navigation markers served as a visual reminder of what parts of the projects we hadn't looked at lately and were likely to forget about. The markers also indicated what topics colleagues were working on right now, because these areas were recently accessed or even modifed.
We further used the CoWeb to generate lists of to do items. We could easily turn these items into a link when we needed to discuss them. Items that eventually showed a dinosaur either were items we had ignored for too long or tasks that were already taken care of.
The very informal use of the CoWeb and the social navigation CoWeb is possible because there are no fixed roles. Edwards points out in [16] that roles in collaboration often are very fluid. The lack of strict roles and strict permissions in the CoWeb allows such fluid role changes: One is not only an author or a reader. Instead it is possible to fix a minor spelling mistake on somebody else's page when I happen to notice it or to add a short comment somewhere else. We believe that this fluidity supports the informal and flexible use of the CoWeb. (Note: There are CoWebs with stricter types of access control).
CoWebs are implemented based on the PWS (Pluggable Web Server) in Squeak, a freely available dialect of Smalltalk which exists for various platforms (PWS is included with Squeak ). In our CoWeb pages are objects that know about their interaction history. Keeping the interaction history within the server restricts us to showing interaction history only for pages on that one server. Even links leading to outside pages cannot show social navigation information.
We have come to realize that keeping the interaction history entirely separate from the CoWeb might be a better approach -- at least for some applications. A possibility is to use intermediaries [17] to add social navigation information to CoWeb and Web pages. An intermediary essentially is a proxy server sitting between a browser and the source of the information (the server) analyzing requests and modifying information returned from the server. Such an intermediary can provide history information just for one user, or for a group. This would allow aggregating interaction history for a specific group of users and therefore be more relevant for that group. A further advantage of this approach is that social navigation information can be generated for larger sections of the Web.
Another shortcoming of our current version is that the social navigation information is relatively inflexible: no matter whether a CoWeb gets a lot of traffic or only a few hits a week, the footprints are based on a static time range. However, like other virtual communities, different CoWebs and even different parts of the same CoWeb can have very distinct rhythms [18]. This leads to some CoWebs appearing as "dead" (showing mostly dinosaurs) although they might be an alive, but slow-moving community. Likewise, very active CoWebs contain mostly red footprints and new markers and thus reduce the utility of the markers.
It is obviously necessary to apply a more dynamic approach to social navigation information where footprints point out information if it is used frequently relative to other information nearby. This approach was used in the author's Juggler system [19]. Juggler also differed from the CoWeb in the fact that it showed interaction history about the link, instead of the page. This made sense in the very spatial setting of the Juggler system. This difference between Juggler and the CoWeb is like showing whether most people entered a store through the door I'm facing vs. showing how many people accessed the store.
Another issue with the present version of the social navigation CoWeb is that every user causes interaction history. As mentioned above, people sometimes use the logfiles to specifically avoid causing footprints on pages. In CoWebs that are open to everybody, visitors from outside a work group cause footprints as much as group members do. Instead we might want certain people to cause more (or less) interaction history than others. For example a teacher's interaction history can prove more valuable for a student trying to understand how to best use an information space than the aggregated interaction history of all fellow students [15]. Similarly in a document I need to edit together with a colleague I might want to see what parts my colleague looked at already so I don't bother working on these.
In this paper we focused on the basic ideas of social navigation support in the social navigation CoWeb. Our system provides a number of additional tools, like a quick navigation popup menu and sonifications of the interaction history. These tools are designed to provide both a visual and an acoustical overview of the interaction history and are described elsewhere [20]. We further employ server-side sonifications to support awareness of the activities on the CoWeb. Of course only a user sitting next to the server benefits from this particular type of sonification. Still we found this type of awareness support particularly useful when coordinating a project through a CoWeb or when monitoring the activities in a group.
We would like to thank Mark Guzdial for creating the original CoWeb and Peter Lönnqvist for his assistance with the current study of the CoWeb. We further would like to thank Nils Dahlbeck, Mark Guzdial, Kia Höök, Colleen Kehoe, Peter Lönnqvist, Lex Spoon, Annika Waern, and the entire Persona group for valuable discussions about social navigation and the CoWeb in particular.