next up previous contents
Next: Known Space Ferrets and Up: Designing a System Previous: Building a Known Space

The Known Space Interface

The interface handler takes care of the map display and user interaction. It decides (with user override allowed) how far apart each page and page cluster should appear to be in the displayed map of Known Space, and what icons should denote each page or page cluster. It also handles the actual user interface that interprets the user's commands to move page clusters around, to enclose pages and page clusters in newly created page clusters, to open page clusters, to delete them, and to move them around within the displayed map of Known Space.

When browsing the web, a user should see not just the information displayed on the current web page, but also a small map of the local neighborhood with nearby nodes described. This graphic should roughly indicate the current webpage's relatedness to other pages the user has already visited and to pages not yet visited by the user but already included in that user's Known Space linkage map. Further, this graphic should stick around on screen as long as the user chooses to stay within a particular neighborhood. By seeing nearby linkages explicitly, the user is better able to manage the search task, and that will benefit the web as a whole as well since it should decrease the number of times users have to revisit a page simply to tamp it into their memories of what's on that particular page.

Instead of the present linear list (or linear lists within folders) of today's bookmarks files, the interface handler displays the pages clustered by relatedness in a two-dimensional space--the user's personal Known Space. The user should be able to modify the icons displaying each page or page cluster and move them about inside Known Space. Such movements should be taken as implicit information about how near or how far the user judges those pages to be semantically and should persist across invocations.

Once a map is created, the user could surf the web through the map. That is, moving to a page cluster (a neighborhood) and selecting it will open that cluster, exposing the pages or page clusters within it. Opening one of those subsidiary clusters could expose yet more subsidiary pages or clusters, and so on. Eventually the user will select a particular page. Having done so, not only is a website download request sent out but the page's Known Space neighborhood is still on screen. The user could always move away from the present neighborhood, but websurfing often consists of trying to find other pages semantically close to the originally selected one.

Eventually, the user should stop thinking of those documents that have been bookmarked as files and start thinking of them as sites. Clusters of those places will unconsciously become neighborhoods.. Clusters of those clusters become districts, and so on. This shift from thinking of data as raw information to a set of physical points in an abstract but perceptible and interactive space should add a new dimension to websurfing, and it should decrease the frustration of searching. Couple that with the advantages of having the system aid the growth and pruning of site clusters by automated means and we have something that can truly be said to be a (small) map of the user's Known Space.


next up previous contents
Next: Known Space Ferrets and Up: Designing a System Previous: Building a Known Space
Gregory J. E. Rawlins
1/13/1998