The WWW size, in terms of content and variation, is enormous. Experienced browsers can be expected to acquire a certain degree of familiarity with specific parts of the network, but this knowledge will be useless in each unexplored server with its own ideosyncratic design and linkage. We therefore presume information-seeking browsers to apply certain associative 'homing' or 'tuning' heuristics to retrieve information from the network, such as 'Hill-Climbing' [Schildt, 1987]. Hill Climbing in hypertext networks could be conceived as a heuristic in which browsers try to locate information in the network by always navigating towards a decreased 'associative' distance between their present position and their goal.
Browsers assesments of 'associative distance' are most likely to be based on a set of general expectations concerning the general associations between concepts and ideas in the world. Browsers share these intuitions as they are part of their common sense understanding of the world.
If one accepts the notion of browsers applying this kind of heuristic to retrieve information from unknown areas of the WWW, then two conclusions can be drawn concerning the desired characteristics of network structure in general.
First, the structure of hypertext networks should not violate user expectations. Hypertext networks should be structured so as to ensure maximal compatibility between the browsers' commmunal knowledge and the structure of links and nodes in the network. This will allow browsers to apply their existing sense of associative relations to the problem of navigating unknown hypertext structures.
Secondly, heuristic browsing at present works only because hypertext network are in fact associative in nature.[Jonassen, 1990] Links in the WWW and other hypertext networks express and model associative relations (as perceived by their authors), just as is the case in many neural networks [Mc Clelland & Rumelhart, 1989][Hinton et al., 1981]. But, in the case of the WWW connections have no strengths or degree of relatedness assigned to them, except perhaps those implicit in the typographic and graphical design of the hypertext pages. Heuristic browsers in other words lack an expression of the associative relatedness between a certain page and the page behind a certain hyperlink. Navigational, cognitive load could probably be decreased by offering browsers a more explicit evaluation of the degree of association assigned to links from network nodes.
In conclusion, hypertext networks and the WWW lack: