Group 4's organization was much like Group 3's, although the resulting hierarchy was deeper, as is demonstrated in Figure 4. The decisions behind this organization were based on providing a consistent and comprehensible interface to the user. Group 4's primary concern in writing the document was content; therefore, many of the user features present in other groups' documents were not developed for this document. Instead, each architecture included additional sections called ``Architectural Properties'' which discussed each architecture in terms of ten general dimensions of characterization and ``Issues'' in which some of the limitations and shortcomings of each architecture were examined.
Group 4 decided very early in the design process to limit the size of nodes to a single idea, usually a paragraph or two. Although this made reading more convenient, it made navigation through the document more troublesome than the groups who had chosen a larger node size. To alleviate this problem, three strategies were implemented. The first was to simply enforce the rigid, hierarchical structure. Each architecture was organized in the exact same way. This organization was reflected in the underlying file structure as well.
Second, a graphical overview of the organization was built. This overview was intended more as a map than an access point and did not have the level of detail that other group's graphical overviews included. Instead, as a third step, Group 4 developed what they termed a ``two-click'' strategy. Every node included links at the bottom of the page to return to the index node of an architecture (when reading about a specific architecture) or to the index node of a feature (when in the topics section); the former are shown as dotted lines in Figure 4. On the index pages, there were non-contextual links to the first page in the document. Thus, from anywhere in the document, it was possible to return to the top of the document through two links.
Another feature of the Group 4 document was the ability to read the document linearly through the use of another link provided at the bottom of each page, a ``NEXT'' button. This link was connected to the node following the current node in the hierarchical indexes. The use of these non-contextual links at the bottom of each document represents an attempt to characterize different types of links within the document, using both textual, content links and links which directly aid navigation. Group 4 thought that these last links would be especially important to someone browsing the document that had little knowledge of cognitive architectures.
