Structuring the Information II

Another program, VTCON, is called into play to provide additional structure beyond that inherent in the graph. The graph establishes paths through the Functions by linking Functions when they are related to each other, but, unlike a road map, a graph

is not necessarily arranged nicely for visual inspection. As it is obtained from the RELATN program, a graph is only a list of what Functions are linked to what other Functions. To draw out the analogy, it is like being in a town and having a list of towns that are next on each road out of town, but not being able to find out whether any of those towns have roads between them without going to one of them or consulting a similar list of roads for each town. If a bird’s eye view were possible, clusters of towns interconnected by roads would be obvious. Unfortunately, for complex graphs, endless visual interpretations are possible, and it is extremely difficult to show one as an optimally arranged “map”. What can be done – and what the VTCON program does – is to find the clusters of Functions (vertices) algorithmically (Figure 13). With that information, the purposes of the map can be achieved.

The clusters are important because they represent primary groupings of Functions. Once the clusters have been found, the planner can choose a Function at will and know which other Functions are of direct concern. Of course, Functions are also linked to others outside their primary clusters or the graph would be unnaturally disjoint. These cross-cluster links provide the basis for higher level, broader-reaching clustering, and VTCON uses them to create a condensation hierarchy (Figure 14). Clusters are themselves clustered

based on Functions held in common and links between Functions in different clusters. Levels of hierarchy are produced with smaller numbers of larger clusters at each succeeding level until the entire graph is condensed into a final cluster,the original set of all Functions. In form, the hierarchical structure is a semi-lattice rather than a tree because Functions can be in more than one cluster and clusters can be themselves members of more than one higher level cluster. This is a very general form of hierarchy and one most appropriate for planning – where it is natural to expect a Function to be performed in more than one Activity. Functionally, the hierarchy is an Information Structure, a specialized structure for synthesis. The actual Information Structure developed for this project is shown in Figure 15.