1. Background

Since the inception of this project in 1972, efforts have been made to bring together the resources, hardware and expertise to permit networks of problems to be portrayed as networks in graphic form. The data is organized to that end. Ideally this technique would be used to communicate with international organizations concerning the networks of problems within their area of concern. By representing networks to organizations, they would be encouraged to identify weaknesses in the pattern of relationships indicated and would be able to make suggestions for specific additions or deletions. Similarly the editorial work on networks of problems would be considerably facilitated. And of course a set of network maps would be an extremely valuable tool as a complement to the text information displayed in this volume.

Despite efforts following the 1976 edition, it was not possible to achieve this for earlier editions. And despite efforts since then, no graphic representations have been possible for this edition. However as will be seen from the inside covers of this volume, it has been possible to experiment with a new software package using the entierty of the problems database. The approach is discussed in Section TZ, with an indication of progress to date. Aside from the technical issues discussed in Section TZ, the concern here is with new ways of exploring information on complex networks of problems. The point has been repeatedly made that information on problems is now overwhelming. Conventional ways of exploring that information are totally inadequate to any insightful overview. Some possibilities are outlined below.

It should be noted that the emphasis here is not on the dynamic approach favoured by "global modellers" in order to represent complex systems. Extensive resources have been allocated to this approach for nearly 20 years without any major breakthroughs of policy significance. Such modelling totally ignores the question of the comprehensibility of any output to those who have to make decisions, even though it concentrates on a small sub-set of quantifiable problems.

The concern here is with the larger challenge of representing the extensive network of problems, whether tangible or "fuzzy", in ways which give a sense of context. The intent is therefore both less ambitious from a theoretical point of view and more ambitious in terms of encompassing the full range of issues with which people are concerned -- whether or not they can be reflected in any systems model. From a mathematical point of view, in contrast with conventional modelling the tools used are those of graph theory, topology and work on partially ordered sets. But the intent is to use these to design appropriate algorithms to simplify the user's understanding of the patterns of problems. The following facilities acquire greater significance with the increasing use of CD-ROM to hold information on large databases and to facilitate access to them.

2. Representation of problem "hierarchies"

There is a need to be able to portray a problem hierarchy fully for inspection. Whilst this can be done in a non-graphic mode (as is currently the case in editorial work), the approach is cumbersome. Ideally users should be able to view the whole hierarchy or zoom into parts of it, accessing the descriptive data when required. This facility is of special interest given that problem hierarchies are not simple. A single problem may be part of several tree structures.

3. Representation of sets of problems

Although relationships between problems may not be known, it should be possible to view those in related subject domains. Again this can be done in a non-graphic mode, but users need the facility of looking at the problems in any domain at different levels of detail, whether the tops of hierarchies only or at any of a succession of levels below. The need for this facility is obvious in the case of geographic information, where the need may be for more or less detail (eg including or excluding towns of smaller size). Increasing detail would thus include in the graphic display the hierarchical links to more specific problems, just as minor roads to smaller towns are included when a more detail geographical map is used.

4. Vicious and serendipitous cycles linking problems

There are two patterns of functional relationships between problems in the database at present: aggravated by/aggravating and reduced by/reducing. Again these can be explored in non-graphic mode but this becomes especially difficult when combined with the question of hierarchical level. Specifically the difficulty lies in determining to which hierarchical level a functional relationship should point. Much greater flexibility is required by users in navigating through the networks in order to get a sense of the pattern of relationships at the appropriate level of detail. Preliminary work has been done on identifying vicious loops linking 3 or more problems in cycles. This is discussed in Section TZ.

5. Configuring patterns of problems

In the features discussed above the emphasis is placed on holding the data in a form which respects its inherent complexity. The user is obliged to navigate through it using the new manoeuvring possibilities. This only responds partially to the user's needs to obtain an integrative overview of the data. Other approaches are possible if the user is allowed to experiment with various distorted presentations of the data (analogous to the need of map-makers to use various "projections" when presenting data on the globe on the surface of a sheet of paper). Some possibilities include:

(a) Grid representation: Forcing the problems onto a 2-dimensional grid, possibly zoned by subject, and allowing problem hierarchies to project vertically according to the number of levels required. Many-levelled hierarchies in a particular zone would then be represented like sky-scrapers compared to "single-story" problems elsewhere. This would trigger an understanding of the "urban" organization of the set of problems, distinguishing between the "built-up, down-town" area and "suburbia". A modified version of this approach could also be used by treating different subject areas as distinct "towns" and providing maps for them, each with its own "down-town" area. In both versions the grid lines can be treated as "roads" with "traffic" of a certain frequency corresponding to the intensity of the relationship between problems. Use of the visual "town" metaphor may evoke new approaches to problems in the light of the well-developed skills of "town management" and "traffic management".

(b) Other two-dimensional projections: As the map-making analogy suggests, there are many other "projections" in two dimensions onto which problem network information might usefully be projected. In each case the emphasis is on using the geometry of the projection to provide an artificially integrated overview and a way of articulating the detail of the networks.

(c) Spherical projections: The complete pattern of problem networks could be projected onto the surface of a sphere as a way of emphasizing the bounded nature of the system within which the problems occur. This may have advantages in rendering explicit symmetry and other features which might render the whole more comprehensible. Approximations to such a spherical representation may be more practical, along the lines suggested by Buckminster Fuller.

(d) Unconventional surfaces: Computer environments increasingly permit the projection of data onto surfaces which do not need to correspond to conventional rules of geometry. This may be especially fruitful to encompass contradictory patterns of information which can best be embedded on dynamic surfaces, or shifting patterns, where some of the integrative power comes from the periodicity or rhythm with which data is portrayed in a particular way.

Given the present conceptual bankruptcy in the face of the complexities of the problematique, there is a strong case for experimenting with unusual ways of portraying networks of problems. There is little hope that major breakthroughs can emerge from agendas formulated in terms of a linear sequence of points, as remains standard practice in policy-making discussions. The hardware and software skills to engage in such experiments are readily available and widely used outside the policy environment.