Iterative Design and the Emergence of Form

A new community cannot simply be inserted into cleared land (it could, but then it is not adaptive, and does not form a community). We envision step-wise growth rather than building everything all at once. The design must be allowed to evolve, and cannot be decided at the beginning. A master plan — in the sense of deciding exactly where future construction is to be placed, and exactly what form it will take — is too restrictive and thus highly incomplete. Social housing that follows this mindset by being planned on paper, and then constructed according to plan fails to form a living environment. Following Alexander, we advocate a process in which every future step is influenced by what exists at that point. 

Careful consideration of the topographic features, the existing vegetation, the entry points, etc. should indicate a loose morphology for the entire settlement at the beginning of the planning process. After getting a very rough idea of the placement of buildings and main access road, then individual lots can be envisioned along the roads, which are themselves still not completely specified. Nothing is yet built, and major decisions take place by using wooden stakes and other markers in the ground. In order to guarantee morphological coherence, what is built is influenced by its environment. This interaction is experimentally determined and cannot be worked out on paper or anticipated, due to the complexity of all the mechanisms involved. In a partially built development, the next house or street segment to be built has to adapt its geometry to what was built previously. 

Any decisions made at the beginning of the project must be regarded as recommendations, and not as rigid dictates (unlike those in a master plan). As the project develops in time, decisions made at the beginning for unbuilt areas will now seem incorrect, no longer relevant, so we need the possibility of changing the design continuously as more building takes place. This is exactly what occurred in historical communities built over a time span of centuries. This adaptive procedure (adapting to human sensibilities about the emerging forms and spaces) generated extremely coherent complex geometries in traditional villages and towns, and that coherence cannot mathematically be achieved all at once.

An iterative process goes back and forth between steps, improving each one in turn. That’s what we are describing in adaptive planning and design: first form the conceptual idea on the ground, then introduce the position and size of future built elements without yet building them, then go back to refine the urban spaces, and so on. It is only in this way that the interaction of all the components with each other, and with their surroundings, can effectively take place. Once components begin to be built, then they become part of the surroundings, and in turn influence all future built elements. 

Healthy urban fabric is an extremely complex system, and it cannot be designed and built in a strictly top-down fashion. Some components could be accomplished top down, by someone who understands the required complexity. The ordering has to be emergent from the process, and not simply an imagined outcome imposed by regulatory fiat. There has to be adaptive capacity that is distributed and pervasive in a process that is inclusive. Cities and neighborhoods are “things that people do together”, where a community exercises its territoriality in a positive manner. Any top-down intervention has to be oriented to facilitating that collaboration, not dictating its terms or forcing it into an overly rationalized container.

Construct an urban quarter for a complex society

Both pattern languages and generative processes and codes (either explicit or implicit) have been around for millennia. Pattern languages were codified into practical form thirty years ago. Codes have been used in traditional architecture, and fixed (non-generative) codes widely implemented by one of the authors (Duany et al., 2010). Fixed codes are form-based and tell you exactly how to structure the geometry of an urban environment. Generative codes are more recent, and have the additional capability of evolving the form with the project. They tell you the sequence of steps but leave the form of the end product unspecified. They also distinguish between an adaptive and a non-adaptive set of codes (i.e. those that either generate, or prevent living urban fabric).

Even though a particular project will require careful adjustment to local conditions, these two methods acting together will serve for most cases. We can begin their immediate application using published material, with on-site experience leading to further refinements in the process. In very broad terms, here is how one can follow our suggestions:

1. Use pattern languages to plan the transportation network long before any building takes place. This is essential for generating village and neighborhood centers. Rigid grids favored by central government do not create the necessary nodal connectivity of the urban quarter.
2. Use pattern languages (and develop new ones appropriate to the locality) to construct an urban quarter for a complex society consisting of children, adults, seniors; and including housing, stores, retail, schools, informal spaces, transportation hubs, etc.
3. Existing simplistic (and consequently antihuman) monofunctional zoning must be rescinded by central government. Without that step, all planning schemes preclude urban life from the beginning, regardless of what they might look like.
4. Encourage construction systems (controlled from the top down) to work with local future residents (working from the bottom up) so as to generate low-cost, higher-quality dwellings.
5. Use pattern languages to rehabilitate existing low-income owner-occupied houses, and to convert current rental units to owner-occupied. This requires an infusion of money, but it also generates construction work.
6. Use pattern languages and the notion of the city as a network to orient interventions globally. Larger-scale and longer-term processes will insure that in addition to building housing, projects are conceived and implemented to complete a sustainable neighborhood, well connected in a larger urban setting.

The process starts with identifying the right land. A major problem is that much informal housing is pushed to marginal and problematic land, on which it can be impossible to upgrade. It is necessary that the architect/planner in charge of the project be knowledgeable in pattern languages and their application. Since most architect/planners today are not, we recommend that, at least for the next several years, governments rely on someone familiar with this material to oversee construction projects. A number of professionals are available with this knowledge, though not enough to satisfy the demand. Hopefully, enough young architects can be trained in the following decades to direct new projects. 

One important point concerns building permissions. Because of the organic variability of different components of the project, it is prohibitive in both resources and time to prepare final drawings and get each one of them approved. Planning permission is nowadays usually given for an explicit documented plan specifying every detail of the design, instead of a general process that can produce similar but individual designs. Alexander solved this problem by getting government permission for a specific building process (a set of building operations, within clearly-defined parameters) that generates similar but distinct results. All products of that process were thus automatically approved without further need for individual permissions (Alexander et al., 1985). It is important to get approval from the authorities for the PROCESS rather than for a set of final drawings. If this is not possible, then it is best to get approval for a generally suitable structure that can then be modified under this process.