Notes on the Synthesis of Form
Notes on the Synthesis of Form is 191 long but a fairly quick read. Alexander includes a number of sections on mathematical analysis and proofs based on set theory - these are not essential to understanding the basic concepts. As Alexander points out in his preface to the 1971 edition, his 1964 book was focused on the process; he subsequently came to realize that the diagrams themselves held great power.
Although I will try to accurately summarize Alexander's arguments, I may have misinterpreted his thoughts or used terminology that has connotations which he did not intend. I have included my crib notes which include quotes that were either particularly good or where I was not confident of my interpretation.
Alexander argues that problems are becoming increasingly complex, to the point where they exceed the cognitive ability of designers to understand the problem and conceive whole solutions 'from scratch'. Often, problems come with multiple, poorly understood requirements that interacted with each other, creating conflicts and contradictions. As a result, designers sometimes rely on the authority of formal 'styles', or retreat to artistic intuition and extreme forms of individualism. Although designers and engineers might take exception, the impression of design as a frivolous and largely artistic pursuit suggests there is some merit to Alexander's comment that "... in an era that badly needs designers with a synthetic grasp of the organization of the physical world, the real work has to be done by less gifted engineers, because the designers hid their gift in irresponsible pretension to genius."
Definition of 'Goodness of Fit'
Alexander starts with the ensemble of the context and the form. It is the success to which form is adapted to the heterogeneous and conflicting forces of the context that determines the degree of 'fitness'. Depending on where the boundaries are drawn, many divisions (contact points) between form and context can be defined. These often occupy different scales or levels in a hierarchy. A whole network of adaptations is required to achieve internal coherence without stress or friction.
It is difficult to specify a complete set of requirements that need to be met to achieve a good fit. A practical approach is to define 'good fit' as the absence of 'misfits', since these are usually what makes the problem obvious and can be ascertained through inspection of prior designs. Although designers may argue over the importance of a particular misfit, they are less likely to disagree on whether the misfit exists.
Only rarely do we have a complete description of the context such that we can determine through analysis or mathematics the fitness of a form separate from its context. Alexander calls these 'selection problems' where the set of solutions can be created symbolically and the solution criteria can also be expressed symbolically. No formal design is required since the context clearly defines the form. In most cases, these conditions are not met and invention/experimentation is required, because:
- we cannot symbolically modify existing designs to solve interesting problems
- we need to test the solution in the real world, or rely on intuition/experience
- as in science, the data (context) is not sufficient to define the hypothesis (form) - other principles are required that lie outside of the context
The Unselfconscious Process
Indigenous architecture is often better adapted to its (albeit simpler) context than our designs, demonstrating integration at multiple levels (construction, operation, maintenance, social context, environmental 'fit'). In contrast, our designs often suffer from excessive simplification (in the pursuit of clarity) or at the other extreme an unintegrated attempt to meet a list of requirements without resolving the inherent conflicts. Indigenous cultures rarely have the concept of a 'designer'. The same person develops the form, builds it and uses it. Individualism is not highly prized. Tradition, folklore, ritual and taboo have a strong influence in limiting 'change for the sake of change' - once a misfit is resolved, the impulse driving further change is eliminated. At the same time, adaptation to misfits is rapid due to the immediacy of the builder to the materials and the usage of the form. Constant maintenance of most designs provides an opportunity for small-scale changes.
It is the combination of sensitive feedback at the local level (immediacy) with damped feedback at the system level (viscosity due to tradition) that appears to explain the ability of the unselfconscious process to support hierarchical, self-organizing systems that consistently produce well-fitting forms. In these balanced (homeostatic) systems, the rate of adaptation is matched to the rate of change at each level.
The Selfconscious Process
Alexander claims that selfconscious cultures encourage the craft of design which breaks the immediacy between identifying and correcting misfits, and also encourages change for its own sake. Whereas the rules for creating form are not explicit or written down in unselfconscious cultures, we pass on design knowledge through formal education of generalized principles, by specialized teachers separate from both practitioners and users. This creates a feedback loop that accelerates the rate of change beyond the ability of the system to adapt. Rather than been in a state of near-balance requiring only minor adjustments at a local level, our systems are moving away from equilibrium. <This reflects the discussion in Into the Cool on how far self-organizing systems are from equilibrium>
Designers are increasingly called upon to solve systems-level problems without the ability to deal with the complex network of conflicting requirements. One 'solution' is to create verbal constructs: sets of requirements that may make sense in terms of language ('safety', 'production cost') but do not reflect the structure of the underlying system. Verbal constructs that are easily expressed tend to carry more weight, regardless of their actual useful. These verbal constructs influence how we solve problems and may prevent us from understanding the root causes of these problems.
Analysis and Synthesis
Alexander describes three design models:
- direct, two-way interaction between the context and the form, with humans as agents resolving misfits ("unselfconscious")
- formation of a mental picture of context, development of a mental picture of the form, translation to the form in the real world ("selfconscious")
- formation of a formal model of the context that "retains only its abstract structural features", development of an "orderly complex of diagrams" from which the mental picture of the form is derived
Alexander argues that a problem defined by a set of requirements ('misfits') can be decomposed into a tree (hierarchy) of nested subsets that represent sub-problems. This decomposition is useful only if these sub-problems are solvable (lead to coherent 'diagrams' that relate form with function) and the sub-solutions can then be fused without requiring significant changes to these sub-solutions. This requires that a specific requirement subset relates to some underlying attribute (identified by highly interconnected 'misfits'), and that the requirement subsets are only weakly linked. The practical implication is that resolution of conflicts between requirements is moved to the early (sub-problem) stage of design where flexibility is greatest.
Alexander describes a model of interconnecting lights that is very similar to Kauffman's simulations in At Home in the Universe. A light that is on indicates a 'misfit' of a specific requirement. A light can turn on a nearby interconnected light. The time it takes for the system to regain stability (all lights off) after one light is flicked on increases exponentially with increasing interconnectedness. However, multiple independent subsystems that are internally richly connected can rapidly regain equilibrium.
Alexander places a number of constraints on requirements:
- have roughly the same scope and significance (constrain a comparable number of forms, are not nested)
- are as independent as possible (this may require significant restatement of the requirement)
- have low three-variable (or higher) relationships (are as specific and small in scope as possible)
The remaining relationships between requirements are assumed to reflect the underlying structure of the problem, typically some physical law or constraint. It is rarely possible to understand these relationship by looking at the correlation between misfits in existing designs. Rather, designers need to look for causal relationships. Although based on the knowledge of the designer (and therefore subjective), the process of uncovering the rationale for these relationships may suggest novel ways in resolving conflict.
Appendix 1. A Working Example
Alexander describes a project to build a plan for a rural village in India. He derives 141 requirements ('misfits' expressed positively) out of 12 verbal constructs (which are only used to help in generating the requirements). For each requirement, the other requirements that are related (positively or negatively) are listed. These linkages were then used to create a three-level tree:
- the overall problem at the top level (superset)
- four sets at level 2
- cattle, bullock carts, fuel
- agriculture production, irrigation, distribution
- communal live (social, industrial)
- private life (shelter, small-scale activities)
- 12 subsets at level 3
A solution for each of the 12 subsets is captured in a diagram, along with a detailed explanation of how the requirements within that subset are related to each other and the proposed solution. The solutions are then consolidated at the second level and finally into the overall village plan.
Although the description of the outcome is very detailed, the process by which the requirements and relationships were determined as well as their decomposition into subsets is not described.
Most articles suggest that pattern languages emerge from identifying patterns in the real world and exploring how they relate to each other. This book proposes a top-down process, which is more compatible with our attempt to use ecosystem principles as a starting point. Alexander describes in detail the pitfalls of 'verbal constructs' that designers use as a short-hand for understanding complex problems. Ecosystem principles may very well be good examples of verbal constructs. Although presented as isolated concepts, they clearly are related in ways that are not explicitly documented. I believe it would be valuable to start with the ecosystem principles that Maibritt explored in her paper, decompose them into as complete and detailed a set of 'misfits'/requirements as possible/practical, and then re-factor these using the process Alexander describes in this book. These subsets could become the basis of ecosystem patterns.
Alexander talks about the growth in diffuse, unorganized knowledge combined with increasingly narrow specialization, points we raised in the first paper presented at SEM. It is increasingly hard to access information. If you can find a specialist, you may not be able to communicate effectively. At the same time, the rate of change is accelerating, tradition and experience are less useful, and there is no time for 'trial and error'.
Alexander emphasizes the importance of boundaries, balance, rates and hierarchies (in both the problem itself and the final solution). The solution must be derived out of a full understanding of the problem and the specific context. This echoes calls by Wendell Berry and others on the importance of 'place'. Alexander makes the task possible by focusing on existing solutions: what works and what does not. He builds on what is there, rather than the more utopian ideal of building a brand new world that is neither anchored in current reality nor leverages proven (if imperfect) solutions.
Lastly, I believe Alexander has created a bridge between human design and the process we see happening in nature. Fil points out that nature does not design, because it is not working towards goals. Yet the outcome often appears to be goal-oriented, something that Schneider/Sagan point out in Into the Cool. The 'unselfconscious' process that Alexander describes seems very similar to how natural selection works. The idea of a homeostatic system that is constantly re-balanced by tweaks and tucks could very well result in the dynamic yet robust ecosystems we see. Alexander argues that the selfconscious process that we associate with modern design not only is incapable of dealing with the complexities around us, but may actually be creating imbalance and chaos. His pattern language approach suggests a way to modify the selfconscious process such that it incorporates the key elements of the unselfconscious process.