Alexander, C. (1964). Notes on the Synthesis of Form.
Status: Read
Superficially, this is a book about design process. What constitutes good design, and how to approach design problems systematically. On a deeper level, this book outlines the process of describing and solving complex system level problems. Alexander formalizes—to the extent possible—the process of form-making. He develops mental models to standardise the problem description, and the method to derive form from that description.
A design problem is solved when there is good fit between the form and its context. In Alexander's formalization, the form is the solution to the problem; the context defines the problem. When we speak of design, the object of discussion is not the form alone, but the ensemble of the form and its context. Fitness then, is a property of the ensemble that describes how well the form fits the context. Ensembles, by this definition, encompasses a range of form-context pairings beyond the typical conception of design: organisms and their environments; specific chess moves and the stage of the match; specific musical notes and the context of a composition.
The form-context boundary is malleable. Part of the form may be redefined into the context and vice versa. In the design of a kettle, the form-context boundary seems clear cut: the kettle, and everything outside the kettle. But if we consider the problem of heating the domestic water supply more broadly, the kettle is now just one part of that form — we've pushed back the context to also include the hot water plumbing. Alternatively, we may decide that it's not the kettle that needs to be redesigned, but the method of heating kettles. In which case, the kettle now becomes a part of the context to be designed for.
The form-context boundary is fractal. The overall ensemble is a collection of nested, overlapping form-context boundaries. Its fitness relies on the stability of its internal structure, and the fitness of its constituent parts. In a perfectly coherent ensemble, we should expect the two parts of any division to complement each other. Designers need to be keenly aware of the changes in each part of the ensemble and to manage multiple form-context boundaries in concert.
Form-context fit is described via negativa — from the negative point of view. It is easier to identify and describe bad fit than it is to describe good fit. Good fit is simply the absence of bad fit. It is almost impossible to characterize a house which fits its context. Yet, it is easy and immediately obvious to name specific kinds of misfits: the rainwater coming in, the front door that visitors cannot find, or the child playing where it can be run over by a car. In this way, design is an error correction process to neutralize the incongruities or forces that cause misifts.
By Alexander's formalization, Form-context fit is binary: there is only fit or misfit. A design problem is not an optimization problem. For most requirements it is sufficient to satisfy them to a level which prevents misfit, or what Herbert Simon calls "satisficing," which is a more accurate description than "optimization" of how we make decisions in complex situations. Only one or two variables, such as cost or time, need optimization. The designer's task then, is to bring each binary variable to the value 0 (fit). It is therefore only important that we specify each variable such that any design can be classified unambiguously as either fit or misfit.
Form-context fit can be found theoretically or experimentally; either by putting forms in contact with the context or by following a complete description of the context. This is best understood by way of example. Suppose we are to invent a stable arrangement of iron fillings in a given magnetic field. The arrangement of iron fillings constitutes a form, the magnetic field the context. A fit form can be achieved by turning on the magnetic field and seeing if the fillings move. Alternatively, a form could also be derived mathematically as we have a complete description of the magnetic field via Maxwell's equations. In this case, the field is simultaneously a description of both the context and the form. In general, however, the theoretical approach is only possible in a subset of design problems where there is an adequate description of the context.
Design is a process of trial and error. It is a process of strategically searching through various forms of the solution-space, rationally discarding those that don't "satisfice" the requirements (see Trial and error is antifragile). For complex problems, a large sample size of candidate forms is therefore necessary—even for experienced designers—to achieve good fit across all form-context boundaries. Related: Learn by doing.
The use of logical structures to represent design problems has led to "a loss of innocence" in design. A logical picture is open to reasonable critique, as its premises are describable and unambiguous. Once assumptions are disclosed, intuitive and non-intuitive methods of problem solving can be described and compared. Problem solving through pure intuition then, can no longer be innocently defended. This stands counter to what many designers believe: that the design process is fundamentally intuitive, and it is hopeless to try and understand it logically.
Alexander's contention is also in contrast to the arguments of Nassim Taleb, who has named this kind of error of rationalism, the lecturing birds how to fly effect. In short, we tend to overestimate the role and necessity of narratable knowledge — knowledge that we acquire in school, can get grades for, and what is explainable — over knowledge that is non-narratable, yet that we can do well. Just because a method cannot be expressed in clear language, does not mean the method is inferior. We do not need to know how to explain bike riding to be able to ride a bike.
In trying to understand the characteristics of design processes that achieve successful fit, Alexander draws a distinction between unselfconscious and self conscious design cultures. The two cultures can be distinguished along three main dimensions