Style is a broad term that could potentially refer to any features of a work, as well as a fluid concept that is subject to change and disagreement.The idea of a style in any discipline is a fluid concept that is always subject to change, and therefore suited to a flexible representation. What is suggested here is that it can nevertheless be accurately represented and emulated. This work has presented an algorithmic method for both deriving a stylistic definition automatically from examples, and using it to generate new designs. Architectural examples were used, and were investigated primarily in terms of their spatial features, but it is intended as a general model in that other forms of input and classification algorithms may be used. Likewise, axial analysis and the aggregation model are not essential to the method, but the principles of feature space reduction and archetype should apply to a variety of analysis and synthesis techniques.
The concept of the archetype proposed is of a defined ideal and of a space in which to measure example designs. It contains only the features most relevant to define that style, but they are not counted as symbolic wholes. Instead one can measure an example’s similarity in degrees, on an objective and continuous scale. This results in a definition of style that is flexible, can evolve, and is based on examples. While fixed, rule-based systems are used as design aids by generating new examples of designs, a flexible, example based method such as this would assist in a very different way. While the archetype may be resistant to symbolic description, so very often are our own mental processes of style recognition, and in many complex problems we can more easily communicate by example than by explicit description. By automatically generalising its representation based on examples presented to it by a designer, such a design aid may propose output based not on rational clarity of process, but on the simple choice of precedents, fashion, taste or a hunch.
The definition of style provided by the archetype is analytical rather than generative, but there is still an obvious role for generative systems to play. The aggregation model iwas chosen for its simplicity and common origin with the analysis in the previous section, but shape grammars and other generative rules could be applied – a likely avenue for future exploration. Their role in this regard however, is as a framework for exploration of many styles rather than a definition of one.
Creative design is ultimately not a matter of rule following, but of judgement, and the model presented here proposes the flexibility this implies may extend to the definition of styles themselves.