Sketch-Trigger: a Specification for a Form Generator and Design Analysis Toolbox for Architectural Sketching

In order to develop design and decision support techniques in the early sketch design phases, weshould (1) experience and (2) observe real behaviour in practice, (3) transform observations intoideas for improvement, (4) develop behaviour models to explain the sketch design activities and(5) to evaluate between the proposals, (6) decide between the alternatives, (7) implement theselected option in a supporting tool. Our paper reports about the results of step 3 in particular inthe first phase of a Ph D project, started this year. Our main objective is to amplify the effects ofthe sketch as a very effective instrument to generate original forms and to stimulate the mind to discover new shapes and meanings in the roughly sketched patterns. Instead of considering the sketch only as a representation of what the designer has in mind as is usually assumed in CAD systems, we see sketching as form activation. Thus, we want also to offer triggering images to spark off the imagination of the designer while generating images which are practically impossibleto create by hand and certainly not at short notice. The main improvement proposed is the use of an evolutionary form breeding system: one or more sketched parent images (either ready-made'partis'or basic schemes drafted by the designer) presented in the centre of the screen, will generate surrounding mutated children as defined at random but constrained by default or customization of the available transformations. By selecting one or more children a next generation will be produced in the same way. At all times the designer can introduce or reduce constraints. To complete the usually offered'classical'symmetrical, spatial and logical operations,we want to introduce dis-functional operations like dislocation, explosion, deformation, anti-logic etc, in short all kinds of antagonistic operations, among them the transformations applied indeconstructionist and post-modern design. Our expectation is that these operations will correspond roughly to the'move'pertaining to a design entity as the operational unit most appropriate for design behaviour research, in particular the analysis of the chunking and parsing behaviour of the designer. The applicability of the'move'approach has been shown experimentally by Habraken and others. Goldschmidt has abandoned the usual typology approach of protocolanalysis based on moves and concentrated on the linking of moves, but has been hampered by the lack of a good representational instrument. This brings us to the representation of moves and linkages as a research instrument. The'linkograph'approach as proposed by Goldschmidt is a first step towards a graphical representation of the designers associative reasoning mode, necessary for tracking the heuristics of designers at the most basic level, but its practical implementation remained as yet incredibly laborious. What is proposed here is an instrument and approach which makes such registration and analysis possible within a structured software environment