The architecture design studio I coordinate starts the semester with a generative drawing exercise. The first design brief requires students to produce random lines using bamboo skewers or other stick-like materials. The forced process aims to promote quick drawings that ignore composition decisions and allows students to make visual judgements in finding and documenting patterns. These patterns inspire form-making with no reference to interior habitation requirements or external influences; these come later. The only connection to architecture is that these patterns must result in a habitable poche wall.
The act of releasing and then drawing skewers introduces randomness into the design process. Unlike art or music disciplines, architects are often conflicted regarding randomness as it removes design agency. While this is true, randomness in architecture offers little meaning, it does stimulate creativity, break conventional thinking and can suggest unconsidered outcomes.
Although the random line process ends in a drawing, it takes students a long time to release and trace lines into a starting composition. I am always interested in how the students follow the rules when teaching this beginning project; they act procedurally to get to a final drawing. The students don’t realise it, but they are making the way a computer program operates. As students plot and join points into lines, they follow a repeatable function that a computer would easily and rapidly complete.
With this in mind, I abstracted the process into a repeatable recipe for the computer, an algorithm capable of producing different digitally generated patterns. The aim became to explore customised software that produces spatial inspiration.
As Benedict Gross et al argue in Generative Design, producing customised software changes a designer’s analogue performance to “orchestrating the decision-making process of the computer” (Gross et al, 2, p4). Rather than producing one drawing, if the designer abstracts a process into rules and runs these with different inputs, it will result in multiple images. Abstracting and decomposing problems into smaller chunks through computer language engages the computer’s talent for repetition, randomness and logic and produces outcomes that any analogue process would require obsessive dedication.
An initial test was how software could produce random lines across a canvas, similar to the bamboo design exercise. Although the outcome was similar, several opportunities arose. One was that the sketch could become interactive, allowing a user to adjust variables and for their inspiration. Another was that the sketch could output an open file format, such as an svg on png, suitable for future design work. The final idea was that the tool could take on the next stage of the design tutorial process, where students generate figure-ground poche diagrams from the generated line patterns.
The sketch regenerates each time the browser refreshes. The next stage is to explore adjustment interaction through a control panel, using a p5.js plugin and an integrated button to export an svg file for inkscape.