Hybridizing historic and emerging technological paradigms to create speculative tools for architectural design
The remaking of obsolete computer programs for interactive manipulation in the present is known as “software reconstructions” (Cardoso Llach and Donaldson 2019) or “algorithmic reenactments” (Galloway 2021). These approaches entail the design and implementation of machines that, through their design and interface, walk their users through the steps and functions of historic computer programs and algorithms. Our research on speculative reenactments builds on these approaches but places emphasis on critical speculation. Its originality lies in the hybridizing of past and new technological paradigms (such as generative artificial intelligence) to create machines that enact a form of historical didactics while being playful and critically evocative. By conceptualizing this work as a reenactment we place emphasis on computation as process and performance. The machines we build render processes visible but also allow us to rethink those machines' spatial expanse; the bodies that move them and which they move.
- Hierarchical Decomposition System 2 [1962 >> 2024]
Our first speculative reenactment is HIDECS 2 (Alexander and Manheim 1962). Written in FORTRAN and ran on the IBM 709 of the MIT Computation Center, the system “decomposed” large lists of design requirements into a hierarchy of smaller requirement groups by examining relationships (“interactions”) between them. It computed for designers a sequence by which to address and “recompose” their design responses to these smaller groupings of requirements. We have designed and built a physical prototype for the hardware, coded the FORTRAN algorithm of the 1962 program in Java, and developed a speculative custom workflow that automates the “recomposition” process through the ChatGPT and DeepAI APIs. This workflow translates the decomposed groups of verbal statements (“design requirements”) into stylized generated images (“diagrams”) and interactively combines these diagrams to produce a final image of a design.
- Planning of Single-Storey Layouts (1964 >> 2024)
We are also working on a reenactment 1964 Algol-based program for automated space planning took that was implemented in an English Electric KDF9 computer (Whitehead and Eldars 1964). The program took as input observed patterns of movement between fixed locations on an existing building and generated a new single-storey floor plan that was optimized for movement. We coded the algorithm in Python and developed two custom-made grid-based 100-cell “screens” for data input and output using photosensors and LED lights respectively. The screens display a schematic version of a floor plan. "Users" can click on cells on that plan to input arrival and departure points of an imaginary occupant moving on that plan. The “output” screen then automatically updates to display and optimized version of that same floor plan. We are working on the design of a performance that reenacts the data collection process.
News & Publications
