A Precursor to Computation in Architecture – An Alternative to the Current Concept Design Process

Abstract

Architectural design was an iterative and structured yet creative process that required the integration of diverse and complex information. In contemporary practice, diminishing mentorship and fragmented workflows presented significant challenges, particularly during the concept design stage. These challenges hindered the efficient coordination of critical details, such as plasterboard dimensions, which impacted material waste and downstream construction processes. While computational design methods offered potential solutions, their integration into real-world architectural workflows remained underexplored.

This research critically examined the existing “As-Is” process of plasterboard dimensional coordination during the concept design stage to identify inefficiencies and propose an alternative computationally assisted workflow. The study followed an action research methodology, incorporating Six Sigma principles to systematically analyse and improve the architectural process. Through a combination of activity observations, a short questionnaire, and focus group discussions, the research captured real-world practice conditions and collaboratively developed a structured improvement strategy.

The thesis presented several key outcomes: (i) the development of an early-stage plasterboard waste-quantification tool to measure process performance, (ii) a comprehensive mapping of the existing design coordination process and a fully documented alternative workflow, and (iii) a preliminary interaction protocol that defined the roles and interactions between architects and computational tools in the coordination process.

To contextualise computational assistance, the research evaluated both non-ML-based and MLbased computational design methods, comparing their effectiveness in task support, efficiency, and usability. While ML-based approaches showed potential for adaptability, they faced limitations due to inconsistent evaluation standards and reliance on training data. The study highlighted the necessity of aligning computational interventions with existing architectural workflows rather than imposing new tools without understanding process conditions

Through the application of Six Sigma’s DMAIC (Define, Measure, Analyse, Improve, Control) framework, the study systematically investigated the plasterboard coordination process, identifying key inefficiencies related to information fragmentation, reliance on assumptions, and cognitive biases. A refined alternative workflow was proposed, integrating computational support while preserving architects’ control over design decisions. This workflow was further operationalised through a structured interaction protocol that balanced human-machine collaboration.

The research made four primary contributions: (i) introducing Six Sigma as a structured framework for investigating and improving architectural workflows, (ii) providing guidelines for process improvement through a combined methodological approach of action research and ethnography, (iii) documenting the critical information required for effective plasterboard coordination and developing a waste-quantification tool, and (iv) proposing an improved workflow and interaction protocol that aligned computational assistance with real-world architectural practice.

By bridging theoretical insights with practical implementation, this study advanced the discourse on computational design in architecture. The findings emphasised the role of computational tools as facilitators rather than prescriptive solutions, advocating for their integration in ways that enhanced, rather than disrupted, existing architectural workflows. Future research should focus on further testing the proposed workflow, expanding its application to other building typologies and materials, and refining methodologies for evaluating workflow efficiency in practice.