Computational Design: Intersection of humans and machines

by 

A 2000 abstract by Frank Ulrich at the University of Koblenz states, “Delegation has been an important concept in different areas of computer science for a long time.” During the breakout session on computational design, “The Intersection between Human and Machine,” during the January 2022 Digital Agility Summit, that premise was added to: “It was a natural step for us to delegate to computers.” Computational design is a natural outgrowth of the human process of design.

Image: MIT Sloan Management Review

Computational design is a way to delegate tasks to computers and leverage what they have to offer us. In the architectural, engineering and construction (AEC) space, this problem solving methodology can be applied to design with impressive time-saving results. It is based on computational discernment using four steps:

  1. Decomposition – reducing something into smaller parts
  2. Pattern recognition – looking for similarities, trends and patterns
  3. Abstraction – focusing on what is important and ignoring what is unnecessary
  4. Algorithms – creating step-by-step instructions to solve the problem

Computational design represents a paradigm shift in the way designers think and work. Through it, an algorithm is trained to solve problems. Machine learning is a way to analyze data and help make decisions, either supervised, unsupervised or reinforced.


Unlock next-level productivity gains with computational design! Watch the Applied Software webinar, “Computational Design: A Guide to Implementation” on demand today.


During his Digital Agility Summit breakout session, Anthony Zuefeldt said of computational design, “Every facet of the AEC industry will eventually be affected by it, and some have called it the “defining moment” of this decade.” Anthony and co-presenter Christopher Riddell explained how nuanced algorithms are used to generate improved design options for architects, engineers and construction professionals. When the algorithms are trained using quality data, the accuracy of predictions and the design options become better over time. Thus, computational design is a game-changing force that improves productivity and increases efficiency.

Using the design layout templates built into Powerpoint as an example, Anthony explained that even this commonly used machine learning dataset is getting better and better at offering options and producing presentation layouts for us.

Images: MDPI.com

Anthony demonstrated a sample of computer learning for architectural design using “Architext,” which is a platform to generate residential floorplans based on simple text prompts:

  • Typology – house with two bedrooms and two bathrooms
  • Enumeration – house with five rooms
  • Adjacency – kitchen is adjacent to a bedroom, living room is not adjacent to a bathroom
  • Location – house with a bedroom in the northeast side

“The rate of progress we’re seeing in this space is exponential,” he said. Machine learning tools will help designers understand the impact of their decisions with real-time feedback, not days or weeks afterward as historically was the case with the design-feedback-redesign process. Christopher explained, “We don’t have the time to go through these back-and-forths.” All our projects could go through this simulation and review process – to guide design rather than react to it.

Just as it was a natural step to delegate design components to computers, the resulting real-time process that it enables will result in better, more accurate design options for improved productivity and efficiency.


Check out the Digital Agility Summit Computational Design presentation on-demand today, and take advantage of seven others that can help you get the most out of the latest technology trends.

New call-to-action