Natural Selection vs Artificial Selection
When we talk about evolutionary design, it’s important to distinguish between two core mechanisms: natural selection and artificial selection.
Natural selection, as used in evolutionary algorithms, refers to a computational process that automatically favours solutions which best meet a set of predefined, measurable criteria - like minimising structural stress, maximising solar gain, or optimising material efficiency. It’s an objective-driven method, where the system evolves by continuously selecting the fittest individuals, generation after generation.
Artificial selection, on the other hand, introduces a very different kind of intelligence into the process - human judgment. Just as breeders of racehorses or greyhounds carefully choose which animals to crossbreed based on experience, instincts, and desired traits, a designer can intervene in the evolutionary loop to select solutions based on aesthetic intuition, cultural meaning, or spatial quality - factors that are difficult to quantify.
In computational design, this process is often implemented through Interactive Genetic Algorithms (IGAs). While the algorithm handles the generative engine, the designer becomes the curator - choosing which options to keep evolving, not just because they meet the data-driven goals, but because they feel right.
This hybrid process - blending machine-driven natural selection with designer-led artificial selection - is what makes evolutionary algorithms such a powerful creative partner. It allows us to move beyond pure optimisation and into the realm of meaningful design exploration.
We’re not just searching for ‘the best’ solution. We’re shaping a narrative. And in that narrative, the human voice still matters.
Case Study: When Intuition Overrides the Algorithm
Here’s an example of a project where artificial selection played a critical role in shaping the outcome.
The Project: Return of the Marshes
This project was part of a larger investigation into ecological urbanism, where we explored architectural interventions that could regenerate the ancient wetlands of southern Mesopotamia. The goal wasn’t just environmental - it was also cultural, symbolic, and spatial. We were working at the intersection of ecology, heritage, and architecture.
To help navigate the vast solution space, we employed an evolutionary algorithm that generated variations of landscape formations and built interventions based on parameters like hydrological flow, shading potential, accessibility, and topographic continuity. These were measurable, and the algorithm did a decent job optimising for them.
But there came a point when the algorithm’s ‘fittest’ solutions - those that technically performed best - felt too sterile. They missed a key quality we were looking for: a narrative expression of place. They lacked emotional resonance.
Artificial Selection
At this stage, we paused the process and reintroduced the human eye. We began curating the population based on spatial rhythm, symbolic alignment with Mesopotamian geometries, and the emotional tone each outcome conveyed - none of which were encoded in the fitness function.
In one generation, we deliberately selected a set of underperforming options because they carried visual and cultural cues we felt were vital: branching patterns that subtly echoed ancient canal networks, or edge conditions that mirrored historical wetland settlements. For their construction, we selected a weaving pattern that echoed their reed construction technique. These selections seeded a new evolutionary cycle - guided by intuition and context, not just computation.
Why It Mattered
That moment of artificial selection transformed the project. It reoriented the algorithm from blind optimisation to culturally meaningful design exploration. It’s a reminder that in computational design, the algorithm isn’t the author - it’s a tool. And sometimes the most powerful thing a designer can do is listen to their intuition and interrupt the machine.
