IGAs in Architecture and Design: Enter Snowflake

Today, computational design environments - like Grasshopper for Rhino - enable architects and designers to create complex systems of parameters and relationships. You’re not just drawing lines or shapes; you’re defining rules and behaviours that generate form. This is the foundation of parametric design. 

But while parametric tools let us define ‘how’ things behave, they don’t help us decide ‘what’ is good. That’s where optimisation tools come in. And when it comes to subjective or hard-to-quantify goals, this is where IGAs truly shine.

Enter Snowflake: A Designer-Centric Evolutionary Plugin

One tool that embodies this designer-algorithm interaction is Snowflake - a plugin developed to work within parametric environments like Grasshopper. Snowflake is built on an evolutionary algorithm (SPEA2) but adapted for subjective evaluation.

Here’s how it works:

1. Define a System: The designer sets up a parametric model - say, a tower with sliders controlling geometry, openings, or other configurations.

2. Run Generations: Snowflake generates a population of design variations using the underlying genetic algorithm.

3. Designer in the Loop: Instead of letting the system choose based on fixed numbers, the designer evaluates each generation - selecting forms based on intuition, goals, or aesthetic preferences.

4. Feedback Drives Evolution: These choices become the ‘fitness’ for the next generation. The algorithm evolves its search based on what the designer prefers - refining results without needing to quantify beauty or spatial experience.

This is designed as dialogue - the machine proposes, the designer responds, and the loop continues.

[Show Snowflake diagrams]

Why IGAs Matter for Architects and Designers

In architectural design, optimisation can’t just be about minimising energy or maximising daylight. It must also accommodate ambiguous goals: identity, cultural expression, spatial character, and material expression. IGAs allow for this ambiguity.

They do not replace the designer. Instead, they amplify the designer’s agency, surfacing possibilities that may never have been imagined manually, while still grounding the evolution in human intuition.

Snowflake, for example, becomes a kind of creative collaborator - not by making decisions on its own, but by inviting the designer’s judgment into each step of the process.

What Makes This Powerful?

• Exploration with Control: Designers can guide the search without knowing exactly what they’re looking for. It’s about steering a direction rather than pinpointing a target.

• Subjectivity as Data: Your preferences, instincts, and gut reactions become the algorithm’s learning material.

• Emergent Discovery: Unexpected solutions arise, which often challenge assumptions and expand creative boundaries.

A Glimpse Ahead

IGAs open the door to a new kind of authorship - collaborative, recursive, and informed by difference. As tools like Snowflake evolve, we’ll likely see even tighter feedback loops, where real-time interaction and aesthetic evaluation form the backbone of generative design systems.

In the next post, I’ll explore how IGAs help balance performance and poetics - showing that optimisation doesn’t have to mean compromise.