Navigating the Generative Possibility Space

Finding your place in the GPS

Lynne Yun coined the term Glyphzian Space to mean the space that encompasses everything surrounding the digital representation of letterforms. From visible forms to Bézier curves to file data structures, there are layers to traverse and territories to explore in what appears to be a solved system.

Working with generative systems, I’ve found there’s a similar territory to map: the generative possibility space—the total set of outputs that a generative system can create.

Let’s say our system is a set of wind chimes. The possibility space includes every combination of notes that could sound as wind moves through them. Soft breezes create sparse melodies and strong gusts create dense clusters. It encompasses all the timing variations, amplitude differences, and everything the system is capable of producing.

Now compare that to something like my Throwies algorithm. The possibility space includes every configuration of sine waves, color combination, and variation in amplitude and frequency that the system allows. It’s a different kind of space—more constrained in some ways, infinite in others.

The size and shape of possibility space varies wildly between systems. But once you have a system that can create infinite outputs, you face a question: How do you explore it?

In our current UI paradigms, I see a few approaches:

Sliders - 1-dimensional. Easy to understand and interact with. You select two opposites and put them at the ends with a toggle for people to decide where they’re leaning. Simple, but limited to exploring one axis at a time.

Axes - 2-dimensional. A combination of two sliders creating four quadrants. You can see relationships between parameters—how increasing density interacts with color saturation, for instance.

Cubes - 3-dimensional. Great for visualizing one’s location in space. Unlike 1D and 2D representations, it requires more design consideration to make it usable. But when done correctly, it can make a system easily understandable.

And that is the key in all of these: finding a way to represent your current state relative to the system overall while also making it easy to select and navigate to another point within the generative possibility space.

This is the central tension—orientation and navigation. You need to know where you are in the space (what makes this output different from others), and you need a way to move through it (how do I get to something more chaotic? more structured? more blue?).

The dimensionality of your interface shapes what you can discover. A single slider might never let you stumble onto the interaction between two parameters. A 3D space might hide combinations that a well-designed series of 2D views would reveal.

For me, the question isn’t whether we can map the entire possibility space, but which slice of it we choose to make visible? And what remains forever out of reach because of that choice.

-Mello