The Oddity of “Designing Space”

The notion of designing space is a peculiar one. After all, isn’t “space” just the emptiness around objects or within containers?

Intuitively, we may speak of space as something tangible we can directly manipulate; after all, we deal with it every moment of our waking lives. But in our everyday physicality, we instead end up manipulating things like walls and furniture directly, all with with the intent of configuring space. To make a room, you have to put up some walls, so you don’t create a room, you create walls, and end up with a room.

To an architect, designing space might mean deciding the massing of a building or laying out programmatic elements like rooms, hallways, or courtyards. To a roboticist, it might mean arranging objects in a warehouse intended as obstacles for robots to train against, identifying zones, or drawing paths. I’d argue that a technologist might design a social network or knowledge base as a kind of “space.” And this doesn’t even account for the added complexities of time.

Despite the ambiguity, “spatial AI” provides theory, concepts, and methods for all of them, to interrogate, model, and speculate on spatial design problems and propositions.

Architectural technologists have been experimenting with generative layout algorithms for the last 20 years, even in complex projects like courthouses, which have several independent circulation flows that must remain separate except in specific intersections, like courtrooms. Will a particular strategy or arrangement of parts really work as intended? Even in this simple question, AI can extend our design decision-making, particularly in predicting how such environments may impact its inhabitants by simulating them.

There are effects from the intrinsic nature of space we may not anticipate. UPS drivers avoid left turns in the US due to such intrinsic properties; left turns are more complicated, costly, and dangerous, just due to the asymmetric nature of our traffic codes, requiring us to drive on the right side of two-way street. This isn’t due to the drivers’ lack of left-turn experience; it’s an outcome of the environment itself.

Simulation as a Key Method

And for designers to begin to understand the impacts of their decisions, building simulations of intelligent environments and agents is the way.

Spatial design in particular is mediated through several layers of media. Architects don’t create spatial experiences directly. While they’re creativity explores a kind of spatial and experiential concept, they act by producing representations like drawings and 3D virtual models, which in turn refer to materials like steel and wood (another medium), which are then put in place by contractors to make buildings (yet another medium), which then are perceived and inhabited by people who experience and interpret the original spatial concept.

This poses a fundamental challenge from a design perspective, one fundamental to spatial disciplines. How might we reasonably validate our design propositions amidst this mediation? And when environments and agents are also intelligent decision-makers, how can we anticipate their emergent behaviors? How can we adjust to changing conditions over time?

I posit that we can create more sophisticated models, not of representation, but of decisions, predictions, and recommendations, vis a vis our definition of artificial intelligence. Such simulacra and simulations can be the comprehensive medium by which we can explore the potential of intelligent environments, the real object of a spatial designers’ goals.