Hello Fellow Magicians,

What is a chair? Seems like a simple question. But as with most seemingly simple questions, it can quickly become complicated.

I once asked this question to interior designers at work, and I quickly learned how nuanced it can be.

Is a bar stool a chair? Hmm. Many would argue no—it’s a stool. A stool, they say, lacks a back, while a chair does not. But does that mean all seating with a back qualifies as a chair? Not quite! Someone might chime in: couches aren’t chairs. So what’s the difference? Couches can seat multiple people.

Ah, we’re getting somewhere. Chairs, then, are single-occupant seating furniture with a back. Does that definition hold up? For the most part, yes.

Now, you might be thinking, “This sounds like an annoying conversation.” And you’d be right. But here’s the thing: conversations like this are essential if you want to succeed in building machine learning capabilities in any industry—especially in AEC.

Ontology: Defining the Essence of Things

This process of working through the “essence” of names, labels, and classifications has a long history in philosophy—specifically, in ontology.

Ontology is the study of being. It is a sub-field of metaphysics - and primarily concerns itself with puzzling through categories of being. Why are sets of things more similar or more different? When are two items the same on a level of analysis and when are they different. The goal of Ontology is to discover the foundational building blocks of the world and characterize reality as a whole in its most general aspects.

This might sound quite lofty - but even if that primary goal is not met - the thought processes inherent to ontological study are highly valuable - and required for developing great Machine Learning and AI systems.

From Philosophy to Practicality

At first glance, ontological questions—like “What is a chair?”—seem abstract or even tedious. But these questions are foundational when building systems that need to interpret the world, particularly in Machine Learning and AI.

Think of it this way: machines don’t have intuition (in fact they don’t have anything). They don’t “know” what a chair is unless we teach them. And we can’t teach them without first agreeing on the boundaries of what a chair is and isn’t. This process—defining the essence of things and their relationships—is ontology.

When you’re developing a model to identify objects in a building, you’re enforcing an ontology, whether you call it that or not. You’re defining a conceptual inventory:

• What counts as a room? What types of rooms are there - and aren’t there

• Is a window part of the room, or an adjacent feature?

• Does a doorway belong to the room or does it connect 2?

These aren’t trivial questions. Your answers shape how the model interprets its input data and how it performs in the real world.

It’s important to know that you can pick different answers for different systems or models, but you can only have one definition for one item for each single application.

Why It Matters

Ontologies aren’t just philosophical—they’re practical tools for creating clarity out of complexity. In the world of ML and AI, clarity is everything. When you define your problem space well, you reduce ambiguity, improve model accuracy, and set a foundation for scalability.

So the next time you want to explore a new ML capability or invest in an AI tool - make sure you have done the important Ontological work first. You need to know what you mean before a computer can know what you mean!

So before you start developing your AI system - take out a pen and paper - get some experts together and have a discussion on Ontology - no matter how dumb it feels when you start arguing over chairs.