The Map Is Not the Territory

Every model is wrong. Some are useful. Understanding the difference between a representation and reality is the foundation of clear thinking.

Alfred Korzybski's Insight

The philosopher Alfred Korzybski coined the phrase in 1931. His point: the word "dog" is not a dog. The concept "inflation" is not the complex set of economic phenomena it attempts to describe. The symbol is not the thing. The model is not reality.

This sounds obvious. And yet we constantly confuse the two.

How We Confuse Maps and Territories

Mistaking the label for the explanation. Naming a phenomenon can feel like explaining it. "He did it because he's impulsive" describes a pattern — it doesn't actually explain why the pattern exists or predict when it'll recur.

Reifying abstractions. GDP is a measurement convention, not a force of nature. "The market" is a shorthand for millions of individual transactions, not a sentient entity with intentions. When we talk about these abstractions as if they were physical objects with causal powers, we've mistaken the map for the territory.

Over-fitting to past data. A mental model built on yesterday's territory may misrepresent today's. The world changes; models often don't update fast enough.

Why This Matters for AI Literacy

Language models are, in a very literal sense, maps. They are statistical representations of patterns in text — not the underlying reality those texts were about. When a model generates a confident-sounding claim about a factual matter, it's producing output consistent with patterns in its training data. It may be right because the map happened to match the territory; it may be wrong for the same reason.

Knowing this helps calibrate trust appropriately. Not paranoia; calibration.

Using Maps Well

A good map-user knows:

• What the map was designed to show (and what it omits)
• At what resolution the map is accurate
• When the map was last updated
• That the map's boundaries aren't the territory's boundaries

These are the right questions to ask about any model — mental, statistical, or economic.