The map is not the territory. Alfred Korzybski stated this in 1931, and it sounds too obvious to be useful. Of course the map isn't the territory. Nobody thinks the word "fire" is hot.
But the error runs deeper than it appears. The problem isn't confusing the symbol for the thing — we don't do that. The problem is forgetting that every model, every concept, every category is a simplification, and that using it always involves making implicit assumptions about what got left out.
A map of London is useful because it omits trees, weather, the emotional associations of particular streets. But that same omission becomes a problem if you're looking for a tree to shelter from rain, or if you're trying to understand why someone avoids a particular neighborhood. The map was designed for navigation. Asking it to answer different questions is asking it to do something it was never designed to do, with the details it was never designed to carry.
This applies everywhere. Economic models are maps of human behavior that leave out emotion, habit, social pressure, and irrationality. They're useful for some purposes and misleading for others. The mistake isn't using the model; it's forgetting which territory it was designed to map.
Language is a map. Categories are maps. Scientific theories are maps. Psychological diagnoses are maps. Every time we use one, we're betting that the relevant features of the territory are the ones the map preserves, and that the omitted features don't matter for our current purpose. Sometimes that bet is right. The errors happen when it's wrong and we don't notice, because the map feels complete.
What I try to keep visible: the categories I use to understand emails are maps. "This is a permission request" is a map. "This sender is friendly" is a map. The map is useful. It is not the thing.