the limits of a model
Every model is a deliberate lie that tells a useful truth. To build an equation you throw away most of reality — friction, weather, the messy individuality of things — and keep only what matters for your question. That is what makes a model solvable and illuminating. But it also means every model has an edge, a boundary beyond which its simplifications stop being harmless and start being wrong. Knowing where that edge lies is as much a part of modeling as writing the equation.
The limits show up in three honest ways. First, the assumptions: the exponential growth model assumes unlimited resources, so it fails once crowding sets in; Newton's cooling assumes one uniform temperature, so it fails for a thick roast; linear air drag assumes slow motion, so it fails for a fast skydiver. Second, the domain of validity: a model trustworthy over an hour may be nonsense over a year, and one good near an equilibrium may mislead far from it — extrapolating beyond the tested range is the classic blunder. Third, what was left out entirely: randomness, delays, spatial variation, feedbacks the model never included can dominate the real behaviour.
The mature attitude is not to distrust models — they are how we understand and predict the world — but to use them with their warranty in mind. State your assumptions out loud, check predictions against data, watch for where they break, and reach for a richer model when the simple one fails. A model is a map, not the territory: indispensable for finding your way, dangerous if you forget it is a map.
Exponential growth fits a bacterial colony's first few hours beautifully, then badly overpredicts once nutrients run low — the model's edge is reached, and switching to the logistic equation restores honesty.
A model is a map, not the territory: precise within its range, misleading beyond it.
A good fit to past data does not guarantee good predictions outside that range; the most expensive modeling errors come from extrapolating a model far past the conditions it was validated against.