On Useful Lies and Forgotten Uncertainties
Why All Models Are Wrong (But We Need Them Anyway)
Jorge Luis Borges, the Argentine writer, once imagined an empire where cartographers had taken precision too far. In their pursuit of perfection, they drew a map so vast it covered the empire itself—point for point, recorded in obsessive detail, overwhelming and exact. The world lay flattened into paper. The map had grown to match the empire—a literal rendering of reality itself.
Of course, the map was of no use. Just a monument to precision without purpose. Imagine a map of Buenos Aires so detailed it included every street, every alley, every crack in the pavement, every uneven tile—a map as physically large as the city itself. Who would even bother unfolding it?
This is the subtle paradox of all models.
Take the Earth itself. From our earliest days in school, we meet it as a sphere—a smooth globe on a desk, a neat circle drawn in chalk. It is not, exactly. The real planet is misshapen and restless: it swells at the equator, flattens at the poles, and wobbles subtly as it spins. Its surface is rough and uneven, marked by mountains and trenches, its shape slowly rewritten by tides and tectonic movements. No simple geometry can fully contain it.
And yet, we keep calling it round.
Not because this is perfectly true, but because it is useful. For most purposes, the simplification works beautifully—for navigation, for astronomy, for weather forecasting, for drawing flight paths. Here, the lie works better than the truth. Each time we model the Earth as a smooth globe, we introduce a controlled distortion, erasing complexity, hiding imprecision in the name of functionality. A perfectly accurate Earth would be as useless as Borges’ empire-sized map.
Simplification, however, is only one way models become usable. Sometimes, instead of subtracting reality, they quietly invent it.
Let’s look now at a different map. A real one this time. You open your phone and scroll through a digital map of Edinburgh. Then one name catches your attention: Oxygen Street. You imagine walking there, exploring the corner cafés, peeking down the alleys.
But Oxygen Street does not exist. It was invented by cartographers long ago, a small fiction tucked into the folds of a map to catch anyone who might copy it—a copyright trap. Here, the map does not merely simplify reality. It creates reality that isn’t there. And yet this invention does not break the map. It is still a good model to guide people through Edinburgh. One false street does not undo the rest.
These examples—the Earth, the fictitious street—are benign, almost quaint. But some models deal with far more complexity. Consider the weather. Weather forecasts feel different from maps. A map claims to show what is. A forecast claims to show what will be. And yet both are models—reductions of reality into something usable, navigable, human-sized.
Behind every simple weather icon—sun, cloud, rain—lies a machinery of equations so vast it might rival Borges’s empire-map in ambition. Meteorologists do not predict a single future; they predict many, each with its own probability. Snow might fall. Ice might form. Winds might veer. Each forecast carries uncertainty.
Yet the interface rarely shows this. We see icons, not distributions. Certainty, not spread. In many newspapers, apps, and online feeds, probabilities are often collapsed into proclamations, often delivered with dramatic flair: snow will bury cities, ice will glaze streets, winds will topple trees. The result is overconfidence, misjudgment, sometimes even chaos.
The forecast works—often remarkably well. Like Borges’s map and Oxygen Street, it succeeds not through perfect accuracy, but through usefulness.
We build models of economies, climates, cities, diseases—each necessary, each simplifying, each incomplete. The problem is not that they are wrong. It’s that we sometimes forget they were never meant to be exactly right. Models aren’t broken versions of reality. They’re useful compressions—tools meant to guide, orient, and suggest, not to exhaust the world. A perfectly faithful representation of reality would be indistinguishable from reality itself, and just as impossible to navigate.
And yet—here’s the strange part—they work. A misshapen planet treated as round still carries aircraft across oceans. A city map with a fictional street still gets people home. Equations describing turbulent air still warn us when storms are coming. These aren’t failures of modeling. They’re small miracles of abstraction—ways of turning a world too big to hold into something we can move through.
But oftentimes a false precision creeps in. Numbers acquire spurious authority. Outputs harden into conclusions. The very tools we built to manage uncertainty start pretending it isn’t there. When approximations are treated as facts and probabilities as certainties, we do not gain knowledge. We lose nuance and judgment.
Models don’t show us reality as it is. They show us reality as we can use it. The task is not to demand perfect models, but to remember their limits—the uncertainty they compress, the messiness they smooth away. And yet, alongside this humility, there is something else entirely: wonder. That a world so vast and unruly can become intelligible, navigable, and actionable through representations that are, by design, incomplete.
Read More:
Book: How to Lie with Maps
Short story (original Spanish): Del Rigor en la Ciencia
Short Story (English translation): On Exactitude in Science