Why Model-Based Science Does Not Mean We Live in a Fake or “Matrix” World

April 21, 2026 |

Science uses models to understand the world. A model is a simplified way of describing what we see so we can make predictions. It does not include everything. It leaves things out on purpose.

We never deal with the full world all at once. Instead, we compare what a model predicts with what we actually observe. If the predictions match what we measure, the model works. If they do not, the model fails.

This leads to a key idea: a model is useful when the details it ignores do not change the result very much. That is what makes simplification possible. But it also means not all simplifications are valid.

There are many possible models we could imagine. Most of them do not work. They fail when tested. Only a small number survive because they keep matching observations across different situations. This shows that models are not just invented freely. They are filtered by something outside them.

That “something” is the world itself. It does not allow just any description to succeed. It pushes back. It forces models to stay consistent with what we measure.

Another important point is that what we call an “object” depends on the model. In one model, air is treated as a smooth flow. In another, it is a collection of particles. These are different ways of describing the same system. The “object” changes with the model.

But even though the descriptions change, some patterns stay the same. These stable patterns are what matter most. They show up across different models and across different conditions. This is what gives us confidence that we are describing something real.

A model fails when it removes the wrong details. If you oversimplify, you lose the structure that actually matters. For example, if you describe traffic as a smooth flow, you miss traffic jams and sudden stops. The model looks clean, but it gives the wrong answers.

This shows that there is no guarantee a model will work. It has to be tested. It has to survive repeated checks. Most models do not.

Now consider the idea that we might be living in a fake or simulated world.

If the world were random or inconsistent, stable models would not exist. Science would not work. But science does work. We are able to build models that keep giving reliable results.

If the world were a perfectly controlled simulation, then it would still have to produce consistent patterns across all tests. In that case, it behaves like a real system in every way we can measure. The difference between “real” and “fake” would not change how anything works.

And even in that case, the constraint would still remain. We cannot choose any model we want. Only certain models continue to work. That limit does not disappear just because we change the label of the system.

So the key point is this: model-based science does not mean reality is fake. It shows the opposite. It shows that not everything is allowed. Only certain models survive, and they survive because they match something stable in the world.

What we call “real” is not every detail we can imagine. It is what stays consistent across different models and different tests.

We do not start with full access to reality. We approach it by building models and seeing which ones hold up. Over time, what remains is what cannot be removed without breaking the results.

That is not illusion. It is constraint. And that constraint is not something we invent or choose. It is what resists change across every model that works. Even if the underlying layer were different from what we think, that resistance would remain. And that resistance—what cannot be varied without failure—is the closest thing we have to what is real.

— Ardan Michael Blum

In this series: 

• What Science Leaves Out to Make It Work

• Newton, Gravity, and the Structure of Physical Description