Complex systems and the inability to predict their behaviour

We have spent a very long time attempting to resolve weather patterns to a level where we can expect what will happen. Yet our models need systematic intervention regularly to maintain accuracy — and even then, they’re unable to predict what will happen consistently.

I’m sure you’ve woken up one day expecting the weather to behave in a certain way, and they were faced with a different weather pattern. This is the butterfly effect in a complex system, where we do not fully understand all the inputs and outcomes, nor do we have the ability to resolve (watch/track) all the information.

In chaos theory, the butterfly effect is the sensitive dependence on initial conditions in which a small change in one state of a deterministic nonlinear system can result in large differences in a later state.

The butterfly effect works like this: a butterfly flaps their wings somewhere in Asia — somehow, it triggers a snow event to crystalise in North America.

Or, as Edward Norton Lorenz said, “When a butterfly flutters its wings in one part of the world, it can eventually cause a hurricane in another.”

It's almost like a tiny domino knocking over something substantially more significant than itself — it doesn’t do it directly, but the small event triggered the larger one.

This is why we need to have constant model intervention as time actually progresses to update the model to the current state of patterns so it can play out, as best it can with the limited inputs and rules it has, what it believes will happen in the future.

Meaning, the further into the future a model predicts the less likely it is to be true.

All of this is a preface to tell you that you should not believe anyone who comes to you with a model and tells you that its predictions are an accurate picture of the world.

Unless you know what you’re reading, how the thing worked entirely, the model is just an elaborate story that a computer has had a hand in developing the outcome of following a base set of rules combined with inputs.

A model is about as useful as Minecraft in estimating what will happen in reality if you’re dealing with a complex system.