Systems & Hidden Structure
Digital Copies for Testing What-Ifs
Also called: Digital twins and counterfactuals
- Personal interest
- Formal theory
- Working interpretation
A useful digital twin, a computer model kept in step with a real system, explores other possible futures and helps compare changes before anyone makes them. It should do more than animate a familiar factory. I want to know whether its behavior stays trustworthy when conditions change.
Which what-if question would teach me the most?
The idea
A digital twin is a working model of a real system, fed with data from it. The interesting use is not watching the copy run. It is asking what would happen under a different choice. Such imagined alternatives are called counterfactuals: statements about what would happen, or would have happened, if something were different. A twin makes some of them testable, within the limits of the model.
What I think (and don't know)
A model is not the system (Models Are Not Reality). A twin rests on assumptions about how things worked when it was built, and it can keep giving confident answers after they stop being true. So I care less about how real a twin looks than about how, and how recently, it was checked against reality. The engine underneath is usually simulation (Simulating a System Event by Event).
Where it connects
I use the same habit outside engineering. Imagining the versions of myself who chose differently (The Selves I Might Have Been) is also a kind of what-if model. It can clarify regret, gratitude or ambition without claiming those other selves exist. Brains may do something loosely similar when they predict what comes next (The Brain That Predicts What Comes Next), though the two have different aims and very different evidence behind them.
An example
Imagine a factory deciding whether to add a second shift on one line. The twin replays a month of orders with and without it, and shows that the extra shift helps only if a test station further down the line also gets more hours. That result is a reason to look closer, not a guarantee.
What this does not establish
A digital copy is a model, not the real system. Its what-if futures are imagined comparisons: they do not show what will happen, and they are not real places or branches.
Questions I'm still exploring
- How do I check that a model still matches reality after conditions change?
- Which what-if questions can a model answer well, and which are beyond it?
- How should a model show its uncertainty so people do not over-trust a smooth picture?
Sources and further reading
- Michael Grieves and John Vickers, "Digital Twin: Mitigating Unpredictable, Undesirable Emergent Behavior in Complex Systems," in Transdisciplinary Perspectives on Complex Systems (Springer, 2017)
Working interpretation: drafted from my notes and interests for review. It is not a direct quotation, and I may still change it.