# Digital Twins

> Digital models kept connected to real counterparts so observations, simulations, and decisions can inform one another.

- HTML version: https://robbiepalmer.me/ideas/digital-twins
- Source: https://www.gov.uk/government/publications/what-a-digital-twin-is-and-how-you-can-contribute/what-a-digital-twin-is-and-how-you-can-contribute

A digital twin is a digital model kept connected to something in the world. The counterpart may be
an object, a process, or a larger system. Observations update the model, and analysis of the model
helps people or software decide what to do next.

The UK government's [introductory guide](https://www.gov.uk/government/publications/what-a-digital-twin-is-and-how-you-can-contribute/what-a-digital-twin-is-and-how-you-can-contribute)
uses a strong definition: the model has a two-way flow of timely data with its real counterpart.
This connection supports testing decisions before making them and examining how actions may affect
the world. The guide also sets an important safety boundary. The real counterpart must remain able
to operate safely and securely without the twin.

## The connection makes it a twin

A static model captures a design or a past state. A simulation explores behaviour under stated
assumptions. A dashboard displays observations. Each can be useful, but none alone maintains a
living relationship with a particular counterpart.

A digital twin combines some of these capabilities around a continuing identity. New observations
change its state. Models estimate what cannot be observed directly or test possible actions. When
an approved action changes the counterpart, later observations show what actually happened. This
feedback loop is what separates the twin from a detailed picture.

"Timely" does not always mean real time. A turbine controller and a household budget need very
different update rates. The right interval comes from the decision being made, the cost of stale
state, and the ability to act safely.

## A twin is an interpretation

The phrase "the map is not the territory" applies directly to a digital twin. A representation
remains separate from the thing it represents. It selects features of its subject for a purpose and
can be incomplete, out of date, or wrong. The Stanford Encyclopedia of Philosophy's entry on
[scientific representation](https://plato.stanford.edu/entries/scientific-representation/)
explains how models let us reason about a target while leaving room for misrepresentation.

A twin selects entities, measurements, relationships, and behaviour for particular questions.
[Ontology engineering](/ideas/ontology-engineering) defines those meanings. [Knowledge
graphs](/ideas/knowledge-graphs) can connect observations and models across systems when their
identities and relationships matter.

This selection should stay visible. A confident visualisation can hide missing sensors, delayed
updates, disputed identities, or a model used outside the conditions where it was tested.
[Data provenance](/ideas/data-provenance) lets a prediction or recommendation point back to its
observations, model version, and assumptions.

## Twins can cooperate without becoming one system

Large environments rarely have one owner or one model. The National Digital Twin Programme does
not propose a single model of the whole United Kingdom. Its guide describes mechanisms for people
to bring together the information and models needed for a specific problem.

[Domain-Driven Design](/ideas/domain-driven-design) gives these boundaries a precise vocabulary.
Each twin can belong to a [bounded context](/ideas/bounded-context) with its own model and
ubiquitous language. A context map records how those languages relate. Adapters translate
identifiers and terms at the boundary when the contexts disagree. Only concepts that carry the
same meaning should enter a shared vocabulary.

This federation preserves ownership. Each twin can use the update rate and level of detail its
decisions require.

Connections also create risk. A compromised observation can corrupt the model. A recommendation
can cause physical harm if software applies it without suitable authority. NIST's security and
trust report treats instrumentation, control, simulation, and testing as part of the same risk
problem. The twin needs authenticated sources, explicit control boundaries, and a safe mode when
the connection becomes unreliable.

## Questions it prompts

* What real counterpart has a continuing identity in the model?
* Which decisions does the twin support, and how fresh must each observation be?
* Can users distinguish observed state, inferred state, and simulated possibilities?
* Which actions may flow back to the counterpart, and who authorises them?
* What happens when sensors, models, or connections fail?
* Can the counterpart continue safely without the twin?

## Sources

* UK Department for Business and Trade. ["What a digital twin is and how you can contribute"](https://www.gov.uk/government/publications/what-a-digital-twin-is-and-how-you-can-contribute/what-a-digital-twin-is-and-how-you-can-contribute).
  First published 29 November 2023, updated 12 January 2024.
* Frigg, Roman, and James Nguyen. ["Scientific Representation"](https://plato.stanford.edu/entries/scientific-representation/).
  *Stanford Encyclopedia of Philosophy*.
* Voas, Jeffrey, et al. [*Security and Trust Considerations for Digital Twin Technology*](https://doi.org/10.6028/NIST.IR.8356).
  NIST IR 8356, 2025.

## Related ideas

- [Bounded Context](https://robbiepalmer.me/ideas/bounded-context.md): An explicit boundary within which one domain model and its language have a consistent meaning.
- [Data Provenance](https://robbiepalmer.me/ideas/data-provenance.md): A record of where data came from, who or what changed it, and how a result was produced.
- [Domain-Driven Design](https://robbiepalmer.me/ideas/domain-driven-design.md): Model software around the language, rules, and boundaries of the problem domain.
- [Knowledge Graphs](https://robbiepalmer.me/ideas/knowledge-graphs.md): Knowledge captured as connected notes, entities, and claims.
- [Ontology Engineering](https://robbiepalmer.me/ideas/ontology-engineering.md): The practice of defining the concepts, relationships, constraints, and shared vocabulary needed to represent a domain.

## Where it appears

- Initiative: [Digital Twin of the Yard](https://robbiepalmer.me/initiatives/digital-twin-of-the-yard.md)
- Initiative: [Digital Twins for Everyday Life](https://robbiepalmer.me/initiatives/digital-twins-for-everyday-life.md)
- Initiative: [Semi-autonomous Software Development](https://robbiepalmer.me/initiatives/semi-autonomous-software-development.md)

---

Markdown index of this site: https://robbiepalmer.me/llms.txt