Intelligence, Explained
On understanding what we ask machines to understand
We ask machines to explain. But what is being asked? An answer is not an explanation. A system can produce the correct output without revealing how it arrived there. It can recognise patterns, predict outcomes and generate responses with remarkable accuracy. That is performance. Understanding is something else.
To explain is to make reasoning visible. To show not only the result, but the path that leads to it. To expose assumptions, sources and the structure behind a conclusion.
Explanation is not about certainty. It is about traceability. This distinction is becoming structural.
As artificial intelligence moves into domains where decisions matter—law, healthcare, governance—the difference between output and understanding can no longer be ignored.
A system that performs can assist. A system that explains can be trusted. Explanation is often treated as an additional feature.
Something that can be layered onto an existing system. A form of transparency that can be enabled when needed.
That assumption is fragile. Some systems are not designed to explain. They are designed to optimise.
As models increase in complexity, their internal logic becomes harder to reconstruct. Not because it is hidden, but because it emerges from interactions that cannot easily be reduced to a single line of reasoning.
The system knows. But it cannot always tell.
A tension emerges. Explanation is expected. Opacity is deployed.
Cogito, ergo sum.
To think was to exist. But it also implied something more: the ability to articulate reasoning, to make thought visible and open to challenge.
René Descartes
That expectation has not disappeared. It has been deferred. Explainability is not a feature. It is a design condition.
The question is not whether machines can explain. It is whether explanation is required. Because in the end, explanation is not about the system. It is about us.
What is accepted. What is trusted. And what is allowed to remain unseen.
Explanation is not about certainty. It is about traceability.
This article is part of The Black Box Divide, a series exploring how Europe is redefining intelligence, accountability and control in the age of AI.
🎨 Credit
Illustration by Altair Media (AI-generated)
✍️ Caption
From noise to outcome. But what remains unseen is the path in between.
