Learning to See

A method for uncovering hidden costs and comparing without misleading boundaries.

A photograph is never the whole story

An electric car emits nothing from its tailpipe because it has no tailpipe. A rooftop solar panel burns no fuel. A nuclear plant releases no black smoke. Each of these three statements is true and, at the same time, misleading: each describes a single instant — the moment of use — and says nothing about everything that happened before (mines, factories, transport) or everything that will happen after (waste, decommissioning, recycling). Learning to see means, above all, distrusting any claim about energy that only talks about that single instant.

This section is not about any one energy source in particular. It is about the method that has to be applied before comparing any of them: where the boundary of the analysis is drawn, what counts inside it and what is left out, and what questions always have to be asked to avoid being misled by a well-chosen number.

Why the system boundary decides almost everything

Any claim about the "cost" or "impact" of an energy technology implies, even if it is never said out loud, a decision about where the accounting starts and ends. If the boundary is drawn only around the power plant or the vehicle in operation, almost any technology with low direct emissions looks flawless. If the boundary is widened to include material extraction, manufacturing, transport, maintenance and end of life — what is known as a life cycle assessment (LCA) — the picture changes, sometimes a lot and sometimes only a little, depending on the technology.

This is not a rhetorical trick or a way of discrediting clean technologies: the conservation of energy and matter guarantees that nothing is generated out of nothing. Every machine that converts an energy source into useful electricity had to be built from real materials, extracted from some real place, using machinery that consumed real energy. The only legitimate question is how much that invisible part weighs compared with the visible part — and that proportion varies enormously between technologies: tiny for some, substantial for others, which is exactly the kind of fact that only shows up once the boundary is drawn correctly.

The second error, subtler than the first: comparing different services

Drawing the boundary correctly is not enough if the comparison afterward mixes apples with oranges. Asking "which pollutes more, an electric car or a petrol car?" without further detail is an incomplete question: per kilometer driven, per car manufactured, over its whole lifetime or only in its first year? charged with what electricity mix? comparing a small city car with a large SUV? Changing any of those conditions can flip the answer. The discipline of comparing "like with like" — the same functional unit, the same service delivered, the same time period — matters just as much as drawing the boundary correctly, and it is much easier to overlook without noticing.

What this method does NOT do

Correctly applying a life cycle assessment does not automatically produce a winner. There is no formula that reduces everything — carbon, water, land, toxicity, time — to a single neutral, objective number; any way of combining those magnitudes into one index requires choosing how much each one weighs, and that choice is a values decision, not a scientific fact. What the method does guarantee is that the comparison is made on the same basis, with the same boundaries stated explicitly, so that anyone can check what was included and what was left out — and decide whether they agree with that choice.

How to follow this section

The leaf The full life of a kilowatt-hour develops the complete method with a step-by-step worked example, shows how much a figure can change depending on where the boundary is drawn, and gives a list of questions that can be applied to any energy technology appearing elsewhere on this platform.