Volume 27 · Part Five · Chapter 18 of 19

When the Proxy Becomes the Goal

Category error as a slow drift, not a moment of error.

How the substitution happens

Nobody decides that the analogy is the thing. What happens is smaller. A proxy is adopted because it is tractable. It works. Papers are written in its vocabulary. The vocabulary becomes the field's shared language, and shared language is how a field thinks. Questions that cannot be posed in it stop being asked, not because they were rejected but because they became awkward to say. Eventually optimisation runs on the proxy — better lattices, prettier visualisations, sharper effective metrics — and the original purpose remains formally intact while playing no active role in any decision.

This is the same structure diagnosed in Volume 26 for alignment engineering, where the measurable stand-in for a goal becomes the goal that is actually pursued. Physics has no immunity to it; it has better error-correction, in the form of experiments that eventually fail.

Three cases from this volume

Constitutive parameters (Chapter 2). A bookkeeping device becomes 'the impedance of spacetime', and then a design target, and the reader is three sentences from a device that cannot exist.

Holography (Chapter 6). 'Dual to' becomes 'is'. The technical work stays honest; the summary sentences do not, and summary sentences are what leave the field.

The braid (Chapters 7–15). This author's own figure, which describes an intuition about relations, has repeatedly been asked to carry evidential weight it cannot. Naming one's own case is not a rhetorical courtesy — it is the only version of this chapter that has any standing.

Equations borrowed

  • The proxy/goal analysis of Volume 26, Chapter 32
  • Goodhart's observation on measures that become targets

Validity band

A claim about how research language drifts. It is a historical and sociological claim, not a physical one, and should be judged by cases.

Falsifier

A field where a proxy has been in heavy use for decades with no drift in its summary language would be a counter-example worth taking seriously.

Where this chapter is weakest

The argument is illustrated rather than demonstrated. Three cases chosen by the author to fit the thesis are not evidence of a general pattern; a systematic study of how status labels degrade across citation generations would be, and does not exist here.

The volume-wide audit of these weak points is collected in Where This Volume Is Weak.