Goodhart's Law

Goodhart’s Law names a recurring failure mode: once a measure is pressured for control or reward, it stops behaving like a reliable measure. The corpus uses it to explain why misaligned incentives and metric substitution are structurally predictable — not accidental moral lapses.

What Goodhart actually said (1975)

Charles Goodhart, a Bank of England economist, stated the original form at a Reserve Bank of Australia conference (July 1975), published in Papers in Monetary Economics:

Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.

Context: UK monetary management in the early 1970s. Econometric relationships — e.g. between interest rates and monetary aggregates — looked stable in historical data until central banks adopted them as intermediate policy targets; under targeting pressure, the relationships broke down. Goodhart’s point was narrow and technical: ex post regularity is not a durable ex ante control lever.

→ Goodhart (1975)

What people usually quote (1997 — not Goodhart’s words)

The familiar one-liner:

When a measure becomes a target, it ceases to be a good measure.

This is Marilyn Strathern (1997), writing on university audit and ratings — explicitly paraphrasing Goodhart via Keith Hoskin’s 1996 account. It is not a sentence Goodhart wrote in 1975. Hoskin (1996) had already condensed the idea as: every measure which becomes a target becomes a bad measure — and labelled it “Goodhart’s Law.”

→ Strathern (1997)

We keep both rows: origin (Goodhart) vs generalised aphorism (Strathern).

Parallel: Campbell’s Law (1979)

Independently, Donald T. Campbell — social scientist — argued in 1979 that quantitative social indicators used for high-stakes decisions become corrupted and distort the processes they monitor (Assessing the Impact of Planned Social Change). Often called Campbell’s Law; overlaps Goodhart/Strathern in policy and organisational design. We cite it as parallel context, not as duplicate proof.

In this corpus

• Metric substitution — organisational dark pattern: the KPI works (numbers rise) while the aim erodes; Goodhart/Strathern supply the mechanism.
• Publication over provenance — papers as the target metric; reproducible chains as the discarded aim.
• Closed-loop control — once hypothesis sensors and comparators are wired to mandatory actuators, they face the same pressure; see meta-loop control. Extreme inversion: Cobra effect.
• Incentive realignment (Event Sourced Science) — attaching reward to clean provenance risks new Goodhart failure if the compensable proxy is wrong; track record and Brier-quality metrics need the same boundary discipline as any other target.

Corpus stance

Context A2 — cite Goodhart for the 1975 origin, Strathern for the popular aphorism; do not attribute Strathern’s sentence to Goodhart. Mechanism lens for incentives and dark patterns, not a substitute for tier or commitment design.