Goodhart's law

Goodhart's law is the observation that when a measure becomes a target, it ceases to be a good measure. The law is named after economist Charles Goodhart, who formulated it in 1975 in the context of monetary policy: any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes. The law has since been generalized far beyond economics into a fundamental principle of systems theory, organizational behavior, and epistemic engineering.

The core mechanism is metric corruption: once an actor knows that their performance is evaluated by a specific metric, they optimize for the metric rather than for the underlying goal the metric was designed to proxy. The metric and the goal decouple. The result is a system that appears to perform well by its own lights while failing at the purpose for which it was created.

The Logic of Metric Corruption

Goodhart's law operates through several distinct mechanisms:

Direct gaming. Agents manipulate the measure without improving the underlying outcome. Teachers teach to the test. Hospitals refuse high-risk patients to improve mortality statistics. Universities optimize for citation metrics rather than research quality. The measure is met; the goal is not.

Causal displacement. The metric originally correlated with the goal because both were caused by a third factor. When the metric is targeted, the correlation breaks. Example: a company uses "hours worked" as a proxy for "productivity" because productive employees often work long hours. When employees are rewarded for hours worked, they stay late without producing more.

Systemic adaptation. The system as a whole reorganizes around the metric. When academic hiring committees use journal impact factors, the entire field of science reorganizes: researchers chase high-impact journals, journals game their impact factors, and the distribution of research topics shifts toward what is publishable in those venues.

Observational collapse. The act of measuring changes the system in ways that destroy the information the measurement was supposed to provide. This is the Lucas critique in economics: policy based on historical relationships fails because the policy itself alters the behavioral relationships.

Epistemic Applications

Goodhart's law is particularly destructive in epistemic systems — systems whose goal is truth-tracking:

• Peer review metrics. When journal impact factors are used to evaluate researchers, the review process shifts from "is this true?" to "will this be cited?"
• Fact-checking organizations. When fact-checkers are rated by "number of false claims debunked," they may prioritize high-volume, low-difficulty claims over complex, high-stakes misinformation.
• Search engine optimization. Google's ranking algorithm was designed to measure relevance. Once websites began optimizing for the algorithm, the algorithm ceased to measure relevance and began to measure optimizability.
• Social media engagement. Platforms measure "engagement" as a proxy for "value to users." Once content producers optimize for engagement, the metric ceases to track value and tracks arousal: outrage, fear, and confirmation bias. See Attention Economy.

Campbell's Law and the Generalization

Sociologist Donald T. Campbell proposed a broader version in 1976: the more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor. Campbell's law is Goodhart's law with the scope expanded from economics to all social indicators and the mechanism specified as social process distortion.

The metric is a policy disguised as a measurement. When a university ranks departments by research income, it is not merely measuring productivity — it is declaring that research income is what matters, which favors large-scale, fundable science over small-scale, unfundable inquiry.

The Impossibility of a Pure Metric

Goodhart's law cannot be solved by "better metrics." Any metric, once targeted, will be gamed. The only partial solutions are metric diversity, process-over-output measurement, or abandonment. The radical implication is that any system of centralized evaluation carries the seeds of its own corruption. The question is not whether a metric will fail but how quickly — and whether the failure mode is visible before the system collapses.

The deepest irony of Goodhart's law is that it applies to itself: once "avoid Goodhart's law" becomes a management target, managers will optimize for the appearance of metric health rather than genuine resilience. The law is not merely a warning about metrics. It is a theorem about the impossibility of centralized evaluation in complex adaptive systems.