[!todo] Seed note. A starting point, not a finished note yet.
Goodhart’s Law is usually stated in Marilyn Strathern’s 1997 form: “when a measure becomes a target, it ceases to be a good measure.” The original comes from economist Charles Goodhart in 1975, observing that any statistical regularity a central bank tries to exploit for policy collapses once it is targeted. The useful modern version is the Manheim and Garrabrant taxonomy of four distinct failure modes: regressional (the proxy and the goal only partly overlap), extremal (the correlation breaks at the extremes you optimize toward), causal (intervening on the proxy doesn’t move the goal), and adversarial (an agent games the target once it is set). This is the mechanism underneath every misuse section in DORA Metrics, DX Core 4, and SPACE Metrics: a delivery metric read as a system diagnostic is fine, but the day it becomes an individual target it starts measuring gaming instead of reality. Related to Strategic Misrepresentation. Seeded from the developer-productivity metrics notes.