Goodhart's law across the corpus: when the proxy becomes the target — generalizes the feedback loop half of → Why reward proxies get gamed: the anatomy of reward hacking

Within Concrete Problems' own taxonomy, one cause is explicitly nested inside another, a relation neither theme's narrative states on its own. The paper's Feedback Loops bullet closes by noting that the phenomenon "can be considered a special case of Goodhart's law, in which the correlation breaks specifically because the objective function has a self-amplifying component" (concrete-problems, §"Feedback Loops:", p. 8). Goodhart's law appears in the source as its own numbered cause, distinct from feedback loops, partially observed goals, complicated systems, abstract rewards, and environmental embedding, yet the text collapses the feedback-loop cause into it rather than treating the two as siblings. The other four causes in this theme keep their independent status; feedback loops alone is demoted, in the paper's own account, from a peer cause to a named instance of the general law this theme's other members track empirically across 2022.