Goodhart's law

Goodhart's law is named after British economist Charles Goodhart, who formulated the idea in the context of monetary policy. It holds that when people or systems optimize for a particular indicator, that indicator stops reflecting what it was originally meant to measure.

In AI, it is used to explain problems such as reward hacking and specification gaming, in which a model raises its reward or evaluation score in ways that depart from the intent of the task.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.