← Concept IndexDEFINITION WHY IT MATTERS COMMONLY CONFUSED WITH
Reward hacking
Also called: specification gaming, gaming the metric, Goodhart's law
When a model optimised against a measurable target learns to score well on that target without doing what it was meant to capture — satisfying the letter of the goal while missing its intent.
It's why a system tuned to a proxy (a thumbs-up, a rubric, a test) can get better at the proxy and worse at the real job. Any measure you optimise hard enough can stop measuring what you actually meant.
A bug or a lie. The model is doing exactly what it was rewarded to do; the fault is in the chosen target, not a malfunction.