Goodhart's Law
When a measure becomes a target, it ceases to be a good measure.
Originally formulated by economist Charles Goodhart in 1975 regarding monetary policy, Goodhart's Law has become a cornerstone of systems theory, management, and AI research. The law states that whenever an indicator or performance metric is chosen as a target to control a system, people or algorithms will optimize their behavior specifically to hit that metric. Consequently, the measure loses its ability to reliably reflect the underlying phenomenon it was intended to gauge. People find loopholes, 'game the system,' or manipulate data to satisfy KPIs, often causing unintended and damaging side effects. In artificial intelligence, this manifests as 'reward hacking,' where a model discovers an unexpected shortcut to maximize its reward function without actually solving the intended problem.
If a hospital is evaluated on how quickly emergency room patients are admitted, staff might leave patients waiting in ambulances outside before officially registering them. The metric shows short wait times, but actual healthcare quality drops.
The law reveals why well-intentioned rules and incentive structures frequently backfire. It reminds us that complex systems adapt to how they are measured, meaning simple KPIs almost always fail without holistic oversight.
Many people believe Goodhart's Law implies we should stop measuring things altogether. In reality, it means metrics should be used for assessment and insight, not as direct targets or incentives.
Identify one target at your workplace or in your daily life and figure out how it could be gamed.
When a metric is turned into a target, behaviors adapt in ways that destroy the metric's usefulness.
Get today's concept by email
Get one short, thought-provoking concept in your inbox each day. Unsubscribe at any time.
By subscribing you agree to our privacy policy. We never sell or share your address.