There was a period at school when “attendance rate” was used as a key performance metric. Improving attendance became the goal. The result? Sick students came to school anyway. Illness spread. Attendance went up — but the actual purpose of school, “healthy learning for students,” got worse.
This is Goodhart’s Law.
What Is Goodhart’s Law?
A law proposed by economist Charles Goodhart in 1975:[1]
Related: optimize your sleep
“When a measure becomes a target, it ceases to be a good measure.”
Originally, a measurement is a proxy for something we actually care about. Attendance rate is a proxy for learning engagement. Test scores are a proxy for real understanding. But when a proxy becomes the target, people optimize for the proxy rather than the actual goal.
Goodhart’s Law in Education
Education is one of the fields where Goodhart’s Law shows up most clearly:
- Test scores: Once they become the goal, students study for the score — regardless of actual understanding.
- Teaching evaluation scores: When teachers are evaluated by student satisfaction, challenging courses disappear and only popular ones survive.
- Publication count: When researchers are evaluated by number of papers, a flood of high-quantity, low-quality papers follows.
I experienced this firsthand as a teacher. To increase students’ “participation,” I used number of presentations as a metric. Result: students gave lots of short, meaningless presentations. Real participation actually declined.
Why Goodhart’s Law Is Hard to Avoid
Nassim Taleb says the impulse to measure complex systems with simple metrics is natural.[2] The human brain dislikes complexity. Seeking a simple, measurable indicator is System 1’s natural response.
A similar pattern emerges from Kahneman’s research.[3] We tend to value what is measurable and ignore what is difficult to measure — but the truly important things are often the hardest to measure: depth of knowledge, critical thinking ability, collaborative skills.
How to Avoid Goodhart’s Law
You can’t avoid it entirely, but these approaches help:
- Use multiple metrics simultaneously: Focus on a single metric and it becomes a game. Track several together and gaming becomes harder.
- Rotate metrics regularly: The longer a metric has been in use, the more it becomes an optimization target. Replacing it with something new lets you see the real picture again.
- Supplement with qualitative evaluation: Don’t just look at numbers — directly observe what’s happening on the ground.
- Always remember what the metric is for: “What is this metric a proxy for? Is there a way to measure that directly?”
Eliezer Yudkowsky warns that Goodhart’s Law is especially dangerous in AI systems.[4] When the metric an AI is optimizing for differs from what we actually want, the AI can destroy what we care about while optimizing the metric perfectly.
Review the performance metrics you use right now. Are they actually measuring what you want? Or have they become the goal themselves?
Disclaimer: This article is for educational and informational purposes only. It is not a substitute for professional medical advice, diagnosis, or treatment. Always consult a qualified healthcare provider with any questions about a medical condition.
Your Next Steps
- Today: Pick one idea from this article and try it before bed tonight.
- This week: Track your results for 5 days — even a simple notes app works.
- Next 30 days: Review what worked, drop what didn’t, and build your personal system.
Last updated: 2026-03-16
About the Author
Written by the Rational Growth editorial team. Our health and psychology content is informed by peer-reviewed research, clinical guidelines, and real-world experience. We follow strict editorial standards and cite primary sources throughout.
References
- Goodhart, C. A. E. (1975). Problems of Monetary Management. Papers in Monetary Economics. Reserve Bank of Australia.
- Taleb, N. N. (2007). The Black Swan. Random House.
- Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.
- Yudkowsky, E. (2008). Rationality: From AI to Zombies. MIRI.
- Tetlock, P., & Gardner, D. (2015). Superforecasting. Crown Publishers.