The article isn't a critique of teams following Goodhart's Law IRL. It's a formalization of concepts between machine learning and Goodhart's Law including a "st

The article isn't a critique of teams following Goodhart's Law IRL. It's a formalization of concepts between machine learning and Goodhart's Law including a "strong" formalization where performance can diverge. I found it quite useful to think about these concepts with respect to machine learning. In a sufficiently large system of humans interacting, there's a lot of emergent stochasticity that is tough to model. Some examples given are of teams working together at a business, but seemed intentionally written to strike familiarity with the reader, not to make suggestions about how people should analyze metrics on a daily basis.