I read something recently here, comparing this phenomenon to the phenomenon of overfitting.
The gist is that we want to leverage our ability to measure things that are easy to measure, which are correlated with things that are hard to measure, and then optimize for those things. The downside being that you will always implicitly be sacrificing the things you don't/can't measure, and these sacrifices tend to occur in greater marginal amounts as the low hanging fruit gets eaten.
There are obviously interventions which are so ill convceived that they have no place being implemented in the first place, but this mechanism applies to every intervention eventually.
The gist is that we want to leverage our ability to measure things that are easy to measure, which are correlated with things that are hard to measure, and then optimize for those things. The downside being that you will always implicitly be sacrificing the things you don't/can't measure, and these sacrifices tend to occur in greater marginal amounts as the low hanging fruit gets eaten.
There are obviously interventions which are so ill convceived that they have no place being implemented in the first place, but this mechanism applies to every intervention eventually.