We are collectively excellent at first-order thinking and remarkably poor at everything beyond it. You want more of something, so you measure it and reward it. You want a problem solved, so you apply the fix that addresses it. In each case the immediate effect is achieved, which is why the approach feels vindicated, and in each case the interesting consequences arrive later and somewhere else.

The pattern is easiest to see with incentives. Reward the number of features shipped and you will get features, along with an architecture nobody wants to touch three years from now. Measure lines of code, bug counts, utilisation, publication numbers: in every case you get exactly what you asked for and, alongside it, a set of behaviours you did not ask for and would not have approved of. This is not because people are gaming the system out of bad faith. It is because a metric is a compressed and lossy description of what you actually wanted, and any sufficiently motivated group or individual will optimise the description rather than the thing. 

What I would take from this is not paralysis. Almost nothing would ever get done if we required a full accounting of the consequences first. The regulations coming out of Brussels are a good example of that. It is a single habit: after deciding, ask what happens next, and then what happens after that. Two steps is usually enough to catch the worst of it, and two steps is more than almost anyone does. So, as a reflection for today: what have you optimised recently? And what is it quietly producing that you did not intend?