The Limits of Optimisation

Modern institutions are devoted to finding the single best possible way of doing things. Routes are optimised, schedules are optimised, and increasingly so are the people inside them.

现代机构致力于找出做每件事「唯一最好的方式」。路线被优化，日程被优化，而人也在被优化 —— 而且越来越如此。

Optimisation works best when the goal is clear, stable and easy to measure. It works worst when the goal is complicated, slow to arrive, or partly invisible.

当目标清楚、容易衡量时，优化效果最好。而当目标复杂、缓慢、或部分看不见时，它效果最差。

A hospital that is judged only by its waiting times will certainly reduce its waiting times. It may well do so by quietly turning away the patients whose cases are hardest to treat.

一个只按「等候时间」考核的医院，一定会把等候时间降下来。但它可能是靠拒收最难治的病人做到的。

The measure becomes the target, and once that happens, it stops measuring what it was originally meant to measure. This does not happen because anyone is dishonest; it happens because people respond to incentives.

指标成了目标；而一旦成了目标，它就不再衡量当初想衡量的东西了。这并不因为有人不诚实，而是因为人会顺着「激励」的方向走。

So the real question is never whether to measure, but which numbers deserve to guide a decision. Some of the most important things we value are precisely the ones that resist being counted.

所以问题从来不是「要不要衡量」，而是「哪些数字有资格指导决策」。我们最看重的一些东西，恰恰正是那些最抗拒被数出来的。