This argument seems common to many debates:
Proposal P arrogantly assumes that it is possible to measure X, when really X is hard to measure and perhaps even changes depending on other factors. Therefore we shouldn’t do P
This could make sense if X wasn’t especially integral to the goal. For instance if the proposal were to measure short distances by triangulation with nearby objects, a reasonable criticism would be that the angles are hard to measure, relative to measuring the distance directly. But this argument is commonly used in situations where optimizing X is the whole point of the activity, or a large part of it.
Criticism of cost-benefit approaches to doing good provides a prime example. A common argument is that it’s just not possible to tell if you are increasing net welfare, or by how much. The critic concludes then that a different strategy is better, for instance some sort of intuitive adherence to strict behavioural rules.
But if what we fundamentally think matters most is increasing welfare, or at least reducing extreme suffering, then the extreme difficulty of doing the associated mathematics perfectly should not warrant abandoning the goal. One should always be better off putting the reduced effort one is willing to contribute into what accuracy it buys, rather than throwing it away on a strategy that is more random with regard to one’s goal.
A CEO would sound ridiculous making this argument to his shareholders: ‘You guys are being ridiculous. It’s just not possible to know which actions will increase the value of the company exactly how much. Why don’t we try to make sure that all of our meetings end on time instead?’
In general, when optimizing X somehow is integral to the goal, the argument must fail. If the point is to make X as close to three as possible for instance, no matter how bad your best estimate is of what X will be under different conditions, you can’t do better by ignoring X altogether. If you had a non-estimating-X strategy which you anticipated would do better than your best estimate in getting a good value of X, then you in fact believe yourself to have a better estimating-X strategy.
Probabilistic risk assessment is claimed by some to be impossibly difficult. People are often wrong, and may fail to think of certain contingencies in advance. So if we want to know how prepared to be for a nuclear war for instance, we should do something qualitative with scenarios and the like. This could be a defensible position. Perhaps intuitions can better implicitly assess probabilities via some other activity than explicitly thinking about them. However I have not heard this claim accompanied by any motivating evidence. And if it were true, it would likely make sense to convert the qualitative assessments into quantitative ones and aggregate them with information from other sources, rather than disregarding quantitative assessments all together.
This post was originally published on Katja’s blog, Meteuphoric.