In effect you’re making a claim that in the general case, less systemic precision is preferable because more precision is brittle However you cannot know the r

In effect you’re making a claim that in the general case, less systemic precision is preferable because more precision is brittle However you cannot know the relative brittleness apriori and you have to have a consistent measure that you’re comparing to Like it seems that you’re suggesting a loose grained low variance forecasting model will only find high likelihood inferences and filter everything less impactful - making it more robust However experiments regularly destroy previous concepts by the merging of conflicting variables - this is hegelian synthesis, or popperian falsification So foundationally that thesis is flawed because it implies no system improves from additional observation - it gets more brittle. The entire theory of the firm, monopoly efficiencies etc… are a counter example disproving this logic That said, all systems are coupled, so if a complex system is optimized around a narrow saddle point then it tipping into a new state may perturb other coupled models which cause cascading adjustments - that’s not more brittle by default. It means you have to observe more coupled interactions and not overestimate likelihood you can’t experimentally validate with a functional system.