I thought the idea was not to encourage defeatism in the presence of difficult-to-predict events, but rather to structure the way that we go about things to mitigate the impact when they do occur. In other words, to be humble about our ability to predict the future, to make sure that we have fall-back positions available when things inevitably go wrong, and to organize ourselves so that we can move quickly and adapt to changes when they occur. When we have to predict the future, we should do it to the best of our ability, using the most sophisticated techniques at our disposal, but we should not act surprised on those inevitable occasions when we fail. Another thing regarding Gaussian distributions:- The Central Limit theorem only gets us so far. Normality, in the datasets that I have worked with, is the exception rather than the rule.

I really, really, really like that aphorism.

To "have the courage to act on your knowledge, but also to have the humility to doubt what we know" might seem like it requires us to be hypocrites. I have felt this a few times before, occasionally concluding that hypocrisy is a necessity, or we at least must permit a certain level of transience and inconsistency in our beliefs in order that we may more effectively deal with a complex and changing world. This is supported, to an extent, by the failures of the strong AI community to build coherent, atomic and unified databases of propositions that are broadly applicable to the world as a whole, and the relative successes of smaller domain-specific "micro models". Similarly, the broader applicability of "Agile" (or responsive) software development methods vs strictly-planned/predictive methods speaks a similar message to the ability of static models to deal with a changing, dynamic world. In any case, this phrase "Strong Opinions, Weakly Held" comes at the issue from a different, and very exciting angle; and is an idea that is well worth talking about.