[wespiser_2018]

What most people really struggle with is uncertainty, and they lack the conceptual framework (distributions) and working understanding of how those distributions behave under various scenarios.

One of the better data scientists I've worked with explained to me that his entire job was 'quantifying uncertainty', and after working in the field I definitely believe it. You could play around with RStan for the weekend and gain an intuition for how evidence changes hypothesis and how a hypothesis could be the "most likely" but still "unlikely" in an absolute sense, all with tremendous uncertainty in your prediction.

Still, when you try to communicate the results of a model, uncertainty is simply hard to communicate when people are interest in using your analysis for a snap decision, and the uncomfortable truth is "we can't tell you" is a fine modeling result.