In the cited article Raphaële Chappe notes:

J.M. Keynes was perhaps the first economist to grasp the full significance of radical uncertainty for economic analysis. Author of A Treatise on Probability published in 1921, Keynes developed the view that it is not always possible to assign a numerical value to probabilities -- a point unrelated to the practicality of measurement (Keynes, 1921). The frequentist approach breaks down in some situations. It is impossible to find strictly relevant frequency ratios because even if a relevant statistical database were to exist, the presence of additional information might make the database inadequate for the computation of a numerical probability.

In his Treatise, Keynes gave the example of the outcome of a Presidential election, or something as ordinary as the probability that we reach home alive if we are out for a walk. In a famous later article, he gave the example of "the price of copper and the rate of interest twenty years hence" (Keynes, 1937). The implication for Keynes' General Theory was that macroeconomic variables such as investment and consumption are shaped by many factors which are not susceptible to probabilistic analysis (fundamentally different from the outcomes of a game of roulette).

Events determined by human decisions (much of which, he argued, are driven by `animal spirits' - ideas and attitudes determined by intangible psychological motivations) do not lend themselves to the calculation of probabilities. We can think of many events that are likely to influence economic variables (such as prices and incomes) and economic decisions (such as the level of investment) but are by their very nature radically uncertain, such as the occurrence of wars, changes of government, inventions, climatic changes.

Yet this insight was completely neglected in the mainstream economics that followed (and arguably marginalized by Keynes's own mainstream interpreters, such as J.R. Hicks). The notion of calculable risk has permeated modern microeconomics, starting with its formalization in the concept of `expected utility theory' introduced by J. von Neumann and O. Morgenstern in 1947. There has been a triumph of risk over uncertainty (Reddy, 1996).  It is something of an irony of history that the importance of the incalculability of risk was first explored in a discipline which has since failed to take the insight into account. This irony is not lost to other social scientists:

"The discovery of the incalculability of risk is closely connected to the discovery of the importance of not-knowing to risk calculation, and it is part of another kind of irony, that surprisingly this discovery of not-knowing occurred in a scholarly discipline which today no longer wants to have anything to do with [it]: economics." (Beck, 2006: 334).

One end of the spectrum of uncertainty is events which have a well defined spectrum of possible outcomes and well defined underlying probability distributions, but where it is excessively costly, time-consuming or otherwise onerous to establish the distribution ahead of making a decision. At the other end of the spectrum of uncertainty you have the Outside Context Event.

Somewhere around the middle you have the sort of event where the span of possible outcomes is not sufficiently well defined that it can be described by a probability distribution, but which is none the less sufficiently predictable that it is possible to devise heuristics and contingency plans.

Those are all "uncertainty," but they are very different sorts of uncertainty. (Indeed it is risk, not uncertainty, that is the extreme edge case of the human condition.)

- Jake

If you only spend 20 minutes of the rest of your life on economics, go spend them here.