risk and uncertainty

Two everyday situations feel uneasy but in different ways. Rolling a fair die, you do not know what number will come up — yet you know all six outcomes and that each has a 1-in-6 chance. Walking into a brand-new business in a country you have never visited, you also do not know how it will go — but here you cannot even list the outcomes, let alone attach honest odds. The first kind of not-knowing is measurable; the second is not. That distinction sits at the heart of every risk decision.

Risk, in the careful sense, means a situation where the possible outcomes are known and we can attach probabilities to them — even if only estimated ones — so the uncertainty can be measured. Uncertainty (sometimes called Knightian uncertainty, after the economist Frank Knight) means a situation where we lack a reliable basis for those probabilities: the outcomes may be unknown, the data too thin, or the world changing in ways no past record captures. A coin flip is risk; the chance that a never-before-seen technology disrupts an industry is closer to uncertainty. Most real problems are a blend, and a good analyst is honest about which part is which.

Actuaries live on this borderline. Their tools — probabilities, expected values, statistical models — work best on genuine risk, where past data and stable patterns let them estimate odds. The danger is treating deep uncertainty as if it were tidy risk: putting a precise number on something that does not deserve one. Sound practice means estimating where you fairly can, widening your margins where you cannot, and stress-testing for the surprises your model never saw.