INEQUALITIES. This point is technical but central, central. The problem with fragilitas is that they focus on equalities (x=.0153745), which gets them into trouble, and diverts them in a vapid and blinding search for certainties, with monstrous model errors. It is more robust to focus on inequalities, such as: probabilities of all these possible outcomes are <= 1 (there is an outcome we may have missed in our imagination, called “Black Swan”), or <1; the probability of such an event >=0, the odds of another Fukushima >= computation by this specific model, etc. First it is harder to make a mistake with inequalities, and in some cases, logically impossible. Now it looks like users of convex heuristics work with inequalities. Jensen’s inequality is an inequality.

Consider the statement: my flight is >= 4 hours. It would be impossible to make a mistake there, compared to someone saying <=6 hours. One side is immune to error, not the other since flight can be delayed by error, but not accelerated by more than a small margin.

We can use mathematics to produce inequalities that we can rely on in the real world.

Fragility and Antifragility being asymmetric exposure to randomness, an inequality comes out of the description: concave and convex outcomes can be expressed by inequalities.

via INEQUALITIES. This point is technical… – Nassim Nicholas Taleb | Facebook.