Nassim Nicholas Taleb

A Black Swan is a high-impact, low-frequency event that is unpredictable in advance, rationalized after the fact, and dominant in determining outcomes. Its statistical engine is the fat-tailed distribution: in "Extremistan," the variance and the mean of a process are dominated by a few extreme realizations, so the historical record undersamples the tail, sample means are unreliable, and standard Gaussian risk machinery (and any point forecast) understates true exposure. The practical consequence is epistemic humility about prediction and a shift of attention from estimating probabilities to bounding consequences. Key works: *The Black Swan* (2007); the formal treatment of fat tails and ruin in Taleb, Read, Douady, Norman & Bar-Yam, "The Precautionary Principle (with Application to the Genetic Modification of Organisms)," arXiv:1410.5787 (2014).

Antifragility is a property distinct from robustness and resilience. The *fragile* is harmed disproportionately by volatility and large deviations (it is concave to the stressor); the *robust* resists and stays the same; the *resilient* recovers; the *antifragile* gains from volatility, randomness, and stressors up to a survivable dose (it is convex to the stressor). Fragility/antifragility is detectable from the second-order (convex/concave) response of a system to the dose of a stressor, independent of any forecast of the stressor itself. This makes "is this system fragile?" answerable when "what will happen?" is not. Key work: *Antifragile: Things That Gain from Disorder* (2012), doi 10.1080/14697688.2013.829244; operationalized for engineered systems by Jones, "Antifragility analysis and measurement framework for systems of systems," doi 10.1007/s13753-013-0017-7.

The decisive distinction is between repeated risks you can survive and absorbing risks that end the game. When a system faces *ruin* (an absorbing barrier with no return), the time-average and the ensemble-average diverge: a strategy with positive expected value can still lead almost surely to ruin if repeated. Therefore exposure to ruin must be constrained *regardless of its estimated probability*, because the estimate is unreliable and the consequence is irreversible. This is the basis of the *non-naive precautionary principle*: confine precaution to the joint case of fat tails and systemic (non-localizable) risk, and there invert the burden of proof onto the party introducing the systemic stressor. Key work: Taleb et al., arXiv:1410.5787 (2014).

Improvement comes more reliably from *removing* fragilities and iatrogenics (via negativa) than from adding clever interventions, because we know fragility more reliably than we predict benefit. The *barbell strategy* manages tail risk by combining extreme safety in the bulk of a position with small, bounded, convex exposures to upside, deliberately avoiding the fragile "moderate-risk" middle that hides tail exposure. *Optionality* (keeping reversible options open, paying small costs to preserve the right but not the obligation to act) is how one harvests positive Black Swans while capping the downside. Key work: *Antifragile* (2012); the decision-under-deep-uncertainty analogues in Haasnoot et al., "Dynamic Adaptive Policy Pathways," doi 10.1016/j.gloenvcha.2012.12.006, and Kwakkel et al., doi 10.1016/j.envsoft.2016.09.017.

Risk-bearing must be symmetric with decision-making: those who make decisions that can impose tail harm on others must themselves be exposed to the downside. Absent skin in the game, agents rationally transfer fragility to others and to the future (hidden risks, deferred liabilities), and forecasts and assurances become cheap because their authors bear no consequence. Skin in the game is both an ethical filter and a statistical one: it removes the actors whose models would otherwise blow up the system. Key work: *Skin in the Game* (2018), within the *Incerto*; engineering analogue in the Safety-I/II antifragility debate, Martinetti et al., doi 10.1080/10803548.2018.1444724.