(with Application to the Genetic Modification of Organisms)

By: Nassim Nicholas Taleb, Rupert Read, Raphael Douady, Joseph Norman, Yaneer Bar-Yam

In: EXTREME RISK INITIATIVE, NYU SCHOOL OF ENGINEERING WORKING PAPER SERIES

Date: 17 October 2014

Key Concepts:

• Precautionary principle: preventing decision makers from putting society as a whole at risk .
  • Potential global, irreversible implications (fat tail, systemic, risk of ruin).
  • Action should not be taken in the absence of near-certainty about safety.
• Proper consideration of risk involves assessing probabilities and consequences.
• Probability distributions: thin versus fat tails.
  • Mild variations.
    • Bottom-up, slow evolution (nature).
    • No individual variation represents large share of sum of population.
    • Limited consequences: independent, local impact.
    • Example: people’s weight (no human is heavier than 10 x average).
    • Harm comes from collective effect of many events.
  • Large variations.
    • Top-down, engineered (man-made, some in nature).
    • Single deviation may take up large share of sum of population.
    • Large consequences: interdependent, cascading, systemic impact.
    • Example: people’s net wealth (single person > 10 x average).
    • Harm comes from the single largest effect.
• Consequences: local versus systemic risks, risk of ruin, fragility.
  • Local versus systemic.
    • Localized, non-spreading impact.
    • Connected, propagating impact.
  • Risk or ruin: non-zero probability of an unrecoverable loss.
    • Low probability wrongly interpreted as acceptable risk on one-off basis.
    • However, over time, as exposure increases, probability of ruin increases.
    • Implication: future ceases to exist = cost is effectively infinite.
    • Traditional risk assessment of cost-benefit analysis nonsensical.
    • Limited use for statistical, experimental or evidentiary analysis.
    • Lack of historic harm is unconvincing (require a very long history).
  • Fragility: harmed by uncertainty.
    • Non-linear response to random events.
    • Ability to withstand (and recover) from small impacts.
      • Large stone (weight of 10 pebbles) hurts more than 10 pebbles.
      • Converse, 10 pebbles don’t hurt as much as one large stone.
      • I’m not destroyed by cumulative effect of many small events.
      • Linked to small, local, bounded risks.
      • Allows for tinkering, progress, adjustment.
• Real world examples:
  • Nuclear energy: PP does not apply (mostly a severe local risk).
    • • GMOs: PP applies, need for severe limits.