Over at The Physics arXiv Blog:
Taleb and co begin by making a clear distinction between risks with consequences that are local and those with consequences that have the potential to cause global ruin. When global harm is possible, an action must be avoided unless there is scientific near-certainty that it is safe. This approach is known as the precautionary principle.
The question, of course, is when the precautionary principle should be applied. Taleb and co begin by saying that their aim is to place the precautionary principle within a formal statistical structure that is grounded in probability theory and the properties of complex systems. “Our aim is to allow decision-makers to discern which circumstances require the use of the precautionary principle and in which cases evoking the precautionary principle is inappropriate.”
Their argument begins by dividing potential harm into two types. The first is localised and non-spreading. The second is propagating harm that results in irreversible and widespread damage. Taleb and co say that traditional decision-making strategies focus on the first type of risk where the harm is localised and the risk is easy to calculate from past data.
In this case, it is always possible to make a mistake when decision-making about risk. The crucial point is that when the harm is localised, the potential danger from a miscalculation is bounded.
By contrast, harm that is able to propagate on a global scale is entirely different. “The possibility of irreversible and widespread damage raises different questions about the nature of decision-making and what risks can be reasonably taken,” say Taleb and co. In this case, the potential danger from a miscalculation can be essentially infinite. It is in this category of total ruin problems that the precautionary principle comes into play, they say.