|
Worse yet, when unpredicted undesired events eventually occur (and they usually do), few analysts return to their original inputs and premises to find out how and where their predictions failed. They compound their original inaccuracies downstream to the next naïve analyst who assumes, without verification, the implied accuracy of the original prediction(s). It often seems that once an analysis has been calculated, any motivation to revise it to conform to reality disappears. It's more likely that errant analysts will dig out some past happening and point to the latest events as "retrocursors," by which to establish the prophecy of their original analyses.
In the referenced paper, Taleb and Pilpel summarize their First Epistemological Problem: "induction and small probability," thus:
If small probability events carry large impacts, and (at the same time) these small probability events are more difficult to compute from past data itself, then: our empirical knowledge about the potential contribution — or role — of rare events (probability x consequence) is inversely proportional to their impact. … We understand so little about catastrophic events, yet these are the events that we talk about the most casually. In risk management terms, the bigger the event, the less we have a clue.
Risk management should attempt to identify and minimize the effects of high-cost events, irrespective of their probability of occurrence. High-probability events are relatively easy to spot, so most efforts are concentrated on low-probability events. In practice, risk analysts "…confidently extrapolate from the seen to the unseen…"6 without bothering to establish their equivalence. Perhaps system safety practitioners should bother themselves with the equivalence issues of the inputs.
|
)
)
