Monday, May 24, 2004

The Precautionary Principle or Pascal's Wager

I was thinking - for no particular reason - about the "Precautionary Principle" propounded by environmentalists and others. Essentially it says that we should err on the side of "safety" when introducing (eg) a new technological development that might harm the environment. Instinctively it certainly seems right but is it strictly rational? ..and how far should we take it? Suppose a development is going to yield economic benefits worth $1 million but has a small chance (p) of doing $1 billion worth of environmental or other damage - shouldn't we (rationally) accept the development if p is less than 1 in 1000? Or does the Precautionary Principle mean that we show set a lower threshhold of say 1 in 2,000? - or 1 in 10,000? - or 1 in 1 million?

Pascal's wager similarly uses an apparently rational risk benefit analysis to settle a non-rational argument. It states that believing in God is a "wager" with an infinite payout (eternal salvation versus eternal damnation) and that we should therefore do so regardless of how high we consider the odds against it. The problem with this argument is that it can be applied just as easily to a belief that Satan is the real god and Christian's are going to hell, or that God exists but gets really cross with people who insist on believing in him in spite of all the evidence he left to prove he didn't.

Perhaps the real reason behind the precautionary principle lies in Greyshade's Law which states that

"No credible risk analysis can contain any assessed probability which is materially lower than the probability that the assessment is a load of bollocks"

Along with its first footnote

"The incidence of clinical schizophrenia is about 1 in 100. The probablility that any one belief you (or anyone else) may hold at the moment is in fact a paranoid delusion is materially greater than one in a million."

So maybe we should -
(1) use rational risk benefit analysis to guide decisions but look really closely at the underlying assumptions and science
(2) pay particular attention to claimed low probabilities - lightning does strike twice in the same place amd the chance of 1 in 100 year floods in consecutive years is much greater than 1 in 10,000.
(3) make sure that no material factor has been left out but disregard unreasonably remote possibilities - if you try to include everything that anyone mighi or ever could possibly consider you will never get a sensible result.