Probability triggers

There’s a link between the question of whether John Terry should be England captain and what to do about Iran’s possible development of nuclear weapons. Both raise the question: upon what level of probability should we act?

This level varies according to circumstance. At one extreme we have the criminal law, which requires that a defendant be proved guilty beyond reasonable doubt. This means we only act upon very high levels of probability - though how high is not (pdf) clear.

But in other cases, we apply the precautionary principle - a small probability of danger is sufficient to act to prevent harm. And in financial markets, a 1-2% probability suffices to trigger action - for example when the “tail risk” of a default or financial crisis causes investors to dump a financial asset.

The fact that, depending on context, we use very different probabilities tells us that our choice of “probability trigger” rests not upon epistemological grounds but rather upon cost-benefit ones.

For example, because we consider the cost of imprisoning an innocent man very high, we set the probability trigger high in criminal cases. Similarly, the cost of default might be so high that it is worth acting to avoid the small probability thereof, so we set the probability trigger low.

And the same might apply to Iran and nuclear weapons. Even if there is only a small probability of Iran getting a nuclear weapon, the cost of this might be so high as to justify military action now to prevent it.

All of this might seem trivial. It’s just expected utility theory.

However, there are some non-trivial effects here.

One problem is that there’s a trade-off here between efficiency and justice. In criminal justice cases, requiring proof beyond reasonable doubt allows some guilty people to go free. This is justified by Blackstone's formulation: “better that ten guilty persons escape than that one innocent suffer”. The converse can also apply. Expected utility might justify a pre-emptive strike on Iran’s nuclear facilities, even if this entails the injustice of killing some innocent people.

And here, two other complications enter.

One is that we often misapply probability triggers. It has become a cliché to say that Terry is innocent until proven guilty. But this is wrong. There is no reason to apply the standard of criminal proof in this context. Stripping a man of the England captaincy is very different from depriving him of his liberty. We don’t need proof that Terry is a racist to remove him from the captaincy. The mere possibility that a man is racist is - when combined with his other character flaws - sufficient to deny him the largely honorific position of England captain. Alan Green - a man I usually have no time for - sees this better than others.

A similar error sometimes happens in discussion of climate change policy. We don’t need proof that man-made global warming is happening to justify reducing carbon emissions. All we need is for the cost of reducing emissions to be smaller than the probability-weighted cost of a climate disaster.

Secondly,  our thinking about these issues is clouded by countless cognitive biases. One particular one here might be the zero-risk bias - our preference for reducing risk to zero. This might lead us to want “proof” - such as in the Terry case - even where none is needed. It might also bias us towards acting against Iran - because we might want to over-pay for reducing to zero the risk of them obtaining a nuclear weapon.

All I’m saying here is that apparently minor cases such as John Terry and major ones such as Iran’s weapons programme raise awkward issues of both probability and ethics - issues which we are not necessarily well-equipped to deal with.

Comments

[Keith]

My problem with this post is that you do not mention uncertainty. Attacking Iran carries uncertain i.e. incalculable risks as it involves future actions of many parties inside and outside Iran. As in the case of Iraq where most of the dangers seem to have been ignored or just not appreciated by the decision takers. So the comparison to criminal or civil justice is an error. If you also ask what good any attack will do in practice there is no convincing answer. Unless you overthrow the Government and occupy the place mere bombardment wont stop only delay any weaponisation. Trying to control the political situation in large foreign countries with an alien culture so as to induce them to adopt the foreign policy you like is extremely hard to do. Bombardment is likely to make them more eager to weaponise the nuclear material. And even if you did invade how long will you stay and at what human and economic cost to invader and invaded? Put a price or utility value on all of that!

[Tom Addison]

I think people are saying "he's innocent until proven guilty" with regards to Terry not when they're discussing whether he should be stripped of the England captaincy or not, but when discussing whether he should be branded as a "massive racist".

The importance of who's captain of England is overblown enough as it is, and I think Terry's already done enough in his life (as that article from The Mirror shows) to merit being excluded from the list of candidates. The bastard doesn't deserve the satisfaction he derives from being England captain.

[Neil]

I can see that your probabilistic-type analysis may be directly relevant to the question of military action, however I think this is only a subset of the initial question you posed:"what to do about Iran’s possible development of nuclear weapons." The latter is a much broader question involving a far wider range of possible responses.

[Andrew]

Issues which we are better equipped to deal with than anything else on earth though. What's your basis for comparison that produces such a downbeat conclusion?

The disagreements are usually over the perception of level of risk. Not surprising given that the only coherent analysis of probability is the Bayesian/subjectivist one.

[Miguel Madeira]

"This level varies according to circumstance. At one extreme we have the criminal law, which requires that a defendant be proved guilty beyond reasonable doubt. This means we only act upon very high levels of probability - though how high is not (pdf) clear.

But in other cases, we apply the precautionary principle - a small probability of danger is sufficient to act to prevent harm."

The "guilty beyond reasonable doubt" can also be considered an application of the precautionary principle - being the "small probability of danger" the probability of sending an innocent to the jail

[Tim Newman]

For example, because we consider the cost of imprisoning an innocent man very high, we set the probability trigger high in criminal cases

I'm not sure you're using the right term here in "probability trigger". Normally, at least in industry, decisions are made on the basis of risk, which is the product of probability and consequence. There is no probability above or below which a decision will be made, i.e. no probability trigger, but there are defined (and often quantified) risk criteria.