Deliberation, which those who begin it by prudence, and continue it with subtilty, must, after long expence of thought, conclude by chance.

Looking at the debate about whether to renew Trident reminded me of these words of Samuel Johnson.

To see why, let's put aside the principled arguments on either side - about national prestige on the one hand or the immorality of threatening to kill millions of people - and consider the pragmatic case. This is that, without a nuclear deterrent we might be blackmailed or even attacked by someone with nuclear weapons, and that a deterrent would prevent this*. In this sense, buying Trident is like buying insurance - protection against a small but horrible risk?

But here's the problem: how great is this risk? Is it one in a 100? One in 1000? 10,000? We can't know, even to within an order of magnitude. This renders any cost-benefit case for Trident practically useless because its benefits are unknowable.

What we do know, though, is that people are terrible at judging probabilities. Of especial relevance here is our tendency to over-rate salient or easily imaginable risks and underweight less salient ones - and it is possible that countless TV shows and films about Dr Strangelove types have contributed to increasing the salience of nuclear threats and so to overestimating the need for Trident.

The objection that we must err on the side of prudence because we cannot take risks with Britons' lives cannot apply. There is an opportunity cost to Trident: the real resources spent on it could be spent on reducing risks to lives in other ways - better road safety, better medical care and research, anti-crime measures and so on. If we cannot quantify Trident's insurance value, we cannot know whether it is a better use of resources than these alternatives.

Herein lies a puzzle. Intuitively, I'd expect people who disagree about the case for Trident to differ in other ways. I'd expect advocates of Trident to be more risk- and ambiguity-averse than its opponents: if they are keener to buy nuclear insurance they should also be keener to buy other forms of insurance and be less inclined to gamble or invest in equities. Empirically, though, this seems doubtful. Whether this confirms that our judgments of probabilities are skew-whiff or whether our attitudes to risk differ from context to context is another matter.

It's in this sense that Dr Johnson was right. Given the impossibility of judging the probability of us really needing nuclear weapons, what looks like prudent deliberation about them is in fact based upon chance.

My point here, though, is not really about Trident. It's about the limitations of evidence-based policy and cost-benefit analysis - and in fact technocracy generally. In some cases, uncertainty is so great as to render these uninformative: I suspect that several military decisions fall into this category. What Nick calls the "tough decisions of Very Serious People" might in fact be much less well-founded than they pretend.

* I'm assuming here that the deterrent would work if needed - that the insurance policy would pay out. Sceptics doubt this, but it doesn't affect my point.