Most people agree that rational behaviour is a good thing, which should be encouraged through education or, if necessary, nudging. However, a new paper by Daniel Pi challenges this. In some contexts, he says, we'd benefit if we were less rational.

Take crime. The rational person weighs the benefit of mugging someone - the financial reward and the buzz of the violence netted off against the feeling of guilt afterwards - against the cost; the probability of being caught multiplied by the punishment.

But we don't really want people to think so rationally because it would lead them to actually mug someone occasionally. It would be better if they had the heuristic "don't mug people." Such a heuristic is, however, irrational in the narrow economistic sense, as it would cause people to reject occasionally profitable actions.

The best policy, then, says Mr Pi, is not to debias people, as nudgeists would like, but to bias them - to encourage the spread of the "don't commit crime" heuristic which, whilst irrational for (some) individuals is socially beneficial.

There's an analogy here with the theory of the second best (pdf). This says that if there is a market failure, welfare can sometimes be increased by moving further away from optimality rather than by moving towards it.In similar fashion, moving away from rationality might enhance aggregate welfare.

This issue might be acute in the field of corporate tax. Companies apply economic rationality in considering how to minimize their taxes, but many people think that such behaviour is detrimental to aggregate welfare. If you agree with such folk - which is another issue - the question then arises of how to increase tax morale, to inculcate a "pay fair tax" heuristic. There are (at least) three possibilities:

- Customer boycotts. These operate on the rational cost-benefit calculus of whether to pay tax, but also help foster the "pay tax heuristic" by encouraging a "doing well by doing good" ethic.

- Arrest and punish high-profile tax-dodgers. This can deter crime through the availability effect; people over-estimate the probability of events they can easily see and recall.

- Introduce an element of arbitrariness into the probability of detection and punishment. If people can't calculate the probability of getting caught and punished, they'll decide to err on the safe side.

These last two can be applied to crime generally.But they run into the objection that it's normally desireable that the law be applied equally and predictably. This, though, just repackages the paradox we began with. Part of the case for having clear and consistent laws is that they permit individuals to make rational plans. If, however, those rational plans are harmful to others, then perhaps rationality is not to be encouraged.