My father once attended a dinner for 12 where an intense argument arose over splitting the bill. After 15 minutes, a happy compromise emerged, at which point one diner, someone an order of magnitude richer than everyone else at the table, piped up with: ‘But I didn’t have any wine.’ This is the perfect example of a valid consideration being introduced at completely the wrong point in the process.
Bear this in mind, because I think there is an optimistic and relatively benign way to look at Britain’s bureaucratic sclerosis without hanging every public-sector employee from their own lanyard.
Provided a decision-making process feels rational, we don’t notice the prospective opportunities it denies to us
According to this more benign diagnosis, it’s not that Britain’s bureaucrats are intrinsically useless or ill-intentioned. It’s simply that, to maintain the pretence of objectivity and hence to protect against criticism or litigation, institutions increasingly choose to hide behind badly ordered, sequential processes designed to look ostensibly rational, while the end result is dumb, slow or impossible.
Why was the speed of HS2 determined first and the route determined second? The two factors are clearly interdependent. It may look organisationally neat to define precisely what you are optimising in advance, but in the real world the best solutions typically emerge through iterative exploration, not some linear process.
In medicine, this path-dependency problem is known as ‘diagnostic momentum’ – the risk that the initial diagnosis a patient receives has a disproportionate effect on subsequent treatment. If you imagine a hospital where every patient was first examined by a dermatologist, you can see why this might cause problems. But no one seems to have noticed that, in government, if everything must be first modelled and vetted by the Treasury, you will end up with a similar distortion in decision-making. Similarly, in the private sector, functions like HR, finance, compliance or procurement have become proscriptive.
Often the more you compartmentalise a decision-making process in order to appear rational, the less effective it becomes. One of the reasons I fervently defend the jury system is that a jury is under no obligation to explain their verdict, which gives them a freedom to wield intuitive skills denied to those who must rationalise a decision.
In the same vein, two decision-making tools used by millions of people may be massively flawed. One is the dating app; the other is the property search website. Both mimic intelligent decision-making processes while being deeply biased.
Bonded by Evolution, by the evolutionary psychologist Paul Eastwick, provides abundant empirical evidence showing that the initial criteria people use online to pick a romantic partner have almost no predictive value as to the value or duration of a real-world relationship.
For instance, two heterosexual men, being shown images of two women, will agree 75 per cent of the time on which of the pair is more attractive. On becoming socially acquainted with these people, however, the figure falls to 53 per cent.
(I once had a highly intelligent American colleague who was rejected for dates on Tinder because she liked Hawaiian pizza. Few people are more disgusted than I am by the idea of pineapple on a pizza, but I don’t think it need be a romantic deal-breaker.)
Rightmove, which is also a dating app of sorts, also seems to demand that you limit your selection of property on a narrow set of arbitrary criteria: place, price, number of bedrooms and so forth. Yet how well do these predict the lasting enjoyment you will derive from a property compared with something really important, like proximity to a really good pub?
Yet provided a decision-making process feels rational, we don’t notice the prospective opportunities it denies to us. And in institutional decision-making, the consequences of a decision matter less than your ability to defend how you arrived at it. AI may free us from this to some extent, but it all depends who writes the question.
Comments