The Wiki Man | 27 March 2010

I don’t know if you have ever been to Paris, but it’s basically a kind of London for girls. I generally try to avoid the place, as you can’t get a decent curry and there’s nothing in the shops unless you are an anorexic dwarf. But a couple of times a year I used to find myself on a breakneck taxi ride between the (pretty crummy) Gare de Lyon and the (crummier still) Gare Montparnasse, trying to catch a train to Bordeaux.

Only after my sixth visit did I learn that there was no need for this ludicrous trip. Instead of changing trains in Paris, you could simply hop off the Eurostar at Lille, have some chips and a pint at the bar, then jump on an almost empty TGV train which runs all the way from Lille to Bordeaux. Simples.

Why did it take me years to find this out? Because, as John Kay writes in his groundbreaking new book Obliquity, computers are incapable of oblique thinking.

Type London to Poitiers into the SNCF website and it will always route you via Paris. Why? Because, in purely chronological terms, the journey via Paris is about 30 minutes quicker. Never mind that it involves a ton of hassle (and expense) changing between crappy stations some miles apart — the computer which chooses your journey for you only has one metric in mind: not comfort, convenience or avoidance of hassle, just speed.

The computer makes the decision directly. It has its measure, and from that it derives what it believes to be a single ‘right’ answer. The intelligent human makes a travel decision (to use Kay’s term) obliquely. The reason the human makes a better decision here is that he knows that, when making a journey, there can be no single, scalar metric which answers the question ‘how should I get from A to B?’ The answer depends on a mixture of incomparable (and often conflicting) factors such as speed, hassle, comfort, price, luggage, time of day and personal preference.

Now there are many problems which can (and often should) be solved by direct thinking methods, and here technology performs wonders. These are generally problems where the question is unambiguously defined, and where there is a single right answer. However, in the years since the invention of the spreadsheet (a technology whose dangerous side-effects have passed almost entirely unremarked) we have developed a business and government culture which has confidence in decision-making only when the decision is represented as having been reached by direct (i.e. inhuman) methods. This obsession with apparent logic requires that any concept be represented numerically as a single value — even when this is a nonsense. Hence, like the SNCF website, we see a whole range of single metrics (shareholder value, patient waiting times, inflation, journey time, A-level pass rates) suddenly acquiring an excessive power, massively distorting the solution of the real problems at issue.

The brutal truth is that most human, social or collective problems can never be solved by pursuit of a single metric. And the metrics we choose are dangerously dependent on the person who gets to define them. To an economist, a poor man is anyone earning less than 60 per cent of median household income. To a psychologist, a poor man is anyone who earns less than his wife’s sister’s husband.