A month ago I wanted to travel to Bath for a 60th birthday party. From Kent, this either involves a Tube journey to Paddington or traversing the south-western stretch of the M25, where — in the rare moments you are not in stationary traffic — you have the even worse experience of driving over 10,000 misaligned slabs of ribbed concrete.

But deep in my hippocampus, I remembered seeing a train to Bath on the departures board at Waterloo. This would let me travel from Seven-oaks to Waterloo East, avoiding the Tube. I looked online, but the website denied all knowledge of such trains. It told me to take the Underground to Paddington and board there.

Following a hunch, I searched Waterloo to Bath via Salisbury. Suddenly there they were — several daily direct trains from Waterloo to Bristol via Bath with First Class Advance tickets selling for bizarrely low prices. Yes, the route is slower than via Paddington, but it is cheap, easy, scenic and WiFi-enabled. The algorithm cares for none of this, of course. It shows the fastest journey and that’s that. It buries the Waterloo to Exeter trains for the same reason.

In the same way, French rail websites always make you travel via Paris whereas the smart money changes trains at Lille. True, you might have to wait for an hour at Lille Flandres station, but who cares when there is a bar called Le Palais de la Bière? (I much admire the Flemish approach to gastronomy, which rests on the much neglected insight that, provided you offer good chips and beer, nobody gives a shit about anything else.)

Taken at a simple level, this story contains a handy travel tip. But I think it poses a more fundamental question. Are websites increasingly leading people to make decisions in an identical informational context —and is this a good thing?

In the old days, some people would buy a house because they saw it in an estate agent’s window, others because they noticed a ‘For Sale’ sign on the house, or perhaps spotted an ad in the local paper.

Today, we uniformly use property websites. And because these impose the same search rules on everyone, it creates strange anomalies in the market. For instance, because the websites make you search high-end houses at price increments of £50,000, it seems any house must be listed at that threshold. Nobody searching high-to-low or low-to-high will find an £875,000 house. So property no longer has a price-demand curve —but a price-demand ziggurat. Good flats may be undervalued outside London, because most people filter by property type before searching. Similarly the restaurant market may become boring if TripAdvisor ratings become too influential. (Really good restaurants often polarise taste.)

The genius of markets lies not only in aggregating people’s preferences but also in aggregating choices made in a wide variety of messily different contexts, and with a variety of different decision-making paths. The danger is that if you make the informational context the same for everyone, and the decision-tree the same for everyone, the information the market receives will increasingly reflect that artificially narrow context, so will no longer reveal the wider preferences of market participants. The individual decisions may seem better, but the market will be dumber overall. This question may even have implications for financial markets where increasingly all public companies are being bullied by institutional investors to deliver the same thing.

Among ants, as I recently learned, anomalies in individual ant behaviour combine to produce outcomes which are optimal for the wellbeing of the nest overall. Perhaps it’s the same with humans: markets are clever precisely because people are silly.