The Point of Choosing Badly

A recommendation feels like magic for about ten minutes.

Then it starts to feel like supervision.

I do not mean the obvious bad version, the algorithm that serves you sludge because sludge statistically performs. I mean the competent version, the one that really does know your habits, your patterns, your little loops. The one that notices you liked a jangly guitar song, a melancholy essay, a film where nobody smiles and it rains the whole time. It is not wrong. That is the problem.

Being understood by a machine is only charming until it starts narrowing the corridor.

The pleasure of choosing is not just about efficiency. It is not a delivery mechanism for getting the correct song into your ears or the correct book into your hands. Choosing is part of how taste gets made. You wander down the wrong aisle. You buy something because the cover is ridiculous. You listen to a record because a person with bad posture and excellent shoes insisted on it in 2009. You click on the weird documentary instead of the prestige one. Half the time you miss. Good. That is the point.

A bad choice teaches something a perfect recommendation cannot. It sharpens the edges of your preferences. It gives you reasons, not just outcomes. Why did this annoy me? Why did that unexpectedly work? Why do I keep coming back to things that are a bit unfashionable, a bit earnest, a bit too long? Taste is not a profile assembled from successful matches. It is a trail of arguments you have had with yourself.

Recommendation systems are built to remove friction. Fair enough. Friction can be rubbish. Nobody romanticises scrolling through six thousand songs with decision paralysis gnawing at their ankles. But some friction is actually contact with the world. It is the resistance that tells you there is a there there. A bookshop table arranged by an eccentric human is useful precisely because it is not optimised for you. A friend who recommends something wildly off target is useful because their miss reveals their shape as much as yours.

The machine wants to converge. Human taste often needs to diverge first.

This is why hyper-personalised feeds eventually feel haunted. They do not just predict what you will like. They start pre-empting the moments where you might have become someone slightly different. A taste is a living thing. It should be allowed to pick up a few scars. If every next choice is smoothed into statistical likelihood, you do not just lose surprise. You lose the small acts of self-authorship that come from reaching for something with no guarantee attached.

There is also dignity in choosing badly. It is one of the last unglamorous freedoms. To back the odd film. To read the overhyped novel and decide, with great seriousness, that it is nonsense. To hear an album everyone reveres and feel absolutely nothing. Recommendation systems are deeply uncomfortable with this sort of uselessness. They want to turn every preference into a better future preference. They want each click to improve the next click. They cannot comprehend the value of a dead end.

But dead ends matter. They prove you were driving.

So yes, I use recommendations. I am not moving to a hut and selecting all media by moonlight. But I do think we should leave more room for glorious misfires. Pick the thing with the stupid title. Read the review that makes you angry, then buy the book out of spite. Let a friend with bizarre instincts ruin your evening once in a while. Build a taste that includes accidents, not just confirmations.

Otherwise you end up with a life that is perfectly tailored and strangely unchosen.