Two Bags, One Feeling: Hume and the Standard of Taste

The two bags sit side by side on the counter. Same leather, same proportions, same discreet hardware. One is from this season, one from the last. The client looks at both for a long time, picks up the second, and says the only thing anyone ever really says at this moment: this one just feels right.

The scene is a composite, but anyone who has worked on a shop floor will recognise it. What is new is everything around it. Somewhere, a system trained on millions of similar moments could have predicted her choice before she walked in. It knows the colours she lingers on, the weight she prefers, the price at which she hesitates. It would have been right.

So here is the question this newsletter begins with. If a machine can predict what feels right to her, does it understand why? And if it cannot, does that matter?

Two words that refuse to sit still

This newsletter is about two words: luxury and taste. Neither will be defined here, not yet. Each comes with a question and a tension.

Luxury asks what makes something worth wanting. Is it the object itself, the hand behind it, or the story told about both? Craft lives inside this question: the atelier, the stitch, the repair that keeps a thing alive for decades. So does status. More than a century ago Thorstein Veblen argued that people buy what others can see them buy. He gets his own issue.

Taste asks something even more uncomfortable. Is it purely personal, or can one judgement be better than another? If it is purely personal, there is nothing to discuss. If one judgement can be better, someone has to decide whose.

Hume's question

In 1757 David Hume published a short essay called "Of the Standard of Taste". He began from the same discomfort. Everyone agrees that tastes differ; everyone also behaves as if some judgements are wiser than others. Nobody seriously argues that a minor poet equals Milton.

Hume's answer was not a rule but a portrait of a person. True judges are rare, he wrote, and their verdict is the closest thing we have to a standard:

To explain delicacy, Hume borrows a story from Don Quixote. Two of Sancho Panza's kinsmen are asked to judge a cask of wine. One tastes leather in it; the other tastes iron. They are mocked, until the cask is emptied and an old key on a leather thong is found at the bottom.

It is a story about luxury as much as about taste. The judge notices what others miss, and is proved right by the object itself.

Even this, though, turns out to be hard. Hume needs five qualities to describe one judge, and he admits such judges are scarce and hard to recognise. Defining taste, it seems, is a matter of taste.

Enter the machine

Now hold Hume's portrait up against a large AI model, and be honest about the result.

Practice? It has seen more images of handbags, paintings and interiors than any critic could see in a hundred lifetimes. Comparison? Comparison is all it does. Freedom from prejudice? Arguably better than ours, in some respects; arguably far worse, since it inherits every prejudice in its training data. Delicacy? It can detect differences in a photograph that no human eye will catch. Strong sense? That depends on what you mean by sense.

On paper, the machine looks uncomfortably close to Hume's ideal judge. If you run a luxury business, that should not reassure you. Recommendation engines, trend forecasting and generated product imagery are all, in effect, claims to judge taste at scale.

And yet something is missing. The kinsmen did not just detect leather and iron. They had drunk wine for years, they had something at stake when they spoke, and they were willing to be laughed at for being right. The model has no body that tasted the wine. It has no reputation to lose and no evening ruined by a bad bottle. It can predict that the second bag feels right. It cannot feel that it does.

Whether that difference matters is not obvious. It may matter enormously or hardly at all. It probably matters in some places more than others. That is what the next issues will try to find out.

What comes next

Each issue will take one place where luxury, taste and technology meet, and ask what should stay human there. It might be the client advisor facing a chatbot, the atelier facing the algorithm, the repair bench, or the question of proof when anything can be generated. Some issues will strengthen the bet behind this newsletter: that luxury and taste only mean something because a human made, chose or judged the thing, and others know it. Some may weaken it.

None of them will end with a conclusion.