From an article in n+1 about the auction house Sotheby’s
Once, long ago, a boyfriend taught me how to spot a cheap dress shirt. “They’re kind of, I don’t know, shiny?” he posited with disdain. Now I cannot look at a broadcloth without squinting for sheen—a wrinkle-resistant treatment, a trace percentage of Lycra. And it wasn’t long into my tenure at Sotheby’s before “good paintings,” like “bad shirts,” began to announce themselves to my eyes.
Art pricing is not absolute magic; there are certain rules, which to an outsider can sound parodic. Paintings with red in them usually sell for more than paintings without red in them. Warhol’s women are worth more, on average, than Warhol’s men. The reason for this is a rhetorical question, asked in a smooth continental accent: “Who would want the face of some man on their wall?”
Here are some more qualities that make a work of art valuable: it is “representative” (it looks like—and was executed in the same era as—other valuable works by the same artist); it has “a good provenance” (important people have owned it before); it is “included in the literature” (critics or historians have written about it in a museum catalogue or book published by a university press). Adolph Gottlieb paintings should have sun-discs in them. Cy Twomblys are best with squiggles. When it comes to Ellsworth Kelly, the more “totemic” the better.
This reminds me of how chick sexers are trained. These are people who pick up newly-hatched chicks and sort them by sex. Seems trivial, but it is supposedly a task which takes years to master. The story goes that novices learn by watching masters do it, and that they aren’t told explicitly what to look out for, but pick up the cues subconsciously. A nice phrase to describe this that I found while looking it up is that it is a skill that’s “hard-earned and not accessible to introspection.” (Though the impossibility of explicitly teaching chicken sexing might be an exaggeration).
Or teaching robots how to choose potatoes. A team from a UK university has built a device to automate the process of sorting potatoes. In this case they’ve chosen a machine-learning paradigm, having human sorters train the device to recognize spoiled potatoes before setting it loose on the potato crop.
Are we at some level like these potato-sorting robots? As humans we like to think that we’re better than our machines because we can think creatively. The task of pattern recognition is straightforward compared to realizing that there is a pattern to be recognized in the first place. How is insight generated? That’s still an open question.