The Taste Monopoly: When Algorithms Become Gatekeepers

For most of the twentieth century, cultural gatekeeping was a human problem. Editors, programmers, buyers, bookers, fallible people with explicit taste, discernible prejudices, and at least in principle someone to argue with. The algorithmic turn promised to dissolve the gatekeeper in favor of the crowd, taste democratized at last. What arrived instead was a gatekeeper with no face, no address, and no appeals process.

The recommendation engine is not neutral. It encodes the preferences of whoever built it, whoever trained it, and whatever objective function it is optimizing for, which is usually not quality, often not diversity, and almost never the long-term flourishing of the culture it is sorting. Engagement is the metric because engagement is measurable. What engagement selects for, over time, is a kind of stimulation that competes for attention rather than rewards it.

The Long Tail That Never Arrives

The early promise of algorithmic culture was the long tail: infinite inventory would finally surface niche work to the audiences who wanted it. In practice, recommendation systems cluster around existing popularity, creating feedback loops that amplify what is already successful. The mechanism that was supposed to decentralize taste has demonstrably recentralized it, concentrating attention on fewer titles, artists, and topics than the editorial regimes it replaced.

This matters for curation specifically because it forecloses serendipity, the experience of encountering something you did not know you needed because you were not looking for it. Algorithmic recommendations excel at serving the known preference and fail structurally at producing the genuinely surprising recommendation. A system trained on what you liked cannot easily produce what would change you.

The Human Remainder

What survives the algorithm is the human curator operating at small scale: the newsletter, the bookshop, the radio show with an actual person behind it. These are not nostalgic forms. They are structurally different tools, inefficient by design, personal in a way that cannot be replicated at scale, and therefore valuable for exactly the reasons the algorithm cannot compete with.

The taste monopoly is real but not total. The resistance is not technological; it is social. It involves choosing to receive recommendations from people rather than processes, and being willing to accept the inefficiency that entails. That inefficiency is a feature. It is where surprise lives.