On algorithms

A feed feels like a mirror, but it's a judgment — a running argument, made by no one in particular, about what deserves your next second.

Curation without a curator

Every previous era of media had an editor you could name and, in principle, argue with. The modern feed has none — just a scoring function, trained on engagement, revised continuously, answerable to no one in particular. That doesn't make it neutral; a system optimized for engagement is still making editorial choices, just ones nobody signed and nobody can be held to. The absence of a visible editor is not the absence of editorial power — it's that power made harder to see, and harder to contest.

Legible to the machine, illegible to us

Recommendation systems optimize for signals — dwell time, click-through, a hundred smaller proxies — that only the system can fully see, weighted in ways that shift by the day. You experience the output: a feed that somehow knows what will hold you. You almost never see the optimization target itself, which means you can't really argue with it, only adapt to it. That's a strange kind of legibility problem: the machine understands the pattern of your attention better than you do, while its own criteria stay opaque to you by design.

The mirror that isn't

It's tempting to treat a feed as a reflection of your own interests — you clicked, so you must have wanted it. But a system trained to maximize engagement will also manufacture the interest it then reflects back at you, nudging toward whatever keeps you watching rather than whatever you'd choose on reflection. The feed doesn't just show you who you are; it quietly edits who you become, one small nudge at a time, and calls the result personalization.

Where this connects

The business model underneath all this — why holding attention is the thing being optimized for in the first place — is the subject of On attention. The same delegation of judgment to a system whose reasoning you can't inspect, applied to decisions you make yourself rather than ones a platform makes for you, is the subject of On AI. And the civic version of this — algorithms quietly making decisions that used to be public and contestable — is On technology and society.