Personalized by Attention, Not by Algorithm

"Personalized" is a word with two meanings that have almost nothing to do with each other. The first meaning is prediction: a system studies your data, sorts you into a category, and serves you what the category tends to want. The second meaning is attention: someone watches what you actually do and remembers it, the way a bartender remembers your order or an old friend remembers the argument you had last spring. The first kind feels vaguely creepy even when it works. The second feels like being known. If you have ever gone looking for a personalized story game — an interactive story made for you, not for a segment you happen to fall into — the difference between those two meanings is the whole question.

Here is the short version. A feed personalizes by guessing who you are from data about people like you; a story can personalize by remembering what you did, because you did it in front of it. The first flattens you into a demographic. The second sharpens you into a protagonist.

The rest of this essay is about why that distinction matters more than it sounds like it should.

The first meaning: personalization as profiling

Consider what happens when a recommendation engine "personalizes" something for you. It does not know you. It knows a shadow of you, assembled from clicks: people who watched this also watched that; users in your age bracket tend to abandon videos after ninety seconds; this thumbnail outperforms that one among viewers flagged as similar to you. Then it makes a bet.

The bet is often good. That is what makes it unsettling rather than merely useless. When the feed serves you exactly the thing you didn't know you wanted, there is a small, cold feeling underneath the pleasure — the feeling of having been anticipated. Not seen. Anticipated. There is a difference, and your body knows it even when your thumb keeps scrolling.

Notice, too, what this kind of personalization is for. It is not trying to know you. It is trying to keep you. The profile exists so that the next item can be chosen well enough that you don't leave. You are not the audience of this system; you are its raw material. The "you" it addresses is a statistical composite — a demographic with your name temporarily attached.

And here is the strange result: the more aggressively a feed personalizes, the more generic the experience becomes. Everyone's feed is different, and everyone's feed is the same — an endless, frictionless sequence of things that people like you tend to like. Prediction can only ever offer you the average of your category. It cannot surprise you with yourself, because it has never actually met you.

The second meaning: personalization as attention

Now consider the bartender. She does not have a dossier on you. She has something much smaller and much better: she was there. You ordered the same thing three Thursdays running, and she noticed, and the fourth Thursday she had it started before you sat down.

That gesture lands differently than the feed's uncanny guess, and it is worth being precise about why. The bartender's knowledge is earned in your presence. It is made of things you actually did, in front of her, over time. Nothing was inferred from strangers who resemble you. There is no category. There is only the record of a particular history that the two of you happen to share.

This is the second meaning of personalized, and it is the older one — the one the word was built for. A personalized letter used to mean a letter someone wrote to you, about things that had happened between you. What made it personal was not the accuracy of its guesses but the specificity of its memory.

Attention, unlike prediction, cannot be manufactured in advance. It has to be paid, in real time, to a real sequence of events. Which is exactly why it feels like being known: because it is.

What a personalized story game actually does

Stories are the natural home of the second meaning, because stories run on memory. A plot is nothing but consequences remembering their causes. The reason a good novel's ending lands is that the book has been quietly keeping track — of a promise made in chapter two, a lie told in chapter five, a door left unlocked in chapter nine.

A personalized story game, in the honest sense of the phrase, is one that keeps track of you that way. Not your demographic. Your deeds. You spared the informant when it would have been easier not to, and forty minutes later, when you need a name whispered to you in a corridor, there he is — not because a system predicted that players like you enjoy loyalty payoffs, but because you spared him, and the story was watching, and it remembered.

The distinction shows up most clearly in what each kind of system does with an unusual choice. To a profiling engine, your strange decision is noise — an outlier to be smoothed back toward the category mean. To an attentive story, your strange decision is the single most interesting thing you have done all evening. It is the moment you stopped being "a player" and became this particular protagonist, the one who did that. Prediction discards exactly what attention treasures.

This is what people are really asking for when they search for a story that feels made just for them. Not a story assembled from their data. A story that was present while they lived it — the way being the main character of a story means the story bends around your choices, not around your profile.

And it explains something else: why an attentive story gets more particular the longer you stay, while a feed gets less. Every hour inside a story that remembers adds detail that belongs to no one but you — the grudge you started, the alias you invented, the bridge you burned. Every hour on a feed adds data that makes you easier to substitute. One process accumulates a self. The other dissolves one. If you have ever closed an app after an hour and felt somehow less specific than when you opened it, you have felt that dissolution firsthand — the case for something better than an evening of scrolling is really just the case for spending attention where attention is returned.

Being guessed versus being known

Put the two meanings side by side one last time.

Profiling asks: what do people like you want next? Attention asks: what follows from what you did? Profiling needs you to be predictable; its whole machinery breaks if you are genuinely surprising. Attention needs you to be consequential; it has nothing to work with until you act. Profiling addresses you in the second person plural, whatever pronoun it uses. Attention addresses the singular you — the only entity in the universe who made that exact sequence of choices.

"Personalized," in the first sense, is something done to you, upstream, before you arrive. In the second sense, it is something that happens with you, and cannot happen any other way. One is a mirror angled to show you what sells. The other is a witness.

Nobody needs to be lectured about algorithms; the feed is fine at what it is for. But it is worth knowing that when the word "personalized" gives you that faint chill, you are not being paranoid — you are correctly detecting the first meaning wearing the clothes of the second.

A story that was there when you did it

Runebook is built on the second meaning. It is an AI story game in which the Storyteller runs the world and you act by typing what you want to say or do — and the story remembers what matters, because it was there when you did it. Threaten a rival in a candlelit hall, and the threat does not evaporate when the scene ends; it waits for you, the way consequences wait in novels. How that kind of memory changes a story over hours and weeks is its own subject — there is a longer essay on story games with persistent memory if you want the mechanics of it.

You can play solo or with friends, in your browser, with narration read aloud and scene imagery as the story unfolds. Fantasy, mystery, romance, horror, historical, manga-style, sci-fi — or a custom story spun from your own idea, which is personalization in the oldest sense of all: a story that starts from you on purpose. Leave whenever life interrupts; the story remembers where you were.

No profile. No category. Just a world paying attention to the one protagonist it has.

Create an account and start your first story — it's included. Start your story