You didn’t tell anyone you wanted it. And yet the recommendation arrived anyway, suspiciously close to a thought you hadn’t fully formed. The common explanation is “it’s listening“; but the more unsettling truth is that it doesn’t need to. It has your pattern, and your pattern turns out to be more consistent, more predictable, and less consciously chosen than most of us would like to believe. This episode explores what behavioral prediction actually reveals about the gap between the self we narrate and the self we actually are and what it means that the most accurate version of “you” might not be the one you have access to.
Podcast: Play in new window | Download (Duration: 11:50 — 29.9MB) | Embed
Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Deezer | Youtube Music | RSS | More
AI Generated Transcript:
There’s a particular kind of unease that comes from seeing an ad for something you wanted, before you’d told anyone you wanted it. Most people have a story like this — a recommendation, a suggestion, a targeted ad that arrived suspiciously close to a thought you hadn’t fully formed yet. And the most common reaction is some version of “it’s listening to me.” But here’s the more unsettling possibility, and the one that’s actually closer to true: it’s not listening. It doesn’t need to. It already knew — not because of anything you said, but because of a pattern you didn’t even know you were part of.
I’m Andrei Barburas, and this is The Unlearning Project — a podcast from barburas.com, where every couple of weeks we take one assumption that usually goes unexamined and we crack it open. Today’s assumption is about self-knowledge — specifically, the quiet assumption that you are the foremost authority on yourself. That nobody, and nothing, could possibly understand your preferences, your tendencies, your likely next move, better than you do — because, after all, it’s your mind, your experience, your interiority. And yet, increasingly, there’s a version of “you” — built entirely from data, from patterns, from behavioral traces — that can predict what you’re going to do with an accuracy that, frankly, most of us couldn’t match if we tried.
Let’s start with what this “digital twin” actually is, because the term can sound more exotic than the reality — though the reality is, in some ways, stranger than the term suggests. It’s not a sentient copy of you somewhere. It’s a statistical model — a profile built from behavioral data: what you click, what you linger on, what you scroll past quickly, what time of day you’re most active, what you buy, what you almost buy and don’t, who you interact with, how long you hesitate before responding to things. None of these data points, individually, seem like they’d reveal much. But in aggregate, across enough data points, over enough time, patterns emerge — patterns that are often far more stable, far more predictable, and far less consciously chosen than we tend to assume about our own behavior.
Here’s the part that I think is genuinely important to sit with: this isn’t really a story about technology being “smart.” It’s a story about how much of what we think of as personal, deliberate, individual choice is actually patterned — shaped by context, habit, timing, and influences we’re not tracking, in ways that are remarkably consistent and therefore remarkably predictable, once you have enough data to see the pattern from outside.
We tend to experience our own decisions from the inside, moment by moment, each one feeling like a small, fresh act of choosing. But a model looking at the same person from outside isn’t seeing individual moments. It’s seeing the shape those moments make over time — and that shape, it turns out, is often far more rigid than the feeling of choosing, in each individual moment, would suggest.
This connects to something worth pulling apart: the difference between self-knowledge and self-awareness. Self-awareness is something like: I know I tend to feel anxious in certain situations, I know I have certain values, I know roughly how I’d describe my own personality. Self-knowledge, in the sense relevant here, is something different — it’s something closer to: given everything about how I’ve behaved, what will I actually do next, in this specific situation, with this specific set of options in front of me? And it turns out these two things — how we describe ourselves, and how we actually behave — often diverge significantly. People consistently report values, preferences, and intentions that don’t fully match their actual behavioral patterns. Not because people are lying, even to themselves, necessarily — but because the conscious, narrative self and the behavioral, pattern-based self aren’t the same system, and they don’t always agree.
A behavioral model doesn’t have access to the narrative self at all. It has no idea what story you tell yourself about who you are, what you value, what kind of person you’re trying to be. It only has access to the pattern. And because the pattern is, in a very real sense, more consistent than the narrative — the narrative shifts depending on context, mood, who’s asking, what image you want to project, even to yourself — the pattern-based prediction can end up being more accurate, in a narrow, behavioral sense, than your own self-description would be.
This is where the discomfort comes from, I think — and it’s worth naming precisely, because the discomfort itself is informative. It’s not really about privacy, exactly, though that’s usually the language used to describe it. It’s something closer to a confrontation with the idea that the “self” you experience — the narrating, deciding, intending self — might not be the most accurate description of “you” that exists. There might be a more accurate one. And it isn’t a person. It’s a pattern, observed from outside, stripped of narrative, stripped of intention, stripped of the story you tell about why you do what you do.
Antonio Damásio’s work is relevant here again, in a slightly different way than we’ve discussed before. Damásio’s research points toward something that runs counter to a very old, very intuitive idea — the idea that there’s a kind of central, conscious “decider” sitting at the controls, weighing options and choosing. What the research increasingly suggests is something messier: that a great deal of what determines behavior happens beneath conscious access entirely — emotional, embodied, automatic — and the conscious narrative arrives afterward, constructing an explanation for a decision that, in some sense, had already been shaped before the explanation was formed.
If that’s true, then a behavioral model isn’t really competing with your “true self” for accuracy. It might be operating on a layer that’s prior to the layer where your sense of self even forms. Which is part of why it can feel so strange when a prediction lands — it’s not that something external guessed your secret. It’s that something external had access to a layer of “you” that your own conscious experience doesn’t have direct access to either.
So what do we actually do with this discomfort? Because I think the common responses tend to miss what’s actually unsettling about it.
One common response is denial — “it’s not that accurate, it’s just coincidence, it’s just good guessing based on broad categories.” And sometimes that’s true — plenty of targeting is broad, generic, more luck than precision. But the denial response tends to break down the moment something lands with enough specificity that “coincidence” stops feeling like a sufficient explanation. And at that point, denial just becomes a less comfortable position to hold, without actually addressing anything.
Another common response is a kind of resigned fatalism — “well, there’s nothing you can do about it anyway, everyone’s data is out there, may as well not think about it.” This response has the appeal of resolving the discomfort by simply not engaging with it. But it also forecloses something that I think is actually useful, which is: thinking about what it means that a version of you exists, externally, that’s optimized for prediction rather than for anything resembling your own interests — and what the gap between those two things actually implies.
Because here’s the thing — your digital twin isn’t trying to understand you, in any sense that would map onto how a person might try to understand you. It’s not curious about you. It’s not trying to help you become who you want to be. It’s a prediction engine, and predictions are useful, commercially, when they predict behavior that can be acted on — usually, behavior that involves you doing something: clicking, buying, staying, scrolling, engaging. Which means the “you” that the model is optimized to know isn’t really all of you. It’s the parts of you that are useful to predict, for purposes that may have very little to do with your own goals for yourself.
This is, I think, where the genuinely useful question lives — not “how do I become unpredictable” or “how do I escape the model,” which I don’t think is realistic or even necessarily desirable. The useful question is something more like: if there’s a version of me that’s more behaviorally consistent than my self-narrative would suggest — what does that version actually want? And is that the same as what I, narratively, want for myself?
Because sometimes the answer is yes — the pattern reflects something genuinely true about you, something your conscious self might even benefit from acknowledging more directly. But sometimes the pattern reflects something more like a groove that’s been worn into your behavior by repetition, by design choices made by systems that benefit from that groove staying exactly where it is — a groove that doesn’t necessarily represent what you’d choose, if you were choosing freshly, rather than repeating a pattern that’s been reinforced, click by click, until it became the path of least resistance.
The discomfort of being “known” by a pattern isn’t really about the pattern being wrong. It’s often about the pattern being right — right about something that’s been quietly shaped, over time, by forces that were never particularly interested in whether that pattern was good for you. And that’s a different kind of discomfort than “my privacy has been violated.” It’s closer to: something has been watching the groove, while I was busy thinking I was choosing.
Here’s the question I want to leave you with. Think about something you do regularly — a habit, a routine, a pattern of behavior — that, if you’re honest, you didn’t consciously choose so much as fall into, and that’s simply continued because it’s familiar. Now imagine a version of you, built entirely from that pattern, with no access to your intentions, your values, or your sense of who you’re trying to be. Would that version of you be someone you’d want representing you? And if not — what would it actually take to change the pattern, rather than just the story you tell about it?
That’s it for this episode of The Unlearning Project. We’ll be back in two weeks with another one. Until then, pay attention to the groove — not just the story you tell about why you’re in it.
