A note before you read: this essay draws on ideas from my new book, Attention Thief, which is available free at the end of this post. But the argument here goes further than the book — it’s where my thinking has landed since.
There is a moment most people have had but rarely examine: the moment you realize you have driven the last twenty minutes without any memory of doing so. You were at the wheel. You were technically conscious. You did not crash. But you were not, in any meaningful sense, there.
We usually explain this as autopilot — the brain’s capacity to handle familiar tasks without conscious supervision. But I think that explanation, while accurate, understates what’s actually happening. It’s not just that you handled the driving automatically. It’s that there was nothing to be conscious of. The route was known. The car was known. Every turn was predicted before it arrived. And where everything is predicted, there is nothing for consciousness to do.
This is not a minor observation about highway hypnosis. It is, I think, a fundamental fact about the nature of mind — and once you see it clearly, it changes how you think about what it means to be alive.
Consciousness as Resolution
Claude Shannon, working on the problem of information transmission in the 1940s, landed on a definition of information that is still, eighty years later, the most precise one we have: information is the reduction of uncertainty. More specifically, the information carried by an event is a function of how unexpected it was. A coin that always lands heads carries no information when it lands heads. A coin that lands heads half the time carries one bit of information per flip. The unexpected is informative. The certain is not.
Apply this to the brain, which is now understood to be — at its core — a prediction machine. At every level of processing, from the primary sensory cortices to the highest reaches of prefrontal cognition, the brain is generating predictions about what will happen next and comparing them against what actually happens. The gap between prediction and reality — the prediction error — is where information enters the system. It is where the model gets updated. It is where learning happens.
And it is, I want to argue, where consciousness happens.
Consciousness, understood this way, is the system’s response to prediction error — the resource deployed when the model cannot handle what’s in front of it automatically. When the drive is familiar and nothing unexpected occurs, the model runs smoothly and consciousness is barely recruited. When something goes wrong — a car pulls out unexpectedly, a strange sound comes from the engine — consciousness arrives with full force, instantly, because the model has encountered something it cannot process on its own.
This is why the narrowness of conscious attention, which is often treated as a limitation, is actually a design feature. Consciousness is narrow because it is expensive — it represents the full deployment of the system’s most sophisticated processing resources. You deploy those resources where they’re needed: at the edges of the model, where prediction fails. You don’t deploy them everywhere, because everywhere else the model is handling things just fine without them.
The Edge of the Model
Here is what follows from this, and what I find genuinely startling when I sit with it: you are only fully conscious where you don’t already know.
Not partly conscious. Not dimly conscious. The depth and richness of conscious experience — the quality of genuine presence, of being truly there — is a function of how much the model is being challenged. Where the model runs smoothly, experience flattens. Where the model is being broken and rebuilt, experience intensifies.
Think about the last time you were genuinely surprised — not startled, but surprised in the deeper sense of encountering something that didn’t fit your existing understanding of how things work. The surprise recruited your full attention. The world became vivid. You were, in that moment, more conscious than you are during most of the hours of most days.
Or think about falling in love. The experience is overwhelmingly vivid — colors seem brighter, time moves differently, the world feels charged with significance. Part of what’s happening is the obvious: elevated arousal, the reward system firing. But part of what’s happening is epistemic. Another person is the most complex system you will ever try to model. A new person — a person you don’t yet know — is a source of prediction errors at every turn. Every conversation violates some expectation. Every reaction reveals something the model hadn’t anticipated. You are, in the presence of a new person you find fascinating, operating at maximum model uncertainty. Consciousness is fully recruited. You are fully there.
And then — slowly, over months and years — the model improves. You learn the person. Their responses become predictable. Their humor, their anxieties, their patterns become known. The prediction errors decrease. And something that people often misidentify as the fading of love is, at least partly, the completion of a model. The consciousness that the relationship once recruited so fully has less to do. The world that was vivid becomes familiar.
I am not saying love is just curiosity, or that long relationships are necessarily impoverished. I am saying that the vividness of early love is a cognitive phenomenon — the phenomenon of maximum model uncertainty — and that understanding this has implications for how you sustain it. You sustain it by keeping the model incomplete. By continuing to put yourself in situations where the other person can surprise you. By asking questions you don’t already know the answer to. By resisting the comfort of having someone fully mapped.
Why Children Are So Alive
Watch a young child encounter something new — a puddle, a dog, a word they’ve never heard. The attention is absolute. The engagement is total. There is nothing held back, no part of the mind elsewhere. Adults observing this sometimes feel a wistful recognition: that quality of presence, they sense, is something they’ve lost.
What they’ve lost is not capacity. It’s frontier. The child’s model is small and young and constantly overwhelmed by the complexity of the world. Nearly everything is at the edge of the model — nearly everything is a source of genuine prediction error. The child is, in the precise sense I’ve been using, almost continuously conscious in a way that adults rarely are. Almost everything is informative. Almost everything recruits the full deployment of attentional resources.
The adult’s model is large and well-calibrated and handles most of what the world delivers automatically. The adult drives to work and doesn’t remember the drive. Eats lunch while reading. Attends meetings while planning. Most of experience passes through the model without leaving much trace — not because the adult is distracted or inattentive, but because the model is doing its job. Handling the known so that consciousness can be reserved for the unknown.
The tragedy is not that we grow up. It’s that we stop finding new frontiers. We stop putting ourselves in situations where the model is genuinely challenged. We optimize for comfort, for the smooth running of the model, for the reduction of prediction error as an end in itself. And we wonder why the years seem to accelerate, why life feels thinner than it once did, why we can look back on a decade and find it strangely empty of vivid memory.
Memory, it turns out, is also a function of prediction error. We remember what surprised us. We remember what violated the model and forced revision. A year in which nothing surprised you — nothing genuinely, deeply surprised you — is a year that will leave few traces.
The Frontier Turned Inward
There is a practice that seems, on the surface, to contradict everything I’ve been arguing. Meditation — specifically the kind that involves sustained attention to the breath, to the body, to the texture of present-moment experience — would seem to be the opposite of model-challenging novelty seeking. It is quiet. It is still. It is, deliberately, the same experience repeated.
But experienced meditators report something that makes perfect sense once you understand the model: the practice discloses complexity that inattention concealed. The breath, attended to carefully, is not the same from one moment to the next. The body, scanned with genuine curiosity, reveals textures and sensations that are, to the untrained attention, simply not there. The practice does not eliminate prediction error — it moves the frontier inward, into domains of experience that the ordinary model treats as background noise.
This is the deepest version of what the contemplative traditions are doing. They are not training people to want less, to expect less, to be satisfied with the familiar. They are training people to find the edge of the model in places where it is always available — in the texture of the present moment, in the space between thoughts, in the body’s continuous and largely unattended stream of sensation. The frontier is always there. Most of us just don’t look for it there.
A Life at the Edge
The thesis I’ve been circling is this: the quality of a conscious life is largely determined by how much of it is spent at the edge of the model — in genuine encounter with what is not yet known, not yet predicted, not yet absorbed into the smooth running of an adequate understanding.
This is not an argument for recklessness or constant novelty seeking. The model matters. Competence matters. The expert who has built a sophisticated model of their domain experiences the edge of that model differently from the novice — more precisely, more richly, more productively. You have to know a lot about jazz to be genuinely surprised by what a great improviser does. You have to understand the conventions of a literary form to feel the full force of what it means to violate them.
The argument is for keeping the model unfinished — for resisting the temptation to declare yourself done with a domain, a person, a question, a practice. For finding the places where your predictions fail and lingering there rather than retreating to ground where the model is adequate. For choosing, when the choice is available, the experience that will revise you over the experience that will merely confirm you.
A consciousness that illuminated everything would resolve nothing. We are conscious at the edges of our models. The examined life is not the life that has arrived at all the answers. It is the life that keeps finding new questions — and stays, as long as it can, in the vivid, uncertain, fully conscious space of not yet knowing.
These ideas are developed at length in my book Attention Thief, available free as a PDF and audiobook. No signup required.