On prediction, direct experience, and the difference between living and predicting

Starting point: the phenomenological and the ontological

It began with an old philosophical distinction. On one side stands the phenomenological: how reality appears to us, how we experience it through our senses. On the other, the ontological: what things actually are, the underlying structure of existence, independent of our perception.

The image was that of a spanner and a nut — the phenomenological spanner trying to grip the ontological nut, and never quite closing around it. The standard philosophical reading (since Kant) is that the two never coincide: the Ding an sich remains forever hidden behind the phenomenon.

But the question raised here was sharper: aren’t they fundamentally the same? And the answer that unfolded was neither a simple yes nor a simple no — it became a wholesale revision of what “perceiving” actually is.

The first turn: the error as point of contact

Predictive processing seemed at first to explain that “not-quite-fitting spanner.” The brain continuously predicts what it will perceive, and only the difference — the prediction error — updates the model.

But the decisive reversal was this: it is the direct contact with reality that causes the error. The error is not a defect, not a failure of the model. It is precisely the only moment at which something from outside genuinely enters. The prediction is the screen; the error is the window. Everything predicted smoothly is mere model — only where the model stumbles do you feel the world itself.

With that, the original intuition reappears in a new form: not because there is no gap between experience and reality, but because the gap itself is the point of contact.

The flaw: the model silts up

But here a problem emerged that put the whole construction under strain. The more experience we accumulate, the more finely tuned the model becomes — so finely that the prediction has already absorbed the deviation before it reaches us. The error is worked away so small that the system no longer registers it as error, but as confirmation. And then we automatically take the prediction to be true.

The unsettling part: from the inside, a model-that-touches-reality feels identical to a model-that-only-confirms-itself. Both: low error, high certainty. Expertise and self-deception look the same from within. The window onto reality silts up — not because reality disappears, but because the model has outrun it.

The break with Friston

Karl Friston turns this mechanism into a law of nature. In his free energy principle, prediction is the very definition of life: any system that persists minimizes its prediction error, or else it falls apart. Predicting is then not something the brain does — it is the only thing it can do, from the fetus onward.

But that is exactly where the theory loses its footing. Three objections, each heavier than the last:

1. It is almost unfalsifiable. Friston calls every reaction, every reflex, every innate tendency “prediction.” This makes the word so broad that it excludes nothing — and a concept that explains everything explains nothing. Its breadth is not a strength; it is a blurring. It erases the very distinction that actually matters: the difference between a body that reacts and an installed model that displaces experience.

2. The machine is born empty. Prediction is real — but it has no content of its own. Empty the brain (figuratively) and there is nothing to predict, because prediction is nothing but stored past thrown forward. No content, no prediction. What is innate is at most the capacity to be filled — not the prediction itself. And that capacity is filled from outside. The prediction is, from start to finish, second-hand.

3. Nature runs largely without it. And this is the decisive counter-evidence.

The evidence from nature

If prediction were the necessary basis of survival, nature would be full of it and direct reaction would lose. The opposite is true.

A newborn fawn stands within an hour and runs within a day. It has no stored past, no content, no time to have built a model — and yet it works. It reacts directly, rawly, sensorily to what is there. If it had to wait until it had stored enough to predict, it would be eaten before the machine could deliver a single guess. It survives precisely because it reacts directly and does not predict.

The same held for our hunting ancestors. The hunter lives on direct perception — the shift in the grass, the sound that is off, the scent on the wind. It is precisely the model telling him what ought to be there that blinds him to what is. Whoever predicts too much sees what he expects and misses what actually moves. The wild rewards direct contact and punishes the silted-up model.

The burden of proof thus reverses. It is not direct experience that must prove it is the basis — nature shows it itself, in every animal that survives without content. Friston must explain why, if prediction is so fundamental, life can largely do without it.

How prediction is installed

A child knows nothing. It recognizes no “mother,” because it does not have the concept. It does not know that a feeling is called “hunger.” It sees a face without knowing what it is. There is nothing to fill in, because there is as yet no content — only presence, state, direct and nameless experience.

What the child does do is pure and basic: to see something new and want to grasp it, taste it, smell it. A natural expedition of experience. It undergoes the experience directly and stores it holistically — as one complete whole, not chopped into the separate categories in which we think.

And there the installation begins. Not abstractly, but in the daily imposition of routine upon rhythm:

• We feed every three hours — not when the child experiences hunger, but when our model says it ought to be hungry.
• We want it to sleep at night — not because its body asks for it, but because it fits our schedule.
• When it cries, we “fill in” — we replace its undivided signal with our category.

Step by step the direct, holistic experience is replaced by our grid. And the engine beneath it is belief: we teach the child to believe its parents, to believe the schoolteacher, later to believe the employer — above its own senses. “No, put on a coat, because I predict that it is cold.” The direct experience is overruled, again and again, by the authority of someone else’s model.

After enough years of believing, the brain carries so much content that it can always supply a fairly accurate prediction — and that prediction now automatically takes precedence over reality.

The core

This is not about denying prediction. Prediction is real. It is about what, precisely, is innate and what is installed:

• There is an original state: direct, raw, holistic, sensory — perhaps already instinctive — contact with reality. This state needs no content to exist. An empty brain can still experience directly.
• There is an empty predictive capacity: innate, but contentless. Without fuel it predicts nothing.
• Through upbringing, routine, authority and language that capacity is filled. And once it is full, the prediction begins to overrule direct experience.
• That overruling feels like truth, but it is second-hand past, not present.
• And because it is installed and not natural, it can in principle be seen through and released — not by regressing to an infant who knows nothing, but by knowing that the layer was put there, and loosening its grip from time to time.

Direct experience has never disappeared. The raw contact — the bottom-up, the touch of reality — still comes in. It is only overstated by a top-down prediction that is no longer our own. And what is overstated can be reversed; what is switched off cannot.

Systemic Oculus

This is what the term “systemic Oculus” can finally mean. Not your eye looking at the world. The systemic eye — the grid that was installed in us, through authority, routine and language, and that now looks on our behalf. The gaze that belongs to the organized, controlled life into which we stepped, away from direct experience.

And in an information society that eye is fed from outside: an endless stream of pre-cooked, self-confirming interpretation, which replaces the costly route — seeking out reality itself and letting it correct you — with the cheap one: consuming confirmation. The override is the old mechanism. The systemic part is who sits at the controls.

The highest thing then attainable is not a return to the infant. It is: to see that the gaze was installed — and to dare, now and then, to let it go, so that reality may come in raw again.