Predictive CodingThe prediction-error account asks how a mismatch between expectation and input changes the next expectation.

What It Is

Someone can know that a reward was deliberately paired with going to the gym and still learn to anticipate that reward. Their conscious explanation of the arrangement and their learned response to it are two separate things, and the response does not wait for the explanation.

Predictive coding is a theory from neuroscience that offers a way to see why. It says that the brain predicts sensory input, compares the prediction with what arrives, and uses the mismatch to update its model, and that the same loop runs in every part of the brain that handles information.

The prediction is issued before the sensory data arrives. When the input comes in, it is compared with that prediction. A match needs nothing further; a mismatch is a surprise, and the surprise prompts an update. The cycle runs by itself, without being started or supervised.

In the theory, conscious perception is the prediction that best explains the sensory evidence. Error correction refines the model until the mismatch is small, so that what you experience is the best-fitting prediction and the raw sensory input serves to correct it.

This is the functional loop described in the article, not a drawing of cortical layers or their timing.

Read this diagram

Generate a prediction → Receive the input → Compare the mismatch → Update the model → Generate a prediction.

The Physical Architecture

The theory describes two streams of signal running in opposite directions. Predictions travel down, from the parts of the brain that deal in larger patterns toward the parts closest to the senses. Evidence travels up from the senses. Where the two meet they are compared, and the mismatch is what gets passed further up:

Top-down (predictions):
  What is expected → sent toward the senses, ahead of the input

Bottom-up (sensory + errors):
  What the senses report → sent upward

Error computation:
  Compare the top-down prediction with the bottom-up input
  Mismatch → Error signal propagates up

In the theory these are physical pathways, and the same arrangement is repeated wherever the brain processes information. Vision, sound, touch, motor control, and abstract thought are the same loop applied to different data. A cue that is followed by a reward can be treated as one more pattern for that loop to predict.

The Critical Timing

A prediction has to arrive before the input it predicts. That order is what makes one signal a prediction of the other:

First:  The prediction arrives
        Sets expected pattern

Then:   The sensory input arrives
        Actual data about reality

Then:   The two are compared
        If match → nothing more to do
        If mismatch → an error signal goes up
                      and the prediction is updated

Simultaneous streams would provide parallel information without that predictive relationship. The gap in time makes one signal a prediction about the other and gives the pair a direction.

Learned associations have the same order. Touching a hot stove and feeling pain a moment later allows the touch to be tied to the pain. Opening DoorDash and seeing a redirect a moment later provides another pairing through which a new association can be learned. The order supplies the "X CAUSES Y" relationship: the earlier event comes to stand for the later one.

Conscious Knowledge vs Learned Circuits

What a person can explain and what they have learned to expect are changed in different ways. Someone can therefore hold an accurate account of a habit while the habit carries on as before.

Call a learned association a circuit, after wiring: once it is in place, the cue leads to the expectation with no decision in between. The word is used for that directness. It makes no claim about what the link is made of.

Recognizing a redirect screen does not remove the association created by a long history of encounters with the DoorDash icon. Knowing that jello is an artificial gym reward does not prevent the pairing from forming once it is repeated.

Conscious reasoning can change the circumstances in which those circuits operate. It can select an environment that avoids the stimulus or arrange a new sequence of cues and rewards. It can also override the response momentarily through substantial effort, but that effort is expensive and hard to keep up.

Prevention architecture uses the first option. Altering the environment prevents exposure to the stimulus, so the old circuit stops being rehearsed while alternative pairings build a new one.

Circuit Formation Through Temporal Pairing

A circuit is built by living through the pairing: the cue, then the reward close behind it, many times over. Understanding an association intellectually does not perform those repetitions. Each time the reward follows promptly the link gets a little stronger, and one occasion or a handful is not enough.

The requirements concern the actual sequence and response:

Take a gym-and-jello pairing. During Days 1-30, the jello supplies the repeated temporal association:

Gym completion → Jello reward (2 minutes later) → The reward is felt
Repeat daily × 30
Aim: Finishing at the gym comes to predict the reward

During Days 31-70, visible progress can begin supplying a natural reward:

Gym completion → Visual progress (mirror) → The reward is felt
Old circuit: gym → jello (still present)
New circuit: gym → visual progress (forming)
Aim: Phase out the artificial reward

The original jello association remains while the visual-progress association forms, allowing the artificial reward to be phased out. Conscious awareness that "I'm engineering my reward circuit" does not change the learning. The circuit updates through what follows what, and how soon, across the encounters.

Why You Can't Think Your Way Out

Thinking and learned associations update in different ways:

Knowing that phone checking is undesirable changes what a person would say about it. The learned association between the phone and its rewards is untouched by that knowledge, and it acts faster than the reasoning does. The conflict reflects that separation rather than a failure to articulate the right conclusion.

Knowing that a cue and a reward are being timed on purpose does not stop the pairing from being learned. The association can be changed, but that requires actual repeated exposure to the new temporal pairing. An AI can explain the mechanism; it cannot go through the repetitions on someone's behalf. Repeating the new sequence is the process that changes the circuit.

Known vs Unknown Unknowns

A model can represent a domain without knowing much about it. That provides a place for uncertainty: predictions about the domain miss often and by a wide margin, and questions such as "What is X?" or "How does Y work?" can be formed. This known unknown produces an active uncertainty signal.

An unknown unknown has no representation in the model at all. There is no parameter slot in which to express uncertainty and no domain from which to generate a question. The absence produces no error signal, so nothing marks the gap.

Search queries come from the model already available. A person cannot deliberately search for a domain whose existence they do not know. Mentors, communities, and customers provide external models that can expose those missing domains.

Isolation therefore narrows the space of hypotheses the person can form. Customer development supplies encounters with gaps that introspection cannot reveal, because the customer's observation introduces something the founder had no representation for.

Related Concepts

• Neural Positivism describes the brain's use of positive signals.
• Computation as Core Language describes computation as a language for mechanisms such as this loop.
• 30x30 Pattern describes a month of daily repetition as the way to form a circuit.
• Prevention Architecture replaces signals through environmental design.
• Nature Alignment addresses the limits of consciously overriding circuits.
• Information Theory describes prediction error as an information signal.
• Startup as a Bug explains why discovering unknown unknowns requires external contact.

Key Principle

An explanation of a learned association and the experiences that change the association do different work. Deliberately arranging a cue and reward changes what can be learned from the next encounter; repeating that encounter, with the reward close behind the cue, supplies the process that an intellectual explanation alone cannot perform.