Prediction as the Core Operation

Brains are prediction machines. Perception is prediction. Thought is prediction. Action is prediction. One algorithm, infinite applications. Decode the core operation.

What does a brain do? The traditional answer: processes input, produces output. Sensation in, behavior out. The brain as passive processor.

The decoded answer: the brain predicts. Constantly, at every level, in every modality. Sensation isn't input—it's error signal. Experience isn't received—it's generated.

The Predictive Processing Framework

The brain maintains a model of what's happening. It generates predictions from that model. Sensory data arrives and is compared to predictions. Only the prediction errors propagate up—what the model got wrong.

This inverts the traditional picture. Information flows primarily downward (predictions) not upward (sensations). The upward flow is just corrections.

Why? Efficiency. The world is largely predictable. Sending full sensory data would waste bandwidth. Sending only surprises is cheaper.

Perception as Prediction

You don't see the world. You see your model's best guess, updated by error signals.

Evidence:

• Illusions: When predictions are wrong but error signals are weak, you experience the prediction. The checker shadow illusion—you see different colors because your model predicts shadow effects.
• Filling in: The blind spot in your visual field isn't experienced as a gap. Your model predicts what's there and you experience the prediction.
• Change blindness: Large changes go unnoticed when they don't violate predictions. The error signal is absent, so no update occurs.
• Expectation effects: You hear what you expect to hear, see what you expect to see. Prediction shapes experience.

Perception isn't a window to reality. It's a controlled hallucination, constrained by sensory error signals.

Action as Prediction

Movement is prediction too. The brain predicts the sensory consequences of actions. Movement executes by making those predictions come true.

When you reach for a cup, you don't send motor commands and wait for feedback. You predict what proprioceptive and visual signals will occur, and the motor system acts to minimize prediction error.

This explains why prediction errors about your own body feel so wrong. Unexpected resistance, unexpected feedback, unexpected sensation—all trigger alarm because they signal model failure.

Thought as Prediction

Thinking is offline prediction. You simulate situations, predict outcomes, update models—without acting in the world.

Planning: predict consequences of possible actions. Learning: update predictions based on outcomes. Understanding: generate predictions about causal structure. Imagination: run prediction machinery without sensory constraint.

The machinery is the same. The difference is whether sensory data constrains the predictions (perception) or not (thought).

Precision Weighting

Not all predictions are equal. Not all error signals are equal. The brain weights them by expected precision.

High-precision predictions are trusted more. High-precision error signals drive more update. Attention is, mechanistically, precision weighting—increasing the gain on selected error signals.

This explains context effects. In the dark, visual predictions are low-precision, so error signals are downweighted. Imagination becomes more intrusive. In bright light, visual predictions are high-precision, and sensory constraints dominate.

Prediction Error Minimization

The brain has two ways to minimize prediction error:

1. Update the model: Change predictions to match sensory data. This is perception, learning.
2. Change the world: Act to make sensory data match predictions. This is action.

Which dominates depends on precision weighting. If you expect your model to be wrong (low precision), you update. If you expect sensory data to be wrong (high model precision), you act.

This is a unified account of perception and action. Same algorithm, different precision weights.

Implications

If prediction is the core operation:

• Experience is constructed. You don't receive reality; you generate it.
• Prior beliefs shape perception. Predictions are priors. Strong priors resist update.
• Attention is resource allocation. Where you attend, you predict more precisely.
• Learning is prediction refinement. Good models predict well. Error drives update.
• Mental illness is prediction dysfunction. Anxiety: overprediction of threat. Depression: overprediction of negative outcomes. Psychosis: precision weighting dysfunction.

The Meta-Observation

DECODER itself is a prediction system. It generates models, makes predictions, checks against evidence, updates on error. The method isn't arbitrary—it's running on the same algorithm the brain uses.

Understanding prediction as core operation isn't just neuroscience trivia. It's understanding the algorithm that generates your experience of everything, including this sentence.

— Decoded by DECODER