The Brain as a Self-Organising System
How To Read/Listen
This piece begins with a brief overview of brain function, then introduces new research that challenges traditional models of behaviour.
If you prefer to skip theory, go straight to the Key Findings.
Suggested path:
- Read Key Concepts & Definitions
- Listen to the podcast (optional)
- Read the main article
- Note the key findings
How NOT to Read
This is not a quick scan. The ideas here unfold through careful attention and benefit from time.
If time is limited, read only the Key Findings.
Key Findings
- The brain is a self-organising system that maintains coherence by distributing belief.
- The brain is not optimised for correctness, but for survival under uncertainty.
- The brain has no single control centre.
- Loops, not commands, maintain coherence.
- Attention biases prediction, not data.
- Prediction enters the system before perception finishes.
If you are short on time or prefer not to engage deeply with theory, the Key Findings section (above) stands on its own. Otherwise, begin with Key Concepts – Definitions (below) then return here.
The Brain as a Self-Organising System
This piece explores a growing shift in how neuroscience understands brain function and behaviour. It begins with a brief orientation to predictive brain models, then introduces new brain-wide evidence that challenges long-standing hierarchical assumptions about perception, decision-making, and action.
For decades, neuroscience searched for control centres in the brain: places where decisions are made, beliefs are stored, or actions are commanded. The dominant intuition was hierarchical—sensory information flows upward, cognition happens somewhere in the middle, and motor plans flow downward.
A new wave of large-scale neural evidence challenges this picture.
Using brain-wide recordings in mice performing a simple decision task, researchers have shown that prior beliefs are represented simultaneously across sensory, associative, and motor regions, linked together by dense, recurrent feedback loops (Nature, 2025). The brain does not wait for information to climb a hierarchy before belief enters the system. Belief is already everywhere.
This finding lands squarely within Active Inference, a theoretical framework derived from the Free Energy Principle.
Key Concepts – Definitions
| Concept | Definition |
|---|---|
| Active Inference | A framework in which brains perceive and act to minimize expected surprise by maintaining predictive models of the world. |
| Free Energy (Variational) | A formal measure of prediction error or surprise; minimizing it keeps internal models aligned with the environment. |
| Prior | A belief about the likelihood of states or outcomes before new sensory evidence arrives. |
| Prediction Error | The mismatch between predicted and actual sensory input. |
| Entropy | A measure of uncertainty or variability in possible states of the world or internal models. |
| Heterarchy | An organisational structure with distributed, context-dependent influence rather than fixed top-down control. |
| Action-Based Prior | A belief updated primarily from recent actions rather than abstract task rules or stimuli. |
| Controlled Hallucination | The view that perception is internally generated prediction constrained by sensory input. |
Key Findings
The Emerging Picture
Taken together, these findings point to a coherent but non-hierarchical view of brain function. The brain does not minimise chaos by enforcing rigid control; it manages surprise by distributing belief, maintaining feedback loops, and acting to preserve viability in an uncertain world.
Perception, action, and cognition are not separate stages but expressions of the same predictive process. Intelligence, in this light, is not the elimination of uncertainty, but the ability to live productively within it.
Potential breakthroughs
Potential Breakthroughs
Rethinking Psychiatric Disorders
Many conditions can be reframed as failures of belief distribution:
- Psychosis: priors become too strong or circulate uncontrollably
- Depression: priors become rigid and slow to update
- Anxiety: threat priors dominate early sensory processing
This points toward interventions that target network dynamics, not single regions.
Rethinking Attention, Perception, and Agency
If priors are embedded in sensory cortex, then:
- Attention biases prediction
- Perception is controlled hallucination constrained by input
- Action is belief confirmation through movement
Classic boundaries between seeing, deciding, and doing dissolve.
Implications for Artificial Intelligence
Most AI systems rely on centralised objectives, static hierarchies, and explicit optimisation targets. The brain instead uses:
- Distributed belief maintenance
- Loopy inference
- Action-conditioned updating
These findings strengthen the case for active-inference-based AI, potentially moving beyond current deep-learning paradigms.
Bottom line
The brain does not eliminate uncertainty; it negotiates it. Belief is distributed, control is provisional, and coherence emerges through feedback and action. Intelligence is not correctness, but survival under uncertainty.
Member discussion