Friston's Free-Energy Principle, Carefully

Have you ever encountered a Free-Energy Principle post online and quietly wondered whether the person writing it actually understood what they were describing? The suspicion is often warranted. Friston's framework is genuinely important, genuinely mathematical, and genuinely susceptible to being over-extended by writers who like the way it sounds. This piece is an attempt to hold both the value and the limits carefully.

At its mathematical core, the Free-Energy Principle proposes that any self-organising system — a cell, a brain, potentially an entire organism — will tend to minimise a quantity called variational free energy. In simplified language, the system will act so as to keep itself in states that are consistent with its own model of what it should be. Prediction error, in this framing, is one manifestation of free energy. Reducing it is what the system is quietly organised around doing.

The mathematics is peer-reviewed and published across multiple journals (Friston, 2010; Parr et al., 2022) [Hypothesis]. The framework has been productively applied to perception, action, learning, and, more speculatively, to consciousness. The reason the evidence level here is Hypothesis rather than something higher is not because the mathematics is questionable. It is because the biological interpretations layered onto the mathematics — that specific neural circuits implement this specific computation, that behaviour and cognition can be explained through this lens — remain actively debated in the field.

What the framework is genuinely useful for is worth naming. It provides a unifying mathematical vocabulary for phenomena that previously required separate theories. Perception, action, learning, and homeostasis can each be described in the same terms, which is intellectually powerful whether or not the biological implementation ultimately turns out to be exactly as proposed. And it has generated concrete predictions that have been tested in imaging studies, which is what a productive scientific framework should do.

What it is less useful for is holding as a settled fact about how minds work. The temptation, particularly in popular writing, is to treat the Free-Energy Principle as if it were an established explanation of consciousness or emotion. That is not what the mathematics claims. The mathematics is a framework — a way of thinking. Whether the brain literally implements it in the way sometimes described is an open empirical question that the field has not resolved.

A useful way to hold the framework is as scaffolding. It gives you a productive vocabulary for asking questions about how a nervous system might be organised. It does not give you the final answers, and anyone claiming otherwise is running ahead of the evidence.

For the empirical core of predictive processing, see The Brain Is a Prediction Machine. For the concrete visual-cortex evidence base, Predictive Coding in the Visual Cortex is the piece to read next.