We assume perception is a window onto reality as it is. But Oxford neuroscientist Thomas Parr, a close collaborator of one of the most highly cited living scientists, Karl Friston, argues that this assumption has the brain backwards. Under the Free Energy Principle, perception and action are the same process: both work to close the gap between what we predict and what we encounter, either by updating our beliefs or by changing the world to fit them. Curiosity, Parr argues, is as fundamental to behaviour as reward — we act not just to get what we want, but to know our world better.

While reading this article, you are engaging in active inference. With each movement of your eyes, you select the sensory data that best resolves uncertainty about what will be said next. Those visual impressions are synthesized with your brain’s internal world model to form letters, words, sentences, and ultimately meaning. As you hear a sound in the background and turn your head to localize where it has come from, you act to provide your internal model with the information you need to update your percept. In short, perception and action are part of the same process.

The graphic below helps to make the active nature of perception intuitively clear. If you fixate on the cross in the center of the picture, and try not to move your eyes, you will perceive the colours in the periphery fading away. As soon as you allow your eyes to move freely again, the colours re-emerge. This underlines that our world appears the way it does because of the way we choose (consciously or unconsciously) to sample it. Of course, the way we choose to sample the world also depends upon the way in which we perceive it. This reciprocity between action and perception is active inference.

Troxler fading, reproduced under copyright by license from Parr et al, 2019

Active inference is (almost) synonymous with the Free Energy Principle—a theoretical principle developed by neurobiologist Karl Friston to explain how our brains work. It has roots in the work ofnineteenth-century polymath Hermann von Helmholtz, draws from advances made in the twentieth century by physicists including Richard Feynman and Nikolay Bogolyubov, and builds upon the same mathematics used at the turn of the century by pioneers of modern artificial intelligence. While highly influential in psychology and neuroscience, the principle has taken on a much broader role in the life sciences. It is difficult to think of a theoretical programme in biology, since Darwin, of similar scope and ambition.

This level of ambition excites both enthusiasm and scepticism, and rightly so. The Free Energy Principle has been celebrated for its simplicity and criticized for its complexity. Advocates point to the unifying power of the framework, facilitating a common language to solve apparently disparate problems. Critics argue that the principle is ultimately unfalsifiable and therefore scientifically futile.

Simple or complicated?