Definition
Active Inference
Active inferences describes how living beings, from single cells to humans, perceive and interact with their world. It posits that the brain operates as a prediction machine, constantly working to minimise surprise by actively seeking out sensory information that confirms its model of the world.
In essence, active inference is the process of minimising free energy by taking action. It unifies perception and action under a single imperative: make the world less surprising. An agent can act on the world to make its sensory inputs match the predictions of its generative model. This is action.
Example: If you feel cold because you “predict” you should be warm, you act to fulfil that prediction by putting on a sweater. The sensation of being cold is a “prediction” error, and the action corrects it.
Environment
If an agent allows for active inference, it not only optimises its internal state to correctly predict the surrounding environment, but the surrounding environment to match one’s model. This allows the possibility of experimentation, to falsify one’s hypotheses. 1