What is a Markov Blanket?

 According to the free-energy principle (FEP), entities like brains or organisms are defined by a Markov blanket — an abstract, statistical skin that separates something from everything else. These blankets aren’t necessarily a real, physical border. To draw them, we need to represent a system, such as a brain, as a network of states. Connected in a network, these states influence each other. In a brain, the state of one neuron might influence the state of another, triggering it to fire. The FEP says that an object persists by changing its internal states (represented by the yellow circle in the graphic above) to create a model about the world beyond, which is represented by external states. The Markov blanket is defined as the smallest set of “blanket states” that can fully predict the internal states. The blanket states are divided into sensory states that receive information from external states, and active states, which usually (but not always) influence external states. Sensing and acting via the sensory and active states allows objects to update their internal states and influence external states to minimise surprise and persist within their environment.