Markov Blanket

A Markov blanket is the minimal set of variables that statistically separates a node in a Bayesian network from all other nodes outside the blanket. Originally formalized by Judea Pearl, the concept describes a kind of statistical membrane: once you know the state of everything in a node's Markov blanket — its parents, children, and co-parents — the node becomes conditionally independent of everything else in the network. Nothing outside the blanket carries information about what is inside, given the blanket.

In systems theory and Free Energy Principle research, Markov blankets have been reinterpreted as the formal boundary between a self-organizing system and its environment. Karl Friston argues that any system that persists through time and maintains its organization against environmental perturbation necessarily possesses a Markov blanket — the boundary is not just a modeling convenience but a thermodynamic requirement for identity. This move is controversial: critics argue that Markov blankets are always observer-relative, not intrinsic features of the world, and that deriving selfhood from a statistical construct involves a category error.

If Friston is right, every persistent dissipative structure — from cells to brains to economies — is implicitly carving itself off from the world with a Markov blanket. Identity would then be, at root, a conditional independence relation.