The Problem with Markov Blankets and the Principle of Parsimony

Markov blankets are omnipresent, yet their selection poses a challenge due to the abundance of choices and the lack of clarity on the principles guiding their selection. The free energy literature has not adequately addressed the difficulty of determining the appropriate blanket for a given system. The concept of parsimony and simplicity emerges as a solution to this problem, suggesting that selecting the blanket that leads to the greatest simplification of dynamics is crucial. This principle aligns with the scientific method and the concept of coarse graining in physics, aiming to segregate systems into simpler components. Applying this principle to the flame problem reveals that the blanket of a flame lies just outside the reaction zone where exothermic reactions occur. By choosing blankets based on parsimony and simplicity, one can effectively simplify system dynamics, as demonstrated in the evolution of cell membranes and objects in a system through changing blankets and the associated statistics. Ultimately, the blanket serves as a crucial factor in system identification and object type labeling, where the statistics of the blanket determine the identity of the object.