Quantum reference frames and the membrane as boundary

The term Markov blanket comes from Bayesian networks. Friston, Levin, Sengupta and Pezzulo (2015) describe it as a set of states that separates two other sets in a statistical sense.³ It consists of the children of a set, its parents and the other parents of its children.³ The blanket partitions states into internal and external ones. External states can only be seen vicariously by internal states, through the blanket.³ The authors give the cell surface as an example, separating intracellular from extracellular states.³ They also split the blanket itself into sensory and active states.³

Friston (2013) noted that a system can have many partitions and blankets. The blanket of an animal encloses those of its organs, cells and nuclei.¹ He asked whether anything marks living systems out. His answer was speculative: living systems might have the blanket of minimum entropy, because its physical configuration changes slowly relative to the states it separates.² Ramstead, Badcock and Friston (2018) argued that interactions between cells are mediated by their blankets, so a cell ensemble can be described by transactions among blanket states.⁴ Kirchhoff and colleagues (2018) described the same assembly as nested and hierarchical, with larger blankets made of smaller ones.⁵

Fields, Glazebrook and Levin (2021) argued that quantum theory requires every reference frame to be implemented by a physical system, yet forbids describing any physical system to arbitrary precision.⁶ They drew the consequence that moving a frame between observers means moving its physical implementation, and a finite bit string is insufficient.⁷ They therefore treated all frames available to an organism as implemented by its internal dynamics.⁷ Fields and Levin (2022) extended this to perception-action modules. They offered the CheY-P to CheY concentration ratio, which sets flagellar direction in chemotactic bacteria, as an ancient example.⁸

Fields and Levin (2022) applied the two ideas together. From outside a system's blanket, they argued, no experiment can determine which frames the system uses.⁹ The best available result is an empirical model, built in the language of the experimenter's own frames.⁹ On this view even biochemical pathways are theoretical models resting on evidence that may be limited by the procedures used.⁹ The authors held that this is the limit faced by any other science, and that modelling the system within its environment, at several scales, helps.⁹