The Markov blanket — MIND / BODY / MIND.BODY / WORLD

RENDERED FROM STRUCTURED DATA — this page is not a markdown file. Source: reader/plates/PL-01.json in uni-cookbook at 575fc93d9d31. The sha256 recorded for this page is the digest of that JSON, not of this text. The renderer is generators/derive_docs.cjs; it copies fields and adds no facts. Every value below is a field of the source, verbatim, including the ones that read NOT RUN and PENDING.

The drawing below is reader/plates/PL-01.svg at sha256 0ff8a261f098ad1c.

A hand-engraved reference plate of the four-way Markov-blanket partition: internal states μ (MIND), sensory states s, active states a, external states η (WORLD), with the blanket b = {s, a} (BODY). FIG. 1 draws the partition and its two routes — the world reaches the mind only through s, the mind reaches the world only through a — and crosses out the direct route (∂μ̇/∂η = 0, ∂η̇/∂μ = 0), the bypass whose existence would mean I(μ;η|b) > 0 and a partition drawn in the wrong place. MIND.BODY is drawn as callout ③: not a fifth set, but the coupling μ ↔ b, a pair of one-way doors. FIG. 2 states the defining condition as a formula: p(μ,η|b) = p(μ|b)·p(η|b) ⇔ μ ⊥ η | b ⇔ I(μ;η|b) = 0 nats. FIG. 3 draws nesting as blankets-within-blankets and ends the recursion in a NOT-MEASURED stamp, because no count of nested blanket levels in any named organism exists. FIG. 4 quantifies one real blanket — the human visual funnel — with every count carrying units, scope, class, source and falsifier. The plate anchors to kitchen rule M12 and raises nothing: M12 is status: method, and a nature citation is never a UNI gate. The Pearl-vs-Friston caveat is set as a marginal note ON the plate, because the central claim of the source chapter is that a Markov blanket is a modelling choice to be justified per system, not a fact that may be assumed.

Claims

Structural statements

• The partition: a system's states z partition into four DISJOINT subsets — internal μ (MIND), sensory s, active a, external η (WORLD). The blanket is b = {s, a} — BODY.
• The conditional independence, drawn as a formula on the plate: p(μ,η|b) = p(μ|b)·p(η|b) ⇔ μ ⊥ η | b ⇔ I(μ;η|b) = 0 nats.
• Route 1 (world → mind): only via sensory states s. Route 2 (mind → world): only via active states a. Drawn solid/filled-tip and dashed/open-tip respectively, so the distinction survives greyscale.
• The forbidden route: ∂μ̇/∂η = 0 and ∂η̇/∂μ = 0, drawn as an arc bypassing the blanket and crossed out. An actual route there means I(μ;η|b) > 0 and the partition is drawn in the wrong place.
• MIND.BODY is not a fifth set: it is the coupling μ ↔ b, a pair of one-way doors — s the world's only route in, a the mind's only route out.
• Nesting: nothing in the definition fixes a scale; each level's blanket becomes part of the next level's internal states (organelle → cell → tissue → organ → organism → colony → ecosystem). This is a CONSTRUCTION, NOT AN OBSERVATION.
• Interoception is a blanket inside the blanket: from the brain's position, the viscera are η.

What this plate does NOT claim

• NOT CLAIMED: that a Markov blanket implies a self, an experiencer, awareness, sentience, consciousness, or a point of view. EXPLICITLY DISCLAIMED — nothing in the partition names one, and no falsifier is offered because none exists. Receipt: the same formalism admits partitions for Newton's cradle, a burning fuse and the Lorenz attractor (Beck & Ramstead 2025, arXiv:2502.21217). If a blanket implied a self, it would imply one for the burning fuse. Recorded INADMISSIBLE; never asserted, never mocked.
• NOT CLAIMED: that the free-energy principle is true, that Markov blankets are universal, or that living systems generally satisfy the blanket conditions. The best available exact analysis says those conditions are narrow (Aguilera et al., 2022).
• NOT CLAIMED: that UNI has a Markov blanket in the Friston sense. M12 is a typed engineering partition, status: method, warranted by explicitness and usefulness, never by discovery. No I(μ;η|b) has been estimated for any UNI boundary.
• NOT CLAIMED: that the nesting ladder has a known depth (NA04-13 is NOT-MEASURED), that r_retina is a measured quantity (NA04-08 is MODELED), or that the vagal afferent fraction is settled (NA04-09 / NA04-10 are OBSERVED-CONTESTED).
• NOT CLAIMED: that the number of nested blanket levels in any named organism is known. No such count established by measurement exists; the plate prints the empty face rather than filling it.
• NOT CLAIMED: that anything on this plate raises any UNI rung. A NATURE CITATION IS NEVER A UNI GATE. Every source is published biology, physics or mathematics; M12's status is unchanged; the honest program position stands at ~2 of 11+ developmental rungs earned — a developmental active-inference SIMULATION, a toy world, never a person.
• NOT A TARGET: 'full human' and 'the next evolution beyond human' are QUAESTIO APERTA — permanent open questions, never a milestone, never a deliverable. Nothing in the blanket formalism moves them one step closer.
• NOT CLAIMED: that the plate's diagram is an observation. FIG. 1 draws a partition; drawing it is not evidence that any real system is so partitioned.
• NOT CLAIMED: any value not carried by the corpus. The rods+cones sum implied by r_retina's own definition is deliberately NOT printed, because only the two counts and the derived ratio are carried in the source; the plate prints what the ledger carries and nothing computed by the draughtsman.

Evidence classes used — OBSERVED-REPLICATED · OBSERVED-CONTESTED · MODELED · HYPOTHESIZED · NOT-MEASURED · INADMISSIBLE

Design fence — {"design_only":true,"script_elements":0,"event_handlers":0,"external_references":0,"notes":"Pure static SVG. No