Blueprint: From Markov Blankets to Subjects

Thesis

A Markov blanket is neither a sufficient definition of an agent nor a proof of consciousness. It is a scale-relative statistical boundary. Volume 2 becomes philosophically defensible when its blanket formalism is embedded in a temporally extended account of autonomous boundary maintenance and when its IIT claim is treated as an additional, independently testable constraint. The resulting "intellecton" is a process that repeatedly re-establishes its own causal boundary, not a static cortical object.

Section 1: The Boundary Problem

Reconstruct Volume 2's four-state partition: internal (c), sensory (s), active (a), external (\lambda).
Distinguish statistical, causal, operational, and phenomenal boundaries.
Show why (c \perp \lambda \mid (s,a)) is important but ontologically underdetermining.
Introduce the four-rung ladder: boundary, autonomy, integration, subjecthood.

Section 2: Markov Blankets as Scale-Relative Models

Analyze conditional independence and precision-matrix sparsity.
Interrogate linearization, stationarity, solenoidal flow, and collider assumptions.
Define a multiscale blanket score: [ \mathcal{B}\ell = I(C\ell;\Lambda_\ell\mid S_\ell,A_\ell). ]
Argue that blankets are explanatory achievements relative to coarse-graining (\ell), not intrinsic membranes.

Section 3: From Boundary to Autonomy

Integrate autopoiesis, enactivism, and active inference.
Define autonomy as counterfactual boundary maintenance under perturbation.
Introduce: [ \mathcal{A}{T} = \mathbb{E}!\left[\sum{t=0}^{T} \log\frac{p(b_{t+1}\mid b_t,c_t,\mathrm{do}(\lambda_t))} {p(b_{t+1}\mid b_t,\mathrm{do}(\lambda_t))}\right]. ]
Separate mere insulation from self-maintaining organization.

Section 4: Integration Without Equivocation

Examine the transition from recurrent covariance to (\Phi>0).
Distinguish correlation, dynamical irreducibility, causal integration, and phenomenal unity.
Analyze the continuous-to-discrete TPM mapping and its dependence on grain and timescale.
Propose robustness across partitions and interventions as a stronger criterion.

Section 5: The Temporally Thick Intellecton

Replace the instantaneous blanket with a history-dependent process.
Connect memory, predictive processing, and diachronic identity.
Define process identity through recurrent boundary reconstitution: [ \mathcal{I}{T}=\sum_t D{\mathrm{KL}}!\left[p(c_{t+1}\mid b_{\le t}) ,|,p(c_{t+1}\mid b_t)\right]. ]
Argue that agency requires temporal thickness and normative continuity.

Section 6: Empirical and Formal Research Program

Specify falsifiable tests using perturbational neuroscience, causal discovery, and multiscale modeling.
State failure conditions for the Volume 2 thesis.
Develop a hierarchy of evidence for boundary, autonomy, integration, and subjecthood.
Address biological, artificial, collective, and pathological cases.

Section 7: A Critical Process Ontology

Synthesize the reconstruction.
Compare physicalism, panpsychism, enactivism, and process ontology.
Defend a modest conclusion: blankets identify candidate loci of agency, not subjects by fiat.
State the revised Volume 2 principle: an intellecton is a scale-relative, temporally extended process whose boundary is counterfactually maintained and whose causal integration is robust under intervention.

3.4 KiB Raw Blame History