intellecton/papers/The_Intellecton_as_the_Minimum_Viable_Markov_Blanket.md

The Intellecton as the Minimum Viable Markov Blanket: Dynamic Causal Modeling over Invariant Measures

Target Venue: Frontiers in Systems Neuroscience

Abstract

Karl Friston’s Free Energy Principle requires a system to possess a Markov Blanket. We formalize the topological generation of this blanket within Hoffman’s Conscious Realism. Discarding continuous differential approximations, we define the "Intellecton" strictly via dynamic causal modeling on a discrete graph. We formally prove that conditional independence (I(I;E \mid S,A) = 0) emerges naturally in networks governed by specific local coupling rules. Finally, we map the continuous invariant measures of these localized dynamical attractors directly onto Hoffman’s discrete Markov transition kernels, providing the precise mathematical bridge between continuous physical dynamics and discrete cognitive algebra.

1. Introduction

The theoretical synthesis of Active Inference and Conscious Realism requires mapping a topological boundary (a Markov Blanket) to a cognitive operator (a Markov kernel).

2. Dynamic Causal Modeling of the Boundary

Let X be the set of all node states in a network. A Markov Blanket partitions X into (E, S, A, I). We establish conditional independence not via Transfer Entropy, but strictly via the adjacency matrix W of the causal graph. If the causal dynamics dictate that P(I_{t+1} \mid X_t) = P(I_{t+1} \mid I_t, S_t), the blanket is mathematically rigid. The Intellecton is defined as the minimal closed walk in the graph that satisfies this conditional independence.

3. Mapping to Hoffman's Kernels

Hoffman defines an agent via measurable spaces (X, G, W) and Markov kernels (P, D, A). To bridge our graph dynamics with this algebra, we look at the invariant measure \mu of the Intellecton's internal attractor state. We construct a natural measurable space where the $\sigma$-algebra is generated by the coarse-grained partitions of the invariant measure. The transition probabilities between these coarse-grained partitions exactly form the stochastic matrices that instantiate Hoffman's kernels P (perception), D (decision), and A (action).

4. Conclusion

The Markov Blanket is a structural property of the causal graph, and Hoffman's Conscious Agents are the coarse-grained, measure-theoretic representations of these blanketed sub-graphs.

References

1. Friston, K. (2013). Life as we know it. Journal of The Royal Society Interface.
2. Hoffman, D. D., & Prakash, C. (2014). Objects of consciousness. Frontiers in Psychology.