Structural Idealism

Consciousness as the intrinsic nature of physical structure. A formal framework connecting conscious agents, Markov blankets, and information geometry.

Paper B Submitted — Journal of Consciousness Studies

The Hard Problem & Two Frameworks

The Challenge

Two influential frameworks address consciousness from opposite directions:

  • Hoffman's Conscious Agent theory: starts from consciousness, derives physical structure as an interface
  • Friston's Active Inference: starts from physics (Markov blankets), derives agent-like behavior

Both describe similar structures — perception-action loops, internal/external boundaries — but no formal bridge connects them.

Structural Idealism

The functor Θ: ConMB is a mathematically precise, structure-preserving map between the two frameworks. It shows that:

  • The P·D·A kernel factorization implies conditional independence (the Markov blanket property)
  • Composition of agents (⊗) corresponds to composition of blankets
  • Tononi's integrated information Φ connects naturally to the dissociability structure

The position: consciousness is the intrinsic nature of physical structure. Not emergent from it, not separate from it — identical to it, viewed from the inside.

Category Theory

Both frameworks are formalized as monoidal categories. Θ preserves this structure (a monoidal functor).

Information Geometry

The consciousness manifold carries a Fisher-Rao metric. Agent configurations are points; transformations are geodesics.

Integrated Information

Tononi's Φ measures irreducibility. Zero Φ corresponds to dissociability in the categorical framework.

The Functor Θ

A structure-preserving map from conscious agents to Markov blanket systems.

$$\Theta: \mathbf{Con} \to \mathbf{MB}$$ $$\Theta(\alpha \otimes \beta) \cong \Theta(\alpha) \otimes_{\mathbf{MB}} \Theta(\beta)$$

Category Con

Objects are conscious agents $\alpha = (X, G, W, P, D, A)$ — perception, decision, action Markov kernels operating on internal states $X$, actions $G$, and world states $W$.

Category MB

Objects are Markov blanket systems $M = (\Omega, \mu, B, \eta, K)$ — internal states $\mu$ conditionally independent of external states $\eta$ given blanket $B$.

The Functor Θ: Con → MB — Structure-Preserving Map

Click objects and morphisms to explore how consciousness maps to Markov blankets. The functor preserves composition and tensor products.

More Visualizations

Category Theory Diagram

Markov Blanket Simulation

Particles self-organize into internal, blanket (sensory + active), and external zones. Internal states are conditionally independent of external states given the blanket.

Information Geometry Manifold

The consciousness manifold $\mathcal{C} = \mathcal{K}(W \times X, X) \times \mathcal{K}(X, G) \times \mathcal{K}(G \times W, W)$ with Fisher-Rao metric. Agent configurations as points, geodesics as natural transformations.

IIT Integration (Φ)

Integrated information $\Phi$ measures irreducibility. The minimum information partition (MIP) is shown as a dashed line. Higher connectivity leads to higher $\Phi$.

Papers

Paper A

Consciousness as Intrinsic Nature

Philosophical paper arguing that consciousness is the intrinsic nature of physical structure, with formal foundations in category theory.

DOI
Paper B

Conscious Agents and Markov Blankets

Technical paper establishing the formal correspondence Θ: ConMB with complete proofs of compositional properties.

DOI

Key Results

Theorem (Conditional Independence)

The P·D·A kernel factorization of a conscious agent implies the Markov blanket property: internal states are conditionally independent of external states given the blanket.

Theorem (Compositional Correspondence)

For non-interacting agents: $\Theta(\alpha \otimes \beta) \cong \Theta(\alpha) \otimes_{\mathbf{MB}} \Theta(\beta)$. The functor preserves the monoidal structure.