| name | iit-fep-maxcaliber-bridge |
| description | Maximum-Caliber Deviation framework bridging Integrated Information Theory (IIT) with the Free Energy Principle (FEP). Defines information as deviation from constrained maximum-caliber path ensembles, re-derives IIT cause/effect repertoires from variational principles, connects to active inference. Activation: iit fep bridge, maximum caliber, integrated information theory free energy, consciousness framework, constrained entropy maximization. |
Information as Maximum-Caliber Deviation: Bridging IIT and FEP
A mathematical framework defining information as deviation from constrained maximum-caliber path ensembles, enabling re-derivation of IIT 3.0 cause/effect repertoires from variational principles and establishing a bridge to active inference.
Metadata
- Source: arXiv:2605.12536
- Author: Alexander Kearney (Mathematical Institute, University of Oxford)
- Published: 2026-05-03
- Length: 84 pages, 10 figures, 2 tables (Extended Master's thesis)
- Subjects: Neurons and Cognition (q-bio.NC); Artificial Intelligence (cs.AI); Information Theory (cs.IT)
Core Methodology
Key Innovation
Defines information as the deviation ψ of realized dynamics from a constrained maximum-caliber (MaxCal) path ensemble over a finite time horizon. Under this definition, IIT 3.0's cause/effect repertoires emerge directly from MaxCal variational principles — allowing IIT's phenomenological calculus to be re-derived from Constrained Entropy Maximization over Paths (CMEP).
Technical Framework
1. Information as Maximum-Caliber Deviation
- Realized dynamics deviate from a constrained MaxCal path ensemble
- Deviation ψ quantifies information content
- Applies over finite time horizons
- Bridges trajectory-level statistics with causal structure
2. Re-deriving IIT from Variational Principles
- IIT 3.0 cause/effect repertoires derived from CMEP (Constrained Maximum-Entropy over Paths)
- Eliminates need for IIT's axiomatic starting point
- Provides physically grounded basis for integrated information (Φ)
- Connects consciousness calculus to thermodynamic entropy maximization
3. Bridge to Active Inference
- Active inference is mathematically dual to CMEP under Langevin dynamics
- Provides principled route for extending IIT to new dynamical regimes
- Unifies predictive processing with integrated information theory
4. Connection to Predictive Coding
- Under Central Limit Theorem for Markov chains and Large Deviations Theory (LDT) applied to Ising models:
- Information ψ is equivalent to prediction error under predictive coding models
- Explains "hill-shaped trajectory" of Φ observed in neuronal cultures adapting to sensory inputs
5. Thermodynamic Framework of Cognition
- Grounds consciousness in violations of the Fluctuation-Dissipation Theorem (FDT)
- Connects FEP, IIT, and thermodynamic frameworks of cognition
Mathematical Components
MaxCal Path Ensemble
- Path entropy maximization subject to constraints
- Realized trajectory deviation from ensemble quantified as ψ
- Finite time horizon formulation
CMEP (Constrained Maximum-Entropy over Paths)
- Variational principle generating IIT cause/effect repertoires
- Provides physical grounding for Φ (integrated information)
- Dual to active inference under Langevin dynamics
LDT Application to Ising Models
- Large Deviations Theory applied to network models
- Connects information deviation to prediction error
- Links to observable neuronal culture dynamics
Applications
- Consciousness research: Unified mathematical framework for IIT and FEP
- Neural adaptation: Explaining Φ trajectories in adapting neuronal cultures
- Active inference systems: Extending IIT to dynamical regimes
- AI safety: Understanding information processing in artificial systems
- Theoretical neuroscience: Bridging thermodynamic and information-theoretic approaches
Related Skills
- iit-critical-review
- spiking-free-energy-control
- neuro-grounded-foundation-models
- neural-emulator-theory
- thermodynamic-brain-connectivity
- nonequilibrium-brain-dynamics
- nonequilibrium-brain-dynamics-physics
- abstraction-fallacy-ai-consciousness
- ctm-ai-consciousness-blueprint
Pitfalls
- Theoretical synthesis: This is a mathematical bridge proposal, not yet experimentally validated
- Computational complexity: MaxCal path ensembles over finite horizons may be intractable for large systems
- IIT assumptions: Inherits all debates around IIT's foundational axioms
- Oxford Master's thesis: Extended academic work — 84 pages, may contain speculative extensions beyond core claims
- FDT violations: The consciousness-FDT connection is proposed but not proven
Implementation Notes
- Framework is primarily theoretical/mathematical
- Key equations involve path entropy maximization and variational calculus
- For practical applications, start with simple Ising model implementations
- LDT + CLT for Markov chains provide concrete testable predictions
