Active Inference and the Physics of Agency
Summary
Integrates Karl Friston’s Active Inference (Variational Free-Energy Principle) with the Three Thermodynamic Laws of Agency. Law 1 (Control Work): Kybits correspond to precision-weighted prediction errors that agents minimize through active inference—informational complexity = physical work. Law 2 (Agency Decay): Agents require ongoing energetic/informational exchange to minimize free energy; isolation leads to prediction error accumulation and model degradation. Law 3 (Agency Limits): Perfect free-energy minimization impossible due to inherent uncertainty, noise, and model inaccuracies—residual errors are unavoidable. Synthesis: agency = continuous predictive control constrained by thermodynamic reality. Positions Active Inference as compatible framework for axionic physics of agency.
Tags
Cross-References
- Related: Free Energy Principle (Friston), predictive processing [external]
Notes
- Bridges neuroscience (Friston) and physics (axionic)
- Explicit formalization via FEP framework
- Cites Friston extensively (2010, 2013, 2017, 2020)
- Technical integration of two frameworks