Litcius/Paper detail

Mathematical formalism of Ze

Jaba Tkemaladze

2026Longevity Horizon5 citationsDOIOpen Access PDF

Abstract

This article introduces and formalizes Ze, a novel theoretical framework for cognitive architecture and autonomous systems. Ze posits that advanced intelligence requires the maintenance of two distinct, asymmetric generative models of the same environment: a causal (forward) model MA\mathcal{M}_AMA​ and a counterfactual (inverse) model MB\mathcal{M}_BMB​. Each model minimizes its own variational free energy FA\mathcal{F}_AFA​, FB\mathcal{F}_BFB​, and their interaction dynamics define core cognitive processes. A key emergent quantity is the model conflict ΔF=∣FA−FB∣\Delta \mathcal{F} = |\mathcal{F}_A - \mathcal{F}_B|ΔF=∣FA​−FB​∣, which regulates a phase transition between two fundamental regimes: an interference regime (characterized by low posterior divergence I≈0\mathcal{I} \approx 0I≈0 where model outputs are constructively fused), and a localization regime (ΔF>θ\Delta \mathcal{F} > \thetaΔF>θ) where the system commits to a single resolved interpretation s^\hat{s}s^. The framework is extended to include active action selection from model-specific policies, a mechanism for representational growth via "which-path" information, and a "quantum eraser" operator for strategic simplification. We demonstrate that this architecture establishes a strict formal isomorphism with quantum measurement phenomena, notably the double-slit experiment, but is grounded entirely in classical variational inference. The theory reinterprets cognitive "collapse" not as a postulate but as an optimization-driven phase transition and yields the key testable prediction that active, alternating intervention accelerates localization compared to passive observation.

Topics & Concepts

Counterfactual thinkingFormalism (music)Computer scienceGenerative modelCognitive architectureMathematicsAction (physics)Operator (biology)Cognitive modelGenerative grammarQuantumSeparable spaceInterpretation (philosophy)Artificial intelligenceDivergence (linguistics)Bounded rationalityAmbiguityCognitionKey (lock)Theoretical computer scienceCausal modelTheoretical physicsChaos theoryPhase transitionStatistical physicsEmbodied and Extended CognitionCognitive Computing and NetworksChild and Animal Learning Development
Mathematical formalism of Ze | Litcius