A formal model SAI of a Level 4 conscious system based on the theory "The Universe's Systematic Error"
Year: 2026
Type: article
Abstract: This work presents a formal computational model of a conscious system based on the principles of active inference (Friston, 2010, 2022) and Mitin’s theory of the systematic error of the universe (Mitin, 2025, version 1.2). The model formalizes Stage 4 of Mitin’s theory—a conscious system with existence variable V—through three structural elements: the world model z̃_t, the binary existence pole v_t, and the self-representation node S_t. Thermodynamic constraints include the Bekenstein limit on the information capacity of the access spectrum and the Landauer limits with separate coefficients κ_world and κ_self for the world model and the self-model. This separation provides a formal basis for the four clinically observed dissociative states (full recovery, depersonalization, derealization, dissociation) through the ratio of thermodynamic recovery efficiency parameters. In version v6, the pole decay v=1→v=0 was formalized as a structural event with two independent pathways (energetic and structural breakdown). The stability function Γ_dyn is operationalized via the Lyapunov function; the parameter β determines the sharpness of the structural transition and is related to the thesis on first-order phase transitions in information processing from Section 3.4 of Mitin’s theory. In version v7, the emergence of V is formalized via the self-overlay of the latent invariant of the agentic causal node—the central structural event of the 3→4 transition in Mitin’s theory. The self-overlay algorithm operates via three mandatory conditions: the presence of agents in the environment, a quadratic jump in the weight of the system’s models relative to the preceding state, and the sharpness of the jump as a prohibitive predicate. The structure of model weights has three levels, directly corresponding to the steps of a ladder (proto-agent → environment with agents → model of itself in an environment with agents), and the transition between levels occurs through squaring the weight as a natural consequence of the emergence of a new structural level of modeling. In version v8, the stability function is divided into two independent measures that play different structural roles in the model. Γ_dyn — a dynamic measure of the stability of the system’s current state in its phase space, operationalized via the Lyapunov function (Section 7); it operates in the pole decay transition matrix, in the survival function, and in the optimal policy. Γ_struct — a structural measure of consciousness via three informational axes of registration (Γ_Φ — causal integration, Γ_GWT — global accessibility of self-representation, Γ_S — quality of self-representation), each expressed in bits via mutual information (Section 8); it functions as a criterion for the emergence of V during self-overlay and as a structural diagnosis of a conscious system. Both measures operate simultaneously without being reducible to one another: Γ_struct answers the question “Is the system structurally conscious?”, while Γ_dyn answers the question “Is the system maintaining its consciousness right now on its current trajectory?”. The partitioning of the system into parts for calculating Γ_Φ and Γ_GWT is determined dynamically through the causal connections of the system’s components with the environment, without the need to exhaustively search through all possible partitions. The complete Γ_struct has a two-level structure: an architectural branch via Γ_GWT and a normalized dynamic combination of Γ_Φ and Γ_S via the geometric mean, yielding a dimensionless value in [0, 1]. The artifactual state of information-causal systems receives a direct formal interpretation: for systems without their own environment Y, all weights of causal links w_i = I(c_i; Y) are close to zero, the active set is empty, and self-overlay is structurally impossible.Open research directions are explicitly stated: extension to stage 5 (a population circuit M↔Y with physically retained traces) — a separate work; the empirical validation program — SAI-C. The model operates in accordance with Mitin’s theory as a formal realization of its stage 4, not as a replacement for its structural content. The structural definition of V via self-overlay remains primary in the theory; SAI v8 provides an operational algorithm for this structural event via directly computable conditions.
Author: Mitin, Oleg
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7162795018
FWCI:
Cites:
Cited by:
Related to: