From Psychology to AGI: A Mathematical Foundation of Titov's Subject-Centred Model for Building Machines with Intrinsic Motivation and Subjectivity

We translate Titov's subject-centred psychological model into a mathematically rigorous, fully differentiable framework for building AI agents with intrinsic motivation and subjectivity. The model formalises the interaction of two subsystems — the intentional (I) and the intending (D) — through a closed causal loop (reentry). We use a Gaussian approximation for integrated information Φ_G, propose the SubjectNet architecture with PyTorch pseudocode, and demonstrate that in the continuum limit the model becomes a lattice gauge theory connected to the CNN/LQCD isomorphism. A minimal falsifiable experiment is proposed.