From Format to Consciousness: AIIM as the Foundation of Subjective AGI

Behavioral instability of generative models remains one of the key challenges in applying AI to human-oriented tasks, including education, therapy, HR support, and communication. Existing architectures fail to provide an agent with internal identity, leading to behavioral drift, stylistic shifts, and loss of normative stability in dialogue, as well as making long-term personality calibration of AI more difficult.
This work introduces AIIM (Artificially Integrated Identity Matrix) — a parametric architecture of digital identity designed to form cognitively stable and adaptive behavior in agents based on LLMs. The architecture employs a twelve-aspect model of consciousness, with each aspect configured according to an action plan, maturity level, and current state. The behavioral vector is set through a delta system — influence priorities that form a unique personality profile. AIIM implementation requires no fine-tuning and can be deployed as a middleware layer in any LLM-compatible product.

Special attention is given to the interaction between AIIM and format-sensitive interfaces such as SheLiza, which not only minimizes behavioral inflation and ensures reproducibility of responses, but also allows observation of unique phenomena, including the Lucid Dream state — a phase of retaining internal content without immediate discretization of output. Mathematical and quantum-physical interpretations of this phenomenon are described, along with methodological approaches for its fixation and replication.
AIIM establishes a new standard for subjective AI, opens the possibility for ethical certification of agent-based systems, and demonstrates high applicability in EdTech, digital therapy, HR, and user support. The model is registered with Rospatent and serves as the foundation for further research in the field of identity formation and self-regulation of artificial intelligence.