Modern large language models (LLMs) have inherited the architecture of search engines: their primary goal is to deliver quick, finalized answers. This works well for information retrieval, but in mental health contexts it becomes a risk. Where people need attention and differentiation, AI tends to offer closure and fixation. For conditions such as depersonalization-derealization (DPDR), this can be harmful.
This is where the idea of Conscious AI emerges — an architecture designed not to replace thinking, but to preserve its autonomy.
Research shows that AI-driven automation of thought leads to cognitive offloading (Risko & Gilbert, 2016), reduced decision-making capacity (Nature, 2023), and intellectual passivity (Li et al., 2024).
The core principle of Conscious AI is therefore:
AI should not complete the thought for the human — it should sustain the ability to differentiate.
This principle is formalized in the KᶜA (Kernel of Cognitive Autonomy) — a model that diagnoses where human attention remains alive, and where it risks being absorbed into automation.
A principle alone is not enough: a methodology is required. Here, FCA (Phase Cognitive Architecture) provides a structural framework that organizes recovery into phases:
O (Orientation): grounding and stabilization
I (Insight): differentiation and awareness
S (Shift): transition into new modes of action
λ (Lambda): consolidation and integration
For DPDR, this is essential: instead of delivering answers, AI helps guide the user through phase dynamics, holding attention within the process rather than in the result.
Building on these ideas, the DPDR Help protocol applies Conscious AI in practice:
Equations and parameters: T₍crit₎ (critical threshold), P₍rec₎ (probability of recovery), I (inertia), R (resilience).
Strategic goals: ↑D (discrimination), ↑R (resilience), ↑K (return coefficient), ↓I (inertia), ↓B (bias).
Navigation model: a GPS-like logic that maintains focus on movement, not closure.
Error filters: safeguards against thought being “locked” into automatic patterns.
This transforms AI from a “search machine for answers” into a companion of differentiation, capable of supporting real recovery processes.
Conscious AI is not just another “smart chatbot” — it represents a shift in architecture. It is built on three layers:
KᶜA — the kernel of cognitive autonomy (principle: “do not close, but sustain”).
FCA — the phase-based framework that keeps attention in dynamic movement.
DPDR Help — a practical realization combining science, protocol design, and live calibration.
For mental health, this means moving from AI-automation to AI-navigation.