ConsciOS: A Viable Systems Architecture for Human and AI Alignment

ConsciOS proposes a formal systems architecture for studying human and AI alignment as a property of nested control structure rather than only post-hoc behavioral correction. The paper models consciousness and self-regulation as a testable control architecture composed of an Embodied Controller for short-horizon perception-action loops, a Supervisory Controller for policy/frame selection, and a Meta-Controller for long-horizon priors and governance constraints. The v5 manuscript is prepared as a journal-submission revision for a Hypothesis and Theory article. It sharpens the AI-alignment framing, removes nonessential metaphor and branding language, and clarifies the distinction between formal proposal, illustrative instrumentation, and future empirical validation. The core mechanisms remain: a Resonance Engine that combines expected utility, coherence, and cost; an Interoceptive Control Signal (ICS) as a proposed fast feedback channel; and Time-Integrated Coherence (TIC) as a proposed resource for policy complexity and option-availability. The manuscript situates ConsciOS within systems theory, the Viable System Model, active inference, affect science, hierarchical reinforcement learning, and human-in-the-loop AI alignment. It provides formal definitions, algorithmic sketches, falsifiable hypotheses, proposed human-subjects and simulation protocols, governance considerations, and a reproducible toy instrumentation demo. The toy code is included to show how selector variables can be logged and visualized; it is not presented as empirical validation of the architecture. Keywords: consciousness architecture; AI alignment; viable systems model; active inference; hierarchical reinforcement learning; interoception; coherence-based control; human-AI hybrids; cybernetics; systems engineering