IRAM-Omega-Q: A Computational Architecture for Uncertainty Regulation in Artificial Agents

Abstract

Artificial agents can achieve strong task performance while remaining opaque with respect to internal regulation, uncertainty management, and stability under stochastic perturbation. We present IRAM-Omega-Q, a computational architecture that models internal regulation as closed-loop control over a quantum-like state representation. The framework uses density matrices instrumentally as abstract state descriptors, enabling direct computation of entropy, purity, and coherence-related metrics without invoking physical quantum processes. A central adaptive gain is updated continuously to maintain a target uncertainty regime under noise. Using systematic parameter sweeps, fixed-seed publication-mode simulations, and susceptibility-based phase-diagram analysis, we identify reproducible critical boundaries in regulation-noise space. We further show that alternative control update orderings, interpreted as perception-first and action-first architectures, induce distinct stability regimes under identical external conditions. These results support uncertainty regulation as a concrete architectural principle for artificial agents and provide a formal setting for studying stability, control, and order effects in cognitively inspired AI systems. The framework is presented as a technical model of adaptive regulation dynamics in artificial agents. It makes no claims regarding phenomenological consciousness, and the quantum-like formalism is used strictly as a mathematical representation for structured uncertainty and state evolution.

Figures (7)

• Figure 1: Time-course dynamics under low noise ($\eta=0.13$; perception-first ordering). Panels show $S_{vN}(t)$, $\Delta C(t)$, $\mu(t)$, and $\Delta\mu(t)$ (mean $\pm 1\sigma$ across independent seeds; shaded bands). After burn-in, $\Delta\mu(t)$ fluctuates around zero and the mean update converges, indicating closed-loop stabilization rather than monotonic drift.
• Figure 2: Time-course dynamics under high noise ($\eta=0.95$; perception-first ordering). Despite stronger fluctuations, $S_{vN}(t)$ and $\Delta C(t)$ remain bounded and the adaptive update $\Delta\mu(t)$ converges toward zero in expectation (mean $\pm 1\sigma$ across seeds). The convergence occurs strictly below the imposed upper bound $\mu_{max}$, indicating stabilization rather than saturation at the clipping constraint. This demonstrates that stability is not restricted to low-noise conditions and does not depend on a single favorable random initialization.
• Figure 3: Publication-mode phase diagram (perception-first ordering). Left: mean coherence gap over the $(\mu,\eta)$ grid (post--burn-in average, then averaged across runs). Right: susceptibility proxy $\chi(\mu,\eta)$ with the detected critical curve $\widehat{\mu}_c(\eta)$ overlaid (defined by the $\mu$-argmax of $\chi$ at fixed $\eta$).
• Figure 4: Publication-mode phase diagram (action-first ordering). Left: mean coherence gap over the $(\mu,\eta)$ grid. Right: susceptibility proxy with detected $\widehat{\mu}_c(\eta)$ overlay. Relative to Fig. \ref{['fig:phase-pf']}, the regulated regime is shifted toward higher $\mu$, indicating increased regulatory demand when control precedes perceptual consolidation.
• Figure 5: Seed robustness of the critical curve $\widehat{\mu}_c(\eta)$ (perception-first ordering). The complete sweep is repeated across independent random seeds. Thin curves show $\widehat{\mu}_c(\eta)$ per seed; the thick curve shows the across-seed mean, and the shaded region denotes $\pm 1\sigma$ across seeds.