A theory-driven candidate annotation architecture for collective regulation under stress in human-centered...

Computational psychiatry has advanced formal accounts of individual prediction, affect regulation, and maladaptive rigidity, but it still has fewer clinically interpretable tools for representing social and institutional processes under collective stress. This Hypothesis and Theory article proposes a theory-driven candidate annotation architecture for this purpose. It defines five collective-process variables-defensive closure (DEF-C), collective anxiety load (ANX-C), integrative progression (PRO-C), punitive superegoic regulation (SUP-C), and moral frame rigidity (FRAME-C)-as theory-guided interpretive indicators for trained human annotation and later formal modeling. Drawing on Active Inference, PAD-S/CSA, Conceptual Metaphor Theory, Moral Politics, social defense theory, and threat-rigidity research, the framework specifies how historically or institutionally situated material can be parsed through explicit coding units, evidence spans, ordinal ratings, confidence markers, adjudication rules, variable-overlap handling, and prospective disconfirmation criteria. Early modern Iberia serves as a worked coding demonstration, with an internal contrast between layered plurality under constraint in al-Andalus and later confessional/inquisitorial consolidation; the Dutch Republic is added as a brief contrastive horizon. Because the framework is applied to historically mediated material, it treats anachronism, narrative uncertainty, interpretive plurality, and temporal delimitation as methodological constraints rather than residual problems. The article does not claim validated measurement status and does not report inter-coder reliability, fitted state-space estimation, or predictive performance. Its contribution is pre-validational and methodological: it provides an intermediate representation layer designed to make clinically interpretable collective-process hypotheses explicit, open to critical inspection, and suitable for future reliability testing, comparative coding, baseline comparison, and human-in-the-loop computational modeling.