Tagline:A formal architecture for studying how institutional decision systems learn, drift, and realign over time under conditions of delayed feedback. Description:This precursor paper establishes the initial architectural specification that later evolved into the Operating Spine and the broader research programme The Coherence Problem: How Institutions Learn, Drift, and Realign. Part 0 provides the conceptual entry point to the programme, while subsequent papers develop the theoretical mechanisms, measurement approaches, and field protocols introduced here. The paper introduces AI-Augmented Impact Frames, a constrained, closed-loop architecture for studying how institutional decision systems learn over time under conditions of delayed feedback. The framework formalises decision logic, constrains AI to interpretive support, and integrates longitudinal psychometric measurement to enable empirical observation of decision-system dynamics while preserving human decision authority. This record corresponds to the conceptual architecture stage of the programme and does not report empirical results. Subsequent papers in the programme refine the theoretical foundations, mechanisms, and empirical methods introduced here. This version is retained to provide a citable architectural baseline, scholarly traceability, and a timestamped reference point for the conceptual origins of the framework. Later versions and related programme papers are linked in the Related identifiers section of this record. Supporting materials, working documents, and programme structure are available via the Open Science Framework (OSF): https://osf.io/9cvky/
Robin Edgard Ulrik Mertens (Tue,) studied this question.