Core Principles of Institutional Learning: The Structural Requirements for Genuine Update

Contemporary institutions exhibit persistent failure to adapt despite unprecedented investments in data, evaluation and oversight. This paper argues that such failures arise not from deficits of information, expertise or leadership, but from the absence of explicit learning architecture. We introduce Institutional Learning Architecture (ILA) as a structural framework governing the conditions under which institutions can revise internal models, rules and authority without destabilising legitimacy, power or mission continuity. The core failure diagnosed is not resistance to evidence, but the excessive rate at which learning produces destabilising institutional consequences. ILA formalises learning as a temporal cycle defined by observation, update and protection structures, and identifies anti-learning regimes in which evidence accumulates without institutional update. We identify four dominant learning fragility cycles — legitimacy, authority, capital and narrative — and introduce learning-specific decoupling and alignment operators that enable updating without immediate institutional penalty. We show that learning must be decoupled from fragility before it can be aligned to reality, and give conditions under which learning becomes regenerative, producing cumulative adaptive capacity over time. The framework explains recurring failures across public governance, science and AI oversight, and reframes institutional learning as a governance primitive rather than a cultural or behavioural attribute.