Evidence Becomes Dangerous: Why High-Legitimacy Institutions Resist Learning

Institutional learning failure is usually attributed to weak capacity, poor information or misaligned incentives. This paper identifies the opposite condition: institutions with high legitimacy, strong expertise and substantial public trust are structurally more prone to suppressing evidence than institutions with less to lose. We define dangerous learning as evidence that, if incorporated, would undermine an institution's foundational narrative, invalidate prior decisions made under its authority, or redistribute power and capital. In low-legitimacy institutions such evidence is tolerated, because authority is already weak. In high-legitimacy institutions it is actively suppressed, because legitimacy is the asset at risk and it is asymmetrically costly to lose. Drawing on Institutional Learning Architecture, we trace the mechanisms through which this occurs — evidence without authority, advisory containment, narrative pre-commitment, feedback penalisation and model rigidity under success — and show why reform efforts aimed at transparency and accountability intensify rather than relieve them. The argument implies an uncomfortable prediction: the institutions least able to update are those that have most earned the trust placed in them.