External Reader Protocol: Calibration, Downgrade Rules, and Public Use of Structural Intelligence

This paper develops an external-reader protocol for Structural Intelligence (SI). Its central problem is public dependability: a framework may be philosophically strong while remaining weak as a public instrument if non-native readers cannot distinguish observation from inference, strong evidence from weak evidence, grounded diagnosis from speculative extension, or responsible refusal from interpretive failure. SI is especially vulnerable to this problem because its central terms—coherence, contact, answerability, burden, witness, repair, and revision—can be imitated as a style. The paper proposes a practical protocol for public use: a minimal runtime sheet, claim-status ladder, downgrade rules, reader-capture safeguards, calibration cases, a falsification registry, and a public translation layer. It specifies how readers should mark evidence, missing information, falsifiers, cost bearers, revision triggers, witness or trace, claim status, and output state before issuing strong SI readings. It also defines when a reading should be downgraded, blocked, routed to containment-first, or marked as no reading justified. The contribution is not another expansion of SI into a new domain, but a reliability layer for the framework itself. SI becomes stronger when it can preserve refusal power: the disciplined capacity to say “not enough evidence,” “not my jurisdiction,” “provisional only,” “blocked,” or “repair before interpretation.”