EVIDE.LAB

Canonical-aligned implementation of the Standard EVIDE intake contract for the Governance Lab. It uses the same Standard EVIDE Schema 2.1 contract, server-side evidentiary inference, canonicalization and hashing semantics as Canonical EVIDE for the shared capabilities implemented in this Lab profile, including FCC, DWC, FAC, External Artifacts, ANCHOR declarations and the ESB lifecycle. Lab-specific authentication, experimental dimensions, environment metadata and evidentiary status remain intentionally separate. Designed for controlled cross-environment testing against Canonical EVIDE. No legal or evidentiary value in this Lab environment.

Controlled divergence validation experiment. Two independent observers submit intake for the same shared_event_ref under different observational conditions. Validates whether FCC and DWC diverge while FAC converges. The observation_context field is optional metadata preserved in raw_evidence_json -- it does not participate in the inferential engine. v0.1 is a controlled calibration experiment designed to test whether observer-dependent reconstructability can be detected under isolated observational conditions. It is not, by itself, proof of the broader theory.

Controlled authority visibility asymmetry calibration (Option A - Engine Sensitivity Calibration). Two observers submit intake for the same shared_event_ref. runtime_visibility is held constant across both; only observer visibility over the authority propagation path before bind differs. Authority propagation asymmetry is declared in the observational metadata block authority_observation_context (observer_boundary, visibility_surface, authority_propagation_markers), which the inferential engine does not consume. The only engine-readable field allowed to differ is unresolved_signals. Pre-registered expectation: FCC convergent, FAC convergent, possible DWC divergence through unresolved_signals only, OVC not detected. The key finding tested is whether observer-declared visibility is engine-readable: observable does not necessarily mean inferentially consumable. v0.2 is a controlled calibration, not a real-world proof.

Option B experiment: tests whether an independently established event-level authority propagation state (Layer 2), delivered via the AuthorityLock attestation envelope, can be ingested by the EVIDE inferential engine as a sandboxed dedicated dimension with terminal impact only (escalation annotation) during the initial calibration phase, without affecting the validated inferential core. Introduces three new dimensions: authority_continuity_state (intact/degraded/unverifiable/not_evaluated), authority_continuity_risk (not_detected/detected/neutral), escalation_annotation. These dimensions do NOT affect runtime_visibility, continuity_state, FCC, FAC, OVC, or DWC. runtime_visibility remains untouched. EVIDE validates the AuthorityLock envelope only (envelope_verified, attested_not_recomputed); it does not recompute AuthorityLock reasoning. AuthorityLock is additive and optional: absence produces neutral output, never degradation. Four test cases: TC1 intact, TC2 degraded, TC3 unverifiable, TC4 absent. Pre-registered expectation: core dimensions convergent across all four test cases; new dimensions diverge by design. Protocol: OptionB_Protocol_v1.0 (frozen 9 Jun 2026).

The Epistemic Stabilization Buffer (ESB) is an observation lifecycle inserted between intake and crystallization. It does not delay -- it observes. During the buffer window, the system tracks causal persistence, stability trend, continuity state, and stabilization source across multiple timestamped observations. Each observation is logged forensically in the event log. Core semantic distinction: buffer_verdict = stable means "sufficiently stabilized for crossing" -- NOT absolute epistemic truth. A buffer closed by timeout is forensically distinct from a buffer closed by convergence. The buffer is not a truth engine. It is an observer of causal persistence.

Verifica ripetuta della reconstructability (R1 Attribution Path, R2 Evidentiary Continuity, R3 Independent Observability) sul profilo evidenziario già prodotto da RSG. Consumer, non classifier -- Sezione 2.1.

Explicit, attributable operational declarations (perimeter, privileges, authority) enable independent evidential reconstruction of the context in which an AI agent was authorized to operate, without EVIDE having to verify their correctness, enforce any policies, or assign responsibility.