From Scientific Symposium to Information Pollution: An Audit of the OSINT Evidence Chain Regarding Hawking's Visit to the U.S. Virgin Islands and the “Private Island Visit” Narrative (2005–2007)

This dataset provides an auditable evidence chain data product for intelligence and national security text analysis, upgrading traditional “interpretable but non-replicable” text interpretation to an analytical process featuring “one-to-one correspondence between conclusions, evidence, and rules, with reproducibility and accountability.” Through structured extraction and dual-algorithm coding of source materials, the dataset generates tabular outputs containing elements such as claims, evidence, coding outputs, reliability/validity metrics, and gating decisions. Conclusions are constrained by explicit gating thresholds and dispute pool rules to prevent narrative filling or over-inference when evidence is insufficient. Core features of the dataset include: Dual-coding: Generates two parallel coding outputs for the same material to measure analytical stability and assess the impact of analyst/algorithm variance on conclusions. Reliability/Validity Gating: Provides verifiable reliability and validity metrics with threshold settings, documenting rework and downgrade rules for non-compliant cases. Evidence Tiering & Traceability: Each critical judgment is tied to evidence tiers and source types, enabling traceability from conclusions back to evidence and coding processes. Reproducibility & Audit-ready: Data structures and field naming designed for reproducibility and independent auditing, enabling rerunning the analytical loop: “Data → Coding → Scoring → Verification → Sealing Decision.” Applicable scenarios include: OSINT/intelligence text auditing, policy text evidence alignment, narrative conflict resolution, reproducible analytical conclusions, and computable evidence chain research for governance and compliance.