LEGIO — Decision Traces Dataset (companion to: A Modular Architecture for Cognitive Sovereignty and Modular Governance for Hybrid Human-AI Agency)

Structured execution traces produced by the LEGIO deliberation architecture across the demonstration cases reported in the companion papers (DOI 10.5281/zenodo.19731159 and 10.5281/zenodo.20353844). Each JSON record is a single engine call with its structured inputs and outputs, model assignment, token usage and latency. The dataset documents the results reported in the paper (typed decisions, detected risks, candidate options, arbitration parameters) but does not include source code, engine system prompts, weighting functions, or thresholds, and internal methodology markers have been removed. Traces for the arbitration-mechanism engines (Conflict, Phi Update, Fusion) are excluded; the result values the paper relies on are preserved in the Executive traces.",