An Instrument to Evaluate Governance Proposals: AI Policy Analysis at Scale

This paper introduces a transparent framework for comparing AI governance proposals across multiple policy attributes. It combines subject-matter-expert input, empirically informed rubrics, computational text analysis, and evaluations of commercial LLMs. A domain-trained, rubric-calibrated model serves as a benchmark for assessing general-purpose models. The framework is designed to expose policy tradeoffs and embedded normative assumptions rather than declare which proposal is effective or desirable. Its jurisdiction-agnostic design aims to help policymakers, analysts, and researchers compare complex proposals while making attribute selection, rubric construction, and weighting schemes explicit.