Prototyping an AI Governance Capacity Index App Using Claude Code

• In Event: From Text to Evidence: AI-Assisted Analysis and Measurement

Abstract

Governments increasingly need to assess their own capacity to govern AI responsibly, yet the tools available fall short. Global indices such as Oxford Insights' Government AI Readiness Index compare countries, not the states and localities where most AI governance actually happens. Existing maturity models also rarely treat AI governance and data governance as related but distinct capacities. This session's index addresses both gaps, building on the presenter's prior comparative research on Korean local governments. This session demonstrates how Claude Code can turn a policy-theory-grounded framework into a working prototype app. Grounded in public policy capacity theory and organized as a logic model of inputs, processes, outputs, and outcomes, the index covers nine dimensions: strategy, institutional design, decision-making structure, collaborative governance, data management, AI adoption, performance management, and ethics and trust. Claude functions here not as a writing aid but as a measurement instrument, applying a pre-defined rubric to score qualitative text such as strategic plans and organizational charts. The prototype pairs this qualitative scoring with quantitative indicators pulled from open government APIs — Korea's Public Data Portal and Local Finance 365, and a U.S. state's Socrata-based open data portal. Because AI-based coding is still an emerging methodology, the session also addresses how to hold it to standards long applied to human coding: checking consistency across repeated scoring runs, and identifying what further validation, such as inter-rater agreement with human coders, would be needed before treating the index as a genuine research instrument rather than a one-off demonstration. The session runs in five parts: a brief framework overview; a video walkthrough of Claude Code scoring sample text against the rubric; a video walkthrough of the resulting dashboard comparing a U.S. state and a Korean local government; a facilitated discussion applying the workflow to participants' own cases; and a wrap-up with resources and Q&A. By the end, participants will be able to translate an academic framework into a scorable AI-assisted index, combine open government data with rubric-based qualitative coding, and identify what data access and validation this approach requires. This lab is for researchers and practitioners who want practical, replicable ways to use AI not just to write, but to build and validate research instruments.

Author

• Kun Yoon, Pennsylvania State University
  Facilitator