Using evidence. Improving lives. Developing “EBPM Database” to Improve Policy Making Process in Japan Daisuke Moriwaki, CyberAgent AILab @dmoriwaki Yohei Kobayashi, MURC/CSIS

Declaration of Conflict of interest To the best of my knowledge, I declare that I, and/or my co-authors and any of my/our close family members, have not had employment, received research support or other funding from, or had any other professional relationship with, an entity directly involved in the production, manufacture, distribution or sale of tobacco, tobacco products, weapons or arms, or have represented the interests of any such entities in any way. I have no actual or potential conflict of interest in relation to this presentation.

EBPM (Evidence-Based Policy Making) Database • EBPM Database, or EBPMDB – The first “Evidence portal” – Founded 2022 by voluntary experts – Dev on Github. Running on GitHub Pages • Inspired by – What Works Centers (UK) – Campbell Collaboration etc. • Primary Objective – Enhance accessibility to research findings Intervention Policy area Outcome Effectiveness Strength Education Modern teaching practice Test scores Modern teaching practice in Ele. School Detail of the paper Positive Regression with controls Individual ”review” https://cyberagentailab.github.io/EBPMDB/ https://github.com/CyberAgentAILab/EBPMDB/ Introduction of

How to Compile “Reviews” • 70+ reviews of research papers – Robust evidences from articles published in quality journals • Focus on the important policy areas – Low fertility rates followed by education/environment and health • Review writers include university researchers to volunteers from private sectors. • 4000+ visitors incl. researchers & govt. officials

Summary of the “Reviews” • Strength of Evidence – Based on Maryland Scientific Methods Scale – 70+% research use modern causal inference • Effectiveness of Interventions – A quarter of “None” while majority is positive – Side effect (unintended “bad” effect) & mixed

Challenges & Next Step • Challenges – Less interest shown from local govt. officials – Lower recognition of the importance of evidence • Next Steps – Demand side: outreach • Twitter (X) account to promote new reviews • Workshops for local govt. officials using EBPM database – Supply Side: scaling • Lower production cost of reviews with GenAI Workshops for Local City Officials Twitter (X) account (@EBPMdb)