Learning from AVA: Early Lessons from a Curated and Trustworthy Generative AI for Policy and Development Research

要旨

General-purpose LLMs pose misinformation risks for development and policy experts, lacking epistemic humility for verifiable outputs. We present AVA (AI + Verified Analysis), a GenAI platform built on a curated library of over 4,000 World Bank Reports with multilingual capabilities. AVA’s multi-agent pipeline enables users to query and receive evidence-based syntheses. It operationalizes epistemic humility through two mechanisms: citation verifiability (tracing claims to sources) and reasoned abstention (declining unsupported queries with justification and redirection). We conducted an in-the-wild evaluation with over 2,200 individuals from heterogeneous organisations and roles in 116 countries, via log analysis, surveys, and 20 interviews. Difference-in-Differences estimates associate sustained engagement with 2.4–3.9 hours saved weekly. Qualitatively, participants used AVA as a specialized “evidence engine”; reasoned abstention clarified scope boundaries, and trust was calibrated through institutional provenance and page-anchored citations. We contribute design guidelines for specialized AI and articulate a vision for `ecosystem-aware' Humble AI.

著者
Nimisha Karnatak
The World Bank, Washington DC, District of Columbia, United States
Mohamad Chatila
World Bank, Rome, Italy
Daniel Pinzon
World Bank, Washington D. C., District of Columbia, United States
Reza Yazdanfar
Nouswise, Inc., Dover, Delaware, United States
Michelle Dugas
World Bank, Washington, District of Columbia, United States
Renos Vakis
The World Bank, WASHINGTON DC, District of Columbia, United States

会議: CHI 2026

ACM CHI Conference on Human Factors in Computing Systems

セッション: Optimizing Interactive Systems

P1 - Room 125
7 件の発表
2026-04-16 20:15:00
2026-04-16 21:45:00