MeasureDev 2026: Open Science in the Age of AI: Balancing Privacy and Transparency
Over the past decade, social scientists and technologists have made significant advancements in transparency and reproducibility: from data and code sharing to pre-analysis plans and replication initiatives. But the field is now at a turning point: artificial intelligence and increasingly complex computational methods are transforming how we analyze data, as researchers face ever-larger and often confidential datasets. Richer data promise to bring more timely and granular insights, but also raise new questions: How can we ensure research remains trustworthy and reproducible? How do we balance openness with privacy and security? And how do we keep up with the pace of technological change?
The 12th annual Measuring Development (MeasureDev) Conference, “Open Science in the Age of AI: Balancing Privacy and Transparency” convened leading researchers, policymakers, and practitioners to review progress over the past decade and chart a path forward for transparency and replicability in light of rapid changes in technology, data access, and computational methods. Co-hosted by the World Bank and the Center for Effective Global Action, MeasureDev brought together experts who are shaping how credible evidence should be generated for development policy.
Program
View selected presentations below.
Adapting Standards and AI Tools to Scale Research Transparency and Reproducibility
Between Efficiency, Recall, and Transparency: An Evaluation of Guided Literature Screening Using LLMs – MetaScreener | Jonas Weinert, World Bank, London School of Economics
TRACE: Trusting Computational Research Without Repeating it | Lars Vilhuber, Cornell University
The AI Replication Engine: Autonomous Verification of Empirical Research in the Social Sciences | Bruno Barbarioli, Institute for Replication
Code in the Machine: Can Large Language Models Reproduce Econometric Analyses? | Aubrey Jolex, Innovations for Poverty Action
AI Tools for Evidence Synthesis
Dev chat | Devika Lakhote, 3ie
Impact AI | Linxi Wang, World Bank
The Global Development Portfolio Atlas: From Fragmented Evidence to Portfolio Intelligence | Janine Aguilera Mesa, Independent Researcher
Generative AI Evaluation in the Development Sector: A Living Playbook | Zezhen Wu, Agency Fund
Keynote
Understanding publication bias in Economics – and what to do about it | Edward Miguel, CEGA, University of California – Berkeley
Balancing Openness and Privacy
Enabling Humanitarian Applications with Targeted Differential Privacy | Nitin Kohli, University of California – Berkeley
Beyond Anonymization: Formal Privacy-Preserving Data Tools for Replicable Randomized Control Trials | Kaitlyn Webb, Pennsylvania State University
Transparency in the Age of AI
AI Reliability for Official Statistics: Benchmarking Large Language Models with the UNICEF Data Warehouse | Joao Pedro Azevedo, UNICEF
Beyond LLMs: Transparency, Ethics, and Interpretability in Geospatial Artificial Intelligence | Virginia Ziulu, World Bank
Economics of Open Science: A Repository-Based Framework for Measuring the Intelligence Value of AI and Code Flows | Silvia Arini, Statistics Indonesia
Lightning Talks
Privacy-Preserving Open Science: Evaluating Hybrid Differential Privacy and Synthetic Data Methods for Equitable Development Research | Patrick Adeyemi Ilori, Uppsala University
Explainability Across the AI Lifecycle: Engineering Trust, Reproducibility, and Accountability in Development Measurement | Mohammed Ba-Aoum, Blue Cross Blue Shield, National Institutes of Health
Confidential-by-Default, Auditable-by-Design: An Evidence-First AI Agent for Consumer-Protection Contract Review | Oumaima Makhlouk, World Bank
