AI for social good
What should count as evidence for AI in policy?
A framework for moving beyond technical performance and asking whether an AI-enabled program improves outcomes that matter.
Notebook
Working notes on AI, evidence, and policy.
Clear thinking about what works, why it works, and how rigorous evidence can guide technology in the public interest.
Writing ideas & essays in progress
Forthcoming
Evidence to policy
From evaluation result to policy decision
How direction, magnitude, implementation, and context turn a research finding into useful guidance for decision-makers.
Implementation
Implementation is part of the treatment
Why training, incentives, infrastructure, and real-world adoption belong at the center of how we evaluate technology.
Research design
External validity in fast-moving technology
What it means to generalize from an evaluation when the product, model, and user behavior may all change before the paper is finished.
Economics of AI
The economics of AI adoption
A field guide to the complements—skills, incentives, management, trust, and infrastructure—that shape whether a technical tool creates value.
Research practice
Reproducibility for policy teams
Simple habits that make quantitative work easier to audit, update, explain, and use under real decision timelines.
Learning
When a null result is still useful
How a careful null result can rule out a theory, redirect resources, and reveal something important about implementation or context.