The initiative will connect governments, nonprofit organisations, tech companies, and world-class economists to rigorously test which AI-based tools are effective, equitable and safe before scaling them, rather than adopting technologies based on hype alone.
Project AI Evidence has already provided funding for eight research studies exploring how AI can help address entrenched social challenges in areas such as education, employment, disaster response and environmental protection. One example includes evaluating AI-assisted teaching tools in Brazil that have helped reduce learning gaps between public and private school students, with the possibility of wider rollout if proven effective.
Researchers will also investigate whether AI can help reduce gender bias in schools, empower career counsellors to uncover more job opportunities for underserved youth, and support early-warning systems for natural disasters. A key goal is to generate usable evidence on “what works, what doesn’t, and why” so policymakers can invest in AI tools that truly improve lives and avoid those that could inadvertently cause harm.
The initiative is backed by major supporters including Google.org, Amazon Web Services, Canada’s International Development Research Centre, and philanthropic funding from Eric and Wendy Schmidt. By building this global network of evaluation expertise and partnerships, J-PAL aims to make AI a responsible and high-impact force for development, particularly in low- and middle-income countries where poverty reduction remains an urgent priority.
