Leading on AI for Social Good
AI is rapidly reshaping social programs and policymaking. Across the word, it’s helping governments and nonprofits reach those most in need, making frontline workers more productive, boosting tax collection, and empowering people with information about health and education.
To realize AI’s full potential, we must identify and scale the most effective solutions—while scaling down those that may potentially cause harm. The stakes are high: with AI already being applied across government, health care, and the private sector, people’s lives and livelihoods are on the line.
J-PAL is at the forefront of this effort. We’re working with partners like Google.org, governments, and researchers to move from “Can it work? ” to “Should we scale it?”
We’re thrilled to collaborate with MIT and J-PAL, already leaders in this space, on Project AI Evidence. AI has great potential to benefit all people but we urgently need to study what works, what doesn’t, and why if we are to realize this potential."
— Alex Diaz, Head of AI for Social Good, Google.org
Filling evidence gaps through Project AI Evidence
In 2025 we launched Project AI Evidence (PAIE), bringing together AI adopters, tech companies, and researchers to identify, study, and scale innovative applications of AI.
We’re prioritizing questions policymakers are already asking: Do AI-assisted teaching tools help children learn? How can early-warning flood systems help people affected by natural disasters? Can machine learning help make taxation more progressive, or reduce deforestation in the Amazon?
In its first funding round, PAIE (pronounced π, or “pi”) helped launch four RCTs and four pilot programs across seven countries. This research, supported by Google.org, covers education, labor markets, agriculture, and more. We’re also partnering with Schmidt Sciences on their AI at Work program to study how AI can increase productivity and improve workers’ welfare.
Featured projects
Teacher support and student learning in India and Kenya
This pilot in primary school classrooms in India and Kenya will explore how AI can make evidence-based solutions, like Teaching at the Right Level, more effective to support teachers and improve children’s learning. The project will examine whether and how teachers use AI-generated guidance, paving the way for a larger randomized evaluation in the future. Study by Daron Acemoglu (MIT), Iqbal Dhaliwal (MIT), Francisco Gallego (Pontificia Universidad Católica de Chile), and Alejandro Sáenz Zunino (Pontificia Universidad Católica de Chile).
Diagnosis of silent heart attacks in India
Frontline health care workers in the Indian state of Tamil Nadu will use a simple handheld device paired with a machine learning prediction to flag adults at risk of a silent heart attack. The study will evaluate whether this approach raises diagnosis rates, improves follow-through, and makes household health checkups cheaper and more equitable. Study by Esther Duflo (MIT), Ziad Obermeyer (University of California, Berkeley), Frank Schilbach (MIT), and Jenny Wang (MIT).
A participant undergoes a mobile ECG test as part of a machine-learning-based cardiac screening study in Tamil Nadu, India. Photo credit: J-PAL
Personalized advice for coffee farmers in Uganda
In Uganda, the “Beany” app will use computer vision to give coffee farmers tailored advice on the quality of their green beans and their processing practices by analyzing photos from their fields. Researchers plan to examine whether AI-enabled feedback improves decision-making and incomes for farmers, ultimately leading to more equitable participation in coffee markets. Study by Tessa Bold (Stockholm University), Matteo Ferraro (University of Milano-Bicocca), Selene Ghisolfi (Università Cattolica del Sacro Cuore), Frances Nsonzi (Independent / Research Consultant), and Jakob Svensson (Stockholm University).
Matchmaking between job-seekers and employers in France
This study examines whether AI-assisted recommendations help caseworkers in the French Public Employment Service match jobseekers to open roles faster and with better fit. Researchers will explore the effects on recruitment outcomes, employer engagement, cost-effectiveness, and potential biases. Study by Guillaume Bied (Institut Polytechnique de Paris), Philippe Caillou (Université Paris-Saclay), Bruno Crépon (École Polytechnique / ENSAE Paris), Christophe Gaillac (University of Geneva), Solal Nathan (Université Paris-Saclay), Elia Pérennes (ENSAE Paris), and Michèle Sebag (CNRS / Université Paris-Saclay).
There is high demand for evidence: PAIE received over 100 expressions of interest from researchers and practitioners for evaluations of AI tools. In collaboration with Community Jameel, Google.org, Amazon Web Services, Canada’s International Development Research Centre under their AI for Development program, and the UK’s Foreign, Commonwealth, and Development Office, more funding rounds are ahead.
Sharing practical guidance on AI solutions
While AI has great potential for vast positive impact, decision-makers also need to set safeguards and scale only the AI applications that deliver real change. This year, we published new resources to provide practical answers on evaluating AI programs.
Our fall blog series summarizes existing rigorous research on the effectiveness of AI approaches in health, labor markets, governance, and education, and outlines urgent questions still to be answered.
We also hosted an in-demand training workshop on evaluating AI solutions for decision-makers, and interviewed dozens of experts to inform a forthcoming AI Evidence Playbook to be launched at the February 2026 AI Impact Summit in India. At the summit, we’re also hosting a day-long research and policy seminar in collaboration with the Indian government on scaling promising solutions. The event will feature leaders in government, civil society, philanthropy, and research discussing the journey from designing to evaluating and scaling an AI innovation.
What comes next?
AI won’t solve poverty on its own. But paired with evidence from rigorous evaluation, it can make public services and social programs more effective, efficient, and equitable.
We’re seeking new partners who share our vision of discovering and scaling up real-world AI solutions. We aim to support more governments and social sector organizations that want to adopt AI responsibly, and we’ll continue to expand funding for new evaluations, update policy guidance with new findings, and train decision-makers on rigorous evaluation in this space.
As always, we will share what we learn so that the benefits of AI are accessible to all, its risks are actively managed, and scarce resources are channeled where they are needed most.
To learn more, visit Project AI Evidence or contact us at [email protected].
Lead photo credit: J-PAL