OpenAI has selected 14 research projects for the inaugural cohort of its Economic Research Exchange, a program that pairs external economists with the company's research team to study how AI is reshaping education, work, firm behavior, innovation and inequality. The selected studies span experiments and administrative data projects across the United States, Kenya, Brazil, Japan and Denmark.
Aaron Chatterji, OpenAI's Chief Economist and a Distinguished Professor at Duke University, said the program exists to fill a gap in public debate. "The public debate about AI's economic effects needs more evidence. It's our job to empower economists to do that." He added: "This is the beginning of the work."
Education projects test AI adoption and access
Three of the 14 projects focus on education as a subject of study or as a variable shaping AI outcomes. Ricardo Perez-Truglia, Professor of Economics at UCLA, and Zoe Cullen, Associate Professor at Harvard Business School, will examine the future of work, education and innovation through large-scale experimental evidence on advanced technologies. The cohort announcement does not identify study populations, locations or sample sizes.
Edward Miguel, Distinguished Professor of Economics at UC Berkeley, will investigate whether education shapes AI adoption in Kenya and whether adoption widens existing divisions. Germán Reyes, Assistant Professor of Economics at Middlebury College, will study generative AI and employment in Brazil, building on prior work measuring generative AI's effects on college student learning and AI tutoring in Peruvian secondary schools. Both projects ask questions central to AI for Education research: how the technology changes learning, access and opportunity.
Six projects examine labor markets and employer decisions
Labor-market impacts account for six projects. Valentina González-Rostani of USC will study collective bargaining and worker adjustment under generative AI. Anders Humlum of Chicago Booth will link ChatGPT usage data with Danish administrative records. Paul Novosad of Dartmouth and Sam Asher of Imperial Business School will map AI adoption and labor-market change across 10,000 cities.
On the employer side, Kadeem Noray of Harvard Business School, Alexander Cline of UC Irvine and Savannah Noray of Harvard Kennedy School will examine AI adoption and demand for tacit skills using evidence from ChatGPT Enterprise. Yasuhiro Tamba of Seinan Gakuin University will study whether generative AI is redesigning Japan's new-graduate hiring system. Reyes' Brazil study completes the group.
From ChatGPT access to automated science
Two projects address unequal access and benefit. Daniel Björkegren of Columbia University will investigate how people in the poorest groups use ChatGPT. Miguel's Kenya study also examines access.
Three projects concentrate on knowledge production. Andres Algaba of Vrije Universiteit Brussel will study the rate and direction of automated science. Antonin Bergeaud of HEC Paris will examine whether ideas are becoming easier to find. Balázs Kovács of Yale will test whether AI-assisted research tools broaden the prior work scientists discover or concentrate attention on the same sources. These studies connect with broader AI Research efforts to measure how AI changes discovery.
Two measurement projects round out the cohort. Yechan Park, a Harvard economics PhD candidate, will study how AI-generated productivity gains are distributed. Duke economists Felix Tintelnot and Federico Huneeus will examine AI adoption and measurement across production networks.
Why this matters for educators
The education-focused projects will produce evidence on questions schools and universities are already facing: whether AI widens or narrows gaps, how it changes learning outcomes and how it affects hiring and career trajectories. For educators, the useful signal is not in any single finding, since the studies haven't produced results yet. It's in the fact that these questions are now being tested with controlled experiments and administrative data instead of anecdotes. That evidence base will shape how institutions decide what to teach, how to assess learning and how to prepare students for labor markets that are themselves being reshaped by the same technology.
Your membership also unlocks:
