This project studies AfriGPT, an SMS-based AI chatbot offered by the telecom Africell in Sierra Leone. AfriGPT delivers unlimited chats with a Large Language Model (LLM) over SMS. Users can chat with AfriGPT even if they do not have internet access, cannot afford data, or do not own a smartphone. With over 6,000 daily interactions in Sierra Leone alone, AfriGPT reaches a population that is largely absent from existing research on AI use. Its users differ fundamentally in income, education, digital skills, and connectivity, making AfriGPT a unique window into how AI is adopted and used by digitally underserved communities.
The AfriGPT project studies how people use such low-resource AI systems, and their social and economic effects. Through an unprecedented data partnership with Africell, the project studies full chat histories and user metadata. In its first phase, the project provides large-scale, systematic evidence on AI use in a low-income country. In particular, it compares AfriGPT usage profiles to those of ChatGPT and Claude and investigates whether AfriGPT is used as a substitute for web search. In its second phase, the project will implement randomized controlled trials that examine how the chatbot impacts information access, learning, and skills acquisition.
The project makes three core contributions. Empirically, it fills a major gap in the global AI literature by documenting real-world AI use among populations that are currently excluded from most datasets and policy debates. Methodologically, it establishes ethical and legal standards for analysing and sharing sensitive chatbot data. Substantively, it provides evidence relevant to digital inclusion and AI policy. Without targeted research and deliberate interventions, AI adoption will mirror existing digital divides. Current AI deployment systematically bypasses the 2.6 billion people without internet access, and risks exacerbating existing global inequalities. The project advances our understanding of whether AI can expand access to information and skills for the billions of people who remain digitally underserved.
Schmidt Sciences is a nonprofit organization founded in 2024 by Eric and Wendy Schmidt that
works to accelerate scientific knowledge and breakthroughs with the most promising, advanced
tools to support a thriving planet. The organization prioritizes research in areas poised for
impact including AI and advanced computing, astrophysics, biosciences, climate, and
space—as well as supporting researchers in a variety of disciplines through its science systems
program.
Africell Press Release and LinkedIn post
DPhil Student
Project role: Co-Primary Investigator
Johanna is a DPhil student in Information, Communication and the Social Sciences. Her research explores how institutions and resources shape AI adoption.
DPhil Economics, Department of Economics, University of Oxford
Project role: Co-Primary Investigator
Before joining the Dept of Economics, Joseph worked on forecasting and information economics at the Global Priorities Institute. He has designed & implemented RCTs in Sierra Leone, India, and Afghanis
