A new study published in PLOS Digital Health finds that while artificial intelligence (AI) tools show significant promise for strengthening research capacity in low- and middle-income countries (LMICs), the technology risks reinforcing the very inequities it could help dismantle if deployed without proper governance frameworks and meaningful local leadership.
The study, conducted by researchers from the Yale School of Public Health, Georgetown University, Queen Margaret University, Spark Street Advisors, and the World Health Organization, reviewed the existing academic literature on the use of AI for strengthening research capacity. The authors found that AI can help researchers in low-resource settings overcome barriers in data analysis, literature management, and scientific writing. The authors also warned that the development of these tools remains concentrated in high-income countries.
“Without immediate action, AI threatens to become the next mechanism through which global health inequities are entrenched rather than eliminated,” said Nina Schwalbe, who led the study. She is affiliated with the Center for Global Health Policy and Politics at Georgetown University and Spark Street Advisors. “We need clear guidelines and guardrails now, not after the damage is done. The window to shape how these technologies are governed is narrowing, and the communities with the most at stake must have a seat at the table.”
The study identified five key areas where AI can support research capacity: (1) data analysis and research productivity, (2) literature reviews and knowledge management, (3) training and capacity strengthening, (4) expanding access to methodological expertise, and (5) writing support. However, the companion review on decolonization revealed a persistent pattern of power imbalances rooted in historic legacies that shape who sets research agendas, who receives funding, and whose knowledge is considered valid.
“One of the most striking findings is how thin the evidence base remains,” said Dr. Brian Wahl, a researcher at the Yale School of Public Health. “Out of all the papers identified in our systematic review, only eight met the inclusion criteria, and all exhibited significant methodological limitations. We urgently need rigorous, empirical research on how AI tools actually perform in LMIC research environments, not just enthusiasm about their potential.”
The authors call for increased funding for AI governance in global health research, including the development of guidance from the World Health Organization on the equitable use of AI tools for research capacity strengthening. They emphasize that effective AI integration requires sustained investment in digital infrastructure, data governance frameworks, and training programs led by interdisciplinary teams that center local expertise.
The study also highlights specific risks, including AI-generated “hallucinations,” algorithmic bias rooted in training data from high-income countries, threats to data privacy, and the potential for dependency on proprietary AI systems controlled by a small number of technology companies—raising concerns about new forms of technological colonialism.
The full study, “Artificial Intelligence for Research Capacity Strengthening: Two Reviews and a Pathway to Shift Power in Global Health,” is available in PLOS Digital Health at https://doi.org/10.1371/journal.pdig.0001302.
