More than 150,000 people have signed up for Coach Mpilo, a free health coach they can reach through WhatsApp—but only about 15% of its users are men. In KwaZulu-Natal, South Africa, researchers are testing whether artificial intelligence can help close that gap by encouraging more people to seek HIV testing and care.
South Africa has about 8 million people living with HIV, more than any other country. Yet men and adolescents are less likely to get tested or visit clinics, which can leave them unaware of an infection and without care. Stigma and long working hours can make it harder for men to seek help. The HIV-LIFT project is trying a different route: helping people invite friends and family to use health services.
HIV-LIFT stands for Leveraging Innovation and Friendships to Improve HIV Testing. The multiyear study brings together Wits University in South Africa, Harvard University’s Teamcore research group, the World Health Organization and other partners. It is led by Alastair van Heerden of Wits University. Harvard researchers, led by computer science professor Milind Tambe, are testing a machine-learning system within the existing Coach Mpilo app.
Coach Mpilo provides confidential information and support about HIV testing and treatment, as well as other health concerns. Users can receive a financial incentive for referring people they know to services. But funds for incentives are limited, so the trial asks a practical question: which users should be encouraged to make referrals first?
The AI system is designed to identify Coach Mpilo users who may be most likely to refer people in groups that are often missed, including men and adolescents. Researchers hope that focusing the available incentives this way will help the program reach people who might otherwise not seek testing or treatment. The trial is testing the approach in real communities, not just in a computer model.
That real-world setting has already shaped the work. The Harvard team first explored a more complex simulation method, but its results were difficult for ordinary users to understand. The researchers then adapted their approach to better fit the needs and limits of the program. “There’s no point in having fancy things that look good on paper but cannot be used in practice,” said Davin Choo, a postdoctoral fellow who helped develop the algorithms and works with partners running the trial.
The effort is meant to support, not replace, clinics, health workers or trusted relationships between people. Cheryl Case Johnson, a World Health Organization technical officer and innovation lead, is also contributing expertise in HIV network-based interventions—the use of social connections to help people reach services.
The KwaZulu-Natal trial will show whether carefully chosen prompts and incentives can help Coach Mpilo’s users bring more people into HIV care. If the approach works, it could offer health programs a practical way to use limited resources to reach groups that have often been left out, while keeping human support at the center of the effort.