AI Tool Prioritizes Peer Referrals to Expand HIV Testing Access

In South Africa, men and adolescents are among those less likely to seek testing or attend clinics, leaving infections undiagnosed and delaying treatment. To help reach these groups, researchers have developed an artificial intelligence approach designed to strengthen peer referrals for HIV services.

The work is part of HIV-LIFT, or Leveraging Innovation and Friendships to Improve HIV Testing. The multiyear initiative includes Wits University in South Africa, Harvard’s Teamcore research group at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS), the World Health Organization, and other partners. A field trial is now underway in KwaZulu-Natal province, where the program is targeting populations that have traditionally been harder to reach.

The artificial intelligence component has been added to Coach Mpilo, a free and confidential health coach and chatbot available through WhatsApp. Coach Mpilo is designed for South Africans and provides information and support related to HIV testing, treatment, and other health issues. The platform has more than 150,000 enrolled users, but only about 15% are men.

The intervention uses a financial incentive that encourages existing users to refer friends or family who may need HIV services. Because resources are limited, the Harvard team developed a machine-learning and optimization approach to identify which users should be nudged first. The model is designed to recognize users who are most likely to refer people in underserved groups, including men and adolescents.

The planned study will compare eligible users selected at random with users selected through an additional artificial intelligence filter. The system will use a fixed priority list developed in advance from data collected during earlier incentive campaigns. In retrospective evaluations involving approximately 16,000 historical nudges collected over two months, the approach was estimated to generate more successful referrals than any other strategy.

The researchers emphasized that the system is intended to supplement clinics, health workers, and peer relationships rather than replace them. The study will assess whether the priority-group approach can be implemented permanently within Coach Mpilo or potentially adapted for other applications. The prioritization framework was also designed to be simple, transparent, and usable under real-world system constraints.

“It's great to see real-world deployments of computer science. It's also interesting to see that when real-world constraints come in, you have to adapt. 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 has led the development of the algorithms and is working closely with partners on the ground running the field trial.

“We hope that our work will ultimately contribute to some improvement in how HIV-affected people may be cared for, or how people can get access to HIV information,” said lead researcher Milind Tambe, the Gordon McKay Professor of Computer Science at SEAS.

Related Links
Wits University
Harvard SEAS