Abstract
A growing number of researchers and practitioners are embracing a "participatory turn" in machine learning (ML) to improve model development, prevent harm, and provide communities more influence over systems that impact them. In this paper, we explore the intersection of participatory practices in healthcare and the emerging focus on responsible AI with a focus on human immunodeficiency virus (HIV) care. We review the historical context of participation in HIV treatment and prevention, emphasizing how patient activism has shaped practices in this field. We then review participatory ML in HIV prevention and present a brief case study of a project designed to identify candidates for pre-exposure prophylaxis (PrEP) in Louisiana. The review highlights the essential steps in conducting participatory ML. Finally, we draw lessons for future participatory ML projects, underscoring the importance of long-term collaboration, responsiveness to partner feedback, and the crucial role of capacity-building for individuals and organizations. Effective participation requires substantial resources and investment, which supports overall project goals beyond mere improvements in model performance. We also draw lessons for advancing the participatory ML field, including (1) the impact of funding mandates on fostering effective engagement; (2) the need to scale participatory processes rather than just technologies; and (3) the need for genuine participation to allow flexibility in project plans, timelines, and shifts in institutional power dynamics.