ABSTRACT
Artificial intelligence (AI), particularly Large Language Models (LLMs), holds great promise for addressing healthcare challenges in Africa. However, the lack of triangulation, which incorporates data, investigator, and theory perspectives, compromises the accuracy, comprehensiveness, validity, and reliability of research findings. Without triangulation, studies on LLMs in healthcare risk bias and may overlook their diverse roles. In addition, reliance on AI without methodological rigour has a negative impact on critical thinking and academic writing in addressing healthcare challenges in Africa. The absence of skilled practitioners, academic researchers, or sufficient healthcare infrastructure further impedes LLM integration, thereby slowing the diffusion of AI innovations in healthcare. This chapter investigates how methodological triangulation, through frugal innovation as a lens, may increase the robustness of LLMs so that a well-rounded understanding of AI contributions to healthcare access in Africa can be ensured. It calls for a multidisciplinary approach integrating empirical data, investigator perspectives, and theory. Without this, AI's impact will be limited. Despite the potential of advanced LLMs, their implementation in Africa is challenging due to substantial resource requirements. In contrast, Small Language Models (SLMs), which are lightweight, economical, and operable offline or in regions with inadequate Internet connectivity, present a viable, cost-effective complement by facilitating the localisation and refinement of African health data.
