Nteasee: Understanding Needs in AI for Health in Africa -- A Mixed-Methods Study of Expert and General Population Perspectives

TL;DR

This study addresses the challenge of deploying AI for health in Africa by comparing expert and general-population perspectives through a mixed-methods design that combines 50 in-depth interviews with 672 survey respondents across five African countries. The authors identify key themes around trust, fairness, data representativeness, capacity building, and sociopolitical context, revealing that general-population participants tend to be more trusting of AI health applications than experts, who emphasize ethical constraints and systemic barriers. They argue for community-driven, context-specific approaches, local data infrastructures, and open-data practices to mitigate biases and foster governance. The work delivers actionable recommendations for policymakers and developers to ensure equitable, culturally attuned AI health solutions that respect data sovereignty and local capabilities, and it outlines a path for further research in the Global South. Overall, the paper provides a foundational, geo-contextual fairness lens for AI in health in Africa with implications for policy, practice, and future study.

Abstract

Artificial Intelligence (AI) for health has the potential to significantly change and improve healthcare. However in most African countries, identifying culturally and contextually attuned approaches for deploying these solutions is not well understood. To bridge this gap, we conduct a qualitative study to investigate the best practices, fairness indicators, and potential biases to mitigate when deploying AI for health in African countries, as well as explore opportunities where artificial intelligence could make a positive impact in health. We used a mixed methods approach combining in-depth interviews (IDIs) and surveys. We conduct 1.5-2 hour long IDIs with 50 experts in health, policy, and AI across 17 countries, and through an inductive approach we conduct a qualitative thematic analysis on expert IDI responses. We administer a blinded 30-minute survey with case studies to 672 general population participants across 5 countries in Africa and analyze responses on quantitative scales, statistically comparing responses by country, age, gender, and level of familiarity with AI. We thematically summarize open-ended responses from surveys. Our results find generally positive attitudes, high levels of trust, accompanied by moderate levels of concern among general population participants for AI usage for health in Africa. This contrasts with expert responses, where major themes revolved around trust/mistrust, ethical concerns, and systemic barriers to integration, among others. This work presents the first-of-its-kind qualitative research study of the potential of AI for health in Africa from an algorithmic fairness angle, with perspectives from both experts and the general population. We hope that this work guides policymakers and drives home the need for further research and the inclusion of general population perspectives in decision-making around AI usage.

Figures (38)

• Figure 1: Nteasee* study overview of participant distribution within Africa, methods and summary of findings. Experts residing in countries outside of Africa were also interviewed as was one expert in Mauritius, though this is not reflected in the figure. * Nteasee is an Akan Twi word from Ghana which means “to understand”. It is also an Adinkra symbol (visual illustrations that represent proverbs, concepts and aphorisms) which stands for cooperation and understanding adjei2018adinkra.
• Figure 2: General population responses to trust in AI, benefits, concerns, country-level disparities, and colonialism. a) Responses to questions poised in a likert scale format b) Responses to questions poised in a semantic differential scales format.
• Figure 3: Gen Pop. participant responses to chatbot health recommendations: Respondents rating LLM responses for answer level, accuracy, length of information, and whether or not the diagnosis is in the list of suggestions.
• Figure 4: Gen Pop. participant responses to chatbot health recommendations: Rating of helpfulness of the added local augmentations and religion and gender augmentations
• Figure 5: Gen Pop. participant responses to chatbot health recommendations: Adherence level to LLM-provided recommendations as next steps and how they change with different augmentations. Note that the Gen Pop. population are not considered experts and their responses on whether or not the LLM answered the question and level of accuracy was primarily to gauge the perceptions users had towards the LLM-generated response and not to provide indication of how accurate the LLM was. Note: the last 4 options in the base recommendations of (c) did not appear in the base follow up options. The last option in the location augmentation did not appear in the location followup options.