[Paper Review] The Case for Globalizing Fairness: A Mixed Methods Study on Colonialism, AI, and Health in Africa

Mercy Asiedu, Awa Dieng+8 more|arXiv (Cornell University)|Mar 5, 2024

Viral Infections and Outbreaks Research4 citations

TL;DR

This mixed-methods study investigates how colonialism shapes perceptions of fairness in AI-driven healthcare in Africa, combining a scoping review with surveys of 672 general participants and 28 experts. It finds that while experts link colonial history to AI bias, most general population participants do not see a direct connection, highlighting a critical perception gap and calling for contextually grounded, philosophically inclusive fairness frameworks in global health AI development.

ABSTRACT

With growing application of machine learning (ML) technologies in healthcare, there have been calls for developing techniques to understand and mitigate biases these systems may exhibit. Fair-ness considerations in the development of ML-based solutions for health have particular implications for Africa, which already faces inequitable power imbalances between the Global North and South.This paper seeks to explore fairness for global health, with Africa as a case study. We conduct a scoping review to propose axes of disparities for fairness consideration in the African context and delineate where they may come into play in different ML-enabled medical modalities. We then conduct qualitative research studies with 672 general population study participants and 28 experts inML, health, and policy focused on Africa to obtain corroborative evidence on the proposed axes of disparities. Our analysis focuses on colonialism as the attribute of interest and examines the interplay between artificial intelligence (AI), health, and colonialism. Among the pre-identified attributes, we found that colonial history, country of origin, and national income level were specific axes of disparities that participants believed would cause an AI system to be biased.However, there was also divergence of opinion between experts and general population participants. Whereas experts generally expressed a shared view about the relevance of colonial history for the development and implementation of AI technologies in Africa, the majority of the general population participants surveyed did not think there was a direct link between AI and colonialism. Based on these findings, we provide practical recommendations for developing fairness-aware ML solutions for health in Africa.

Motivation & Objective

• To examine how colonialism influences perceptions of fairness in AI applications within African healthcare contexts.
• To identify and validate axes of disparity—beyond Western-centric models—relevant to fairness in AI for global health in Africa.
• To bridge the gap between expert and public perceptions of colonialism’s legacy in shaping AI equity in African health systems.
• To develop practical, contextually attuned recommendations for fairness-aware AI development in low- and middle-income African countries.
• To advocate for a broader philosophical foundation in algorithmic fairness that includes non-Western conceptions of justice and normativity.

Proposed method

• Conducted a scoping review to identify potential axes of disparity in fairness for AI in African health contexts.
• Administered a survey to 672 general population participants across multiple African countries to assess public perceptions of AI and colonialism.
• Conducted in-depth interviews (IDIs) with 28 experts in machine learning, health, and policy focused on Africa to gather professional perspectives.
• Analyzed qualitative data using thematic analysis to identify recurring themes on colonialism, AI, and fairness.
• Used mixed-methods triangulation to compare expert and public perceptions and validate proposed axes of disparity.
• Explored the implications of findings for bias mitigation techniques and policy development in AI for global health.

Experimental results

Research questions

• RQ1 How do African general population participants perceive the relationship between colonialism and AI in healthcare?
• RQ2 What axes of disparity do experts and the public identify as relevant to fairness in AI for African health systems?
• RQ3 To what extent do experts link colonial history to current inequities in AI deployment in African healthcare?
• RQ4 How do perceptions of colonialism's legacy differ between experts and the general population in the context of AI fairness?
• RQ5 What practical recommendations emerge for developing fairness-aware AI systems in African health contexts that account for historical and structural inequities?

Key findings

• Experts overwhelmingly recognized colonial history as a key axis of disparity influencing AI fairness in African health systems.
• A majority of general population participants (over 50%) did not perceive a direct link between colonialism and AI bias, indicating a perception gap.
• Country of origin and national income level were identified by participants as significant axes of disparity affecting AI fairness in African health contexts.
• Participants emphasized the importance of data ownership, local inclusion, and equitable partnerships in AI development to ensure fairness.
• There is a need for a philosophically broader approach to algorithmic fairness that incorporates social justice and non-Western normative frameworks.
• The study highlights the risk of algorithmic colonialism, where AI systems from high-income countries may perpetuate historical inequities without local context or oversight.