To critically examine the sources and implications of bias in artificial intelligence (AI) applications for audiology in Africa, focussing on cultural, linguistic, technical and ethical dimensions, and their implications for equitable hearing healthcare delivery.
A systematic narrative review of peer-reviewed literature published (2000–2025) was conducted. Empirical studies, systematic and scoping reviews and conceptual papers on AI in audiology and related healthcare technologies were included. Data were extracted and synthesised thematically to identify patterns of bias relevant to African contexts.
Forty studies met the inclusion criteria. Five interrelated sources of bias were identified: (1) data and representational bias due to underrepresentation of African populations, (2) algorithmic and model development bias linked to design and validation limitations, (3) linguistic bias in speech-processing technologies, particularly Automated Speech Recognition, (4) deployment and contextual bias associated with infrastructure and health system constraints in low- and middle-income settings and (5) ethical, governance and regulatory challenges affecting accountability and data sovereignty. Collectively, these biases contribute to inequitable diagnostic accuracy and access to AI-enabled care.
This review advances the concept of “digital ear harm” to describe iatrogenic risks associated with biased AI in audiology and foregrounds Ubuntu as a relational ethical framework for mitigation. Unlike previous reviews that predominantly conceptualise AI bias as a technical or algorithmic challenge, this review demonstrates how data representation, linguistic diversity, cultural context and governance interact to shape healthcare equity in African audiology. It offers a contextually grounded synthesis and strategic directions for equitable, culturally responsive AI development in African audiology.
