Decolonizing LLMs

An Ethnographic Framework for AI in African Contexts

Datos Bibliográficos

ID5948801
AutoresLindsey Dewitt Prat (Bold Insight), Olivia Nercy Ndlovu Lucas (Mantaray Africa), Olivia Lucas (0000-0001-8169-7767), Christopher A Golias, Christopher Golias (Google (United States)), Mia Lewis, M A Lewis
Año2024
Volumen2024
Número1
Páginas46-85
Fecha de publicación2024-11-01
Peer ReviewedSí
Open AccessSí
TipoARTICLE
RevistaEthnographic Praxis in Industry Conference Proceedings (CONFERENCE)
Identificadores de la revistaISSN: 1559-890X • E-ISSN: 1559-8918
EditorialWiley (PUBLISHER • GB)
DOI10.1111/epic.12196
OpenAlexW4407381260
IdiomaEN
Referencias citadas53

This paper examines LLM deployment through African lenses that engage content's complex social, technological, and linguistic landscape. We propose an adaptable framework for LLM research based on an extensive synthesis of research literature and our primary research in Ethiopia, Ghana, Kenya, Nigeria, and South Africa. Our work unearths instances where LLMs could perpetuate digital colonialism or exacerbate existing sociopolitical tensions, as well as how they are contested and adapted. We emphasize the imperative for an embodied ethnographic approach that engages the inherently fluid, flexible, and multicultural nature of language, and connects LLM technologies with the complex social contexts in which it is built and deployed

Ethnography · Sociology · Anthropology · Multimodal Machine Learning Applications · Natural Language Processing Techniques · Topic Modeling · Gender Studies