In Africa, the spread of vaccine hesitancy is associated with rapidly evolving digital narratives, but there is limited evidence on how artificial intelligence (AI) and natural language processing (NLP) have been used to study these narratives and the associated dynamics of confidence and trust. This scoping review mapped peer-reviewed primary studies applying AI, machine learning, and NLP to vaccine-related digital text in African settings. The review documented the data sources, analytical methods, validation, and stated policy or practice relevance of these studies. Using a Population-Concept-Context (PCC)-aligned search of PubMed/MEDLINE and Scopus, and following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) reporting guidelines, 456 records were identified. After deduplication and dual independent screening, 13 studies were included. The included studies spanned North, West, East, and South Africa, with most analyses based on Twitter/X and supplemented by data from Facebook, YouTube, online news comments, and multi-platform social listening tools. Methodologically, the studies could be categorized as follows: (i) sentiment and topic modeling to track hesitancy, confidence, and trust-related attitude dynamics and risk signals during rollout; (ii) supervised classification for misinformation or fake news detection, including work on African languages and dialects; and (iii) operational social listening workflows that translate signals into response recommendations. Automation enabled near real-time monitoring and geographic or temporal segmentation across studies, but validation was commonly internal, and ethical transparency (privacy, consent, and data governance) was inconsistently reported. The potential of African AI-enabled infoveillance for targeted vaccine communication shows potential. However, there is a need for stronger context-sensitive model evaluation, multilingual coverage, and clearer integration pathways into decision-making.
Abdulrauf et al. (Thu,) studied this question.