Discourse Structure and Comprehension in Yoruba-Language AI Explanations: A Controlled Experiment

This study tests whether Yoruba-dominant speakers with deep oral tradition exposure comprehend, retain, and trust AI-generated explanations more effectively when those explanations are structured according to Yoruba oral discourse conventions — story-first framing, community-referenced validation, and a proverb close — compared to standard Western AI explanatory structure. The study addresses a gap documented in a systematic review of 884 African NLP papers (2019–2026): no published study has measured user comprehension of AI outputs as a function of discourse structure alignment in any African language user population. Three hundred and thirty-seven participants are randomly assigned to one of three conditions (Condition A: discourse-aligned with proverb; Condition B: story-aligned without proverb; Condition C: standard AI format) and read matched explanations on medical, financial, and agricultural topics. Five outcome measures are assessed: immediate comprehension, 48-hour retention, cognitive effort, trust, and text modification behaviour. The Oral Tradition Exposure Index (OTE-I) — a purpose-built continuous covariate — is administered to all participants and included in all analyses. Four pre-specified interaction effects are registered: OTE-I × Condition (primary theoretical test), Topic Domain × Condition, Age × Condition, and Group × Condition primary omnibus. Three result interpretation patterns are pre-committed before data collection. This study is the empirical foundation of the African Cognitive AI research programme; its findings determine the warrant for the Proverb Activation Library, discourse-native benchmarks, and three-agent deployment architecture described in accompanying programme documents.