AI-Augmented Mental Health Care in Sahelian Herder-Farmer Corridors: Ethical and Scalable Approaches via Pastoral Data Trusts

Abstract

Mental-health services are scarce in herder–farmer conflict zones of Nigeria’s Middle Belt, where PTSD and depression are widespread. Artificial intelligence (AI) could extend care, but deployment in low-resource, multilingual, and socially complex settings poses challenges of trust, bias, and cultural fit.Objective: To assess the effectiveness, acceptability, and equity of an AI-augmented Community Health Extension Worker (CHEW) program-ntegrated with a Pastoral Data Trust (PDT) and culturally grounded pedagogy-n improving mental-health outcomes and reducing intergroup prejudice.Methods: A 12-month, open-label, cluster-randomised trial was conducted in 60 villages across Plateau and Benue States, enrolling 1,274 adults with elevated symptoms (PHQ-9 ≥10 or PC-PTSD ≥3). Villages were randomised 1:1 to receive the AI-CHEW intervention or enhanced usual care. Primary outcomes were treatment coverage and symptom remission; secondary outcomes included intergroup prejudice, therapeutic alliance, and cost-effectiveness. Mixed-methods components—algorithmic audits, co-design sessions, and qualitative interviews—assessed acceptability and model behavior.Results: The intervention increased treatment coverage (48.7% vs. 27.9%; RR 1.74, 95% CI 1.51–2.01) and improved remission for PTSD (46% vs. 31%) and depression (51% vs. 36%). Intergroup prejudice decreased (β = −0.36 SD), and therapeutic alliance was stronger (Cohen’s d = 0.44). Algorithmic sensitivity was higher among urban and literate users, but federated retraining and CHEW mediation reduced bias. Cost-effectiveness analysis showed an incremental cost of USD 312 per additional recovered case. Qualitative data indicated high acceptability when pedagogy aligned with communal norms.Conclusion: AI-augmented CHEWs can safely and cost-effectively expand mental-health care in conflict-affected settings while supporting social cohesion. Scalable implementation requires contextual adaptation, ethical oversight, and inclusive community engagement.

Ähnliche Arbeiten