Fair and Explainable Pregnancy Risk Prediction for Underserved Muslim Communities

This proposal presents a Track 3 shared machine learning competition for the 6th MusiML Workshop 2026 focused on early pregnancy risk prediction in underserved Muslim communities. Maternal, neonatal and child health outcomes in underserved Muslim communities remain critically poor, particularly in low-resource regions such as Zanzibar, Tanzania. Limited antenatal care access, gender-sensitive healthcare barriers and Ramadan-related nutritional challenges continue to delay timely medical intervention. Rooted in Islamic principles that regard maternal wellbeing as a religious obligation, the proposed shared task challenges participants to develop culturally informed, fair, explainable and computationally efficient AI models for identifying high-risk pregnancies early using maternal healthcare records, antenatal care history, maternal nutrition indicators, socioeconomic variables, healthcare accessibility information and clinical risk factors including proxies for religious practices (e.g., fasting during Ramadan, modesty preferences affecting care-seeking). Models will be evaluated using F1-score and AUC-ROC as primary metrics, with secondary evaluation across fairness, robustness, explainability and cultural appropriateness dimensions. Standardized preprocessing pipelines and open baseline implementations will be publicly released to support reproducible experimentation. This challenge ultimately advances responsible healthcare AI research while promoting equitable predictive modeling trusted within Muslim-majority communities.