Artificial Intelligence Models for Predicting Fertility Transitions: A Systematic Review

Abstract

The demographic transition from high to low fertility has profound societal impacts. Recent years have seen growing interest in applying artificial intelligence (AI) to forecast fertility outcomes at both individual and population levels. This review systematically surveys empirical studies (2010–2025) using AI models to predict fertility-related transitions. I screened peer-reviewed articles and identified 30 studies addressing fertility forecasts or outcomes. Most studies apply machine learning to demographic surveys, vital statistics, or personal health data. I report model types, input features, training data, and evaluation metrics with concrete examples. Tadese et al. (2024) used Random Forest and XGBoost on Nigeria Demographic and Health Survey data (n=37,581) to predict women's fertility preferences, achieving accuracy of 0.92 and AUC of 0.98. Tzitiridou-Chatzopoulou et al. (2024) applied XGBoost to forecast monthly birth counts in Scotland, obtaining MAE of 0.32 (vs 0.44 for ARIMA). Deep models have also shown promise. Xue et al. (2024) developed an attention-augmented LSTM (BRP-Net) using economic and demographic features to predict regional birth rates with RMSE of 0.22. I also summarize fairness and equity analyses. Few studies explicitly audit demographic biases. One preeclampsia risk model reported a correctable bias against Black women. Most models neglect performance evaluations for underrepresented groups. I examine regulatory readiness, clinical adoption challenges, and the need for interpretability and user-centered design. Overall, AI shows promise for fertility forecasting but requires careful validation, fairness auditing, and stakeholder engagement before clinical or policy deployment.