Family caregiving practices play a central role in shaping children’s early developmental outcomes, yet the importance of specific caregiving and the contributions of different caregivers remain less understood. This study applies explainable artificial intelligence (XAI) to examine how various caregiving activities with children —reading books, singing songs, playing, taking the child out, telling stories, and naming, counting, or drawing— carried out by different family members —mothers, fathers, or other adult family members— within Vietnamese households predict early childhood development.
Various machine learning algorithms were evaluated on Vietnam Multiple Indicator Cluster Survey 2020–2021 data to identify the best predictive model. The Random Forest model yielded the best predictive accuracy, with a Mean Absolute Error (MAE) of 2.35, a Mean Squared Error (MSE) of 9.21, a Root Mean Squared Error (RMSE) of 3.04, and a Coefficient of Determination (R²) of 0.48. SHapley Additive exPlanations (SHAP) were then employed to compute feature importance for model interpretability.
Findings suggest a greater importance to the type of activity rather than the identity of the caregiver in child development predication. SHAP values indicate that caregiving activities provide positive contributions to the model’s predictions, regardless of which family member performs them. Conversely, the lack of caregiving engagement is negative, highlighting the developmental risk associated with the absence of stimulation. The results also reveal that caregiving activities exert similar marginal effects for boys and for girls. Across caregiving domains, the positive influence of stimulation is robust, while gender interaction effects are small.

Authors:
Mohamadou Bassirou Jean-Baptiste Sanfo University of Hyogo, Japan

About the Presenter(s)
Dr. Jean-Baptiste SANFO is currently an Associate Professor of statistics and econometrics at the School of Economics and Management, University of Hyogo.

See this presentation on the full schedule – Thursday Schedule

Bookmark