Towards Responsible and Personalized Social Protection: Leveraging Machine Learning for Policy Innovation in Developing Countries – The Case of Morocco

Abstract

This article explores the potential of Machine Learning (ML) to enhance the effectiveness, fairness, and personalization of social protection systems in developing countries, with a particular focus on Morocco. In light of the demographic, institutional, and technological transformations underway, ML emerges as a promising tool to support eligibility assessment, fraud detection, beneficiary segmentation, and long-term forecasting.
The paper proposes a structured five-step methodological framework for the ethical and context-sensitive integration of ML in welfare systems. Drawing from international experiences (Estonia, India, Rwanda, Chile), it identifies key success factors, including data interoperability, human oversight, and transparency. The article then outlines five strategic use cases tailored to Morocco’s current reform of universal social protection, ranging from predictive targeting to territorial outreach.
To assess the social acceptability of these innovations, an exploratory survey was conducted among Moroccan citizens and stakeholders. Results indicate a cautiously optimistic attitude toward ML adoption in public services, tempered by concerns around algorithmic bias and institutional trust. Based on these findings, the paper offers concrete policy recommendations to foster inclusive, transparent, and human-centered AI governance in the social domain.
By combining theoretical insight, comparative evidence, and citizen perceptions, this study contributes to the global ``AI for Social Good'' agenda and informs the digital transformation of social policy in the Global South.