CONSTITUTIONAL IMPLICATIONS OF THE USE OF MACHINE LEARNING IN INDONESIA’S SOCIAL ASSISTANCE SELECTION AND DISTRIBUTION SYSTEM

Abstract

This study examines the constitutional implications of using machine learning in Indonesia’s social assistance selection and distribution system. The main objective is to analyze how algorithmic decision-making may affect citizens’ constitutional rights to social security, welfare, equality before the law, legal certainty, and protection from discrimination. This research applies a qualitative legal method with normative-juridical and socio-legal approaches. The analysis is based on constitutional provisions, statutory regulations, social welfare data governance, and policy documents related to Indonesia’s social assistance system, particularly DTKS and SIKS-NG. The findings show that machine learning may improve targeting accuracy and administrative efficiency in social assistance distribution. At the same time, it may reproduce or intensify existing problems in welfare data, especially when the system relies on incomplete, outdated, biased, or unevenly collected information. Algorithmic discrimination may occur indirectly through proxy variables such as residence, housing condition, employment status, digital access, and household composition. This study argues that machine learning should be positioned only as a decision-support tool, not as an autonomous decision-maker. Its constitutional legitimacy depends on data quality, explainability, meaningful human oversight, contestability, independent audit, and clear institutional accountability. The contribution of this study lies in framing machine learning-based social assistance as a constitutional issue, not merely as a technical matter of prediction accuracy or administrative efficiency.