This study aims to analyze the use of big data and machine learning in public policy formulation by positioning citizen participation as a foundation of democratic legitimacy. The study responds to the growing assumption that data-driven policy is more objective, efficient, and rational, while it may also narrow public participation when governmental decisions rely excessively on digital data and algorithmic recommendations. This research employs a qualitative method with a normative-conceptual approach and library research. The data sources consist of legal materials, policy documents, and academic literature related to big data, machine learning, public policy, digital government, algorithmic governance, and citizen participation. The analysis is conducted through qualitative content analysis and normative interpretation to assess the relationship between analytical technology and participatory principles within the public policy cycle. The findings show that big data and machine learning can strengthen problem identification, agenda setting, policy formulation, implementation, and policy evaluation. However, these technologies also create risks of technocratic policymaking, data bias, underrepresentation of vulnerable groups, weak accountability, and the reduction of citizen participation into mere digital data. This study argues that data-driven policy must preserve public consultation, data correction, citizen objection, decision explanation, and public deliberation. The contribution of this study lies in framing citizen participation as a normative limit on the use of big data and machine learning in public policy formulation.