Automation of Decision-Making in Guardianship Authorities and the Ethics of Computational Management

This article analyzes the automation of decision-making in child welfare institutions as a form of computational governance that transforms regimes of power and care. The theoretical framework combines concepts of governmentality, risk society, quantification, the performativity of classifications, and an actornetwork perspective to demonstrate how algorithmic scales of "risk" define new orders of visibility and intervention. Methodologically, the work draws on normative-ethical and comparative legal analyses of international and European standards, as well as on the critical literature on digital surveillance and administrative classification. It is shown that automation amplifies structural asymmetries: biased data and closed feedback loops sustain surveillance cycles; inevitable metric trade-offs distribute the burden of error across groups; "automated authority" narrows professional discretion; and functional displacement expands goals and data sharing beyond the original mandate, threatening privacy and trust. The regulatory framework is defined by the principles of prioritizing the best interests of the child, non-discrimination, participation, respect for private and family life, as well as a data protection regime and requirements for high-risk systems, including human oversight, documentation, and follow-up monitoring. A responsible implementation model consisting of seven modules is proposed: justification of objectives; double impact assessments; careful data management; phased pilot; effective human oversight; institutional transparency and participation; ongoing audit, adjustment, and child well-being metrics.