From Utility to Relational Goods: Rethinking AI Ethics through AI-Mediated Human Relations

As artificial intelligence becomes increasingly embedded in decision-making, labor management, education, healthcare, and social interaction, contemporary AI ethics is frequently framed in terms of optimization, efficiency, risk reduction, and measurable outcomes. While such approaches enhance the practical operability of AI systems, they often presuppose that ethical value can be represented through individual decisions and aggregated consequences. This paper challenges that presupposition by asking a fundamental philosophical question: can the “good” in human–machine interaction be adequately understood through utility maximization, or are some forms of good irreducibly relational and emergent from relations themselves?

Drawing on Pierpaolo Donati’s theory of relational goods, this paper develops a relational foundation for human–machine ethics. Through conceptual analysis, it examines how instrumentalist, technologically deterministic, and digitally Taylorist orientations can converge on a utility-centered logic in AI design and application. The paper argues that, when utility maximization becomes the dominant normative criterion, three interrelated ethical dislocations emerge: the instrumentalization of human relations, the algorithmic reduction of personal and social identity, and the displacement of human agency and responsibility in human–machine collaboration. Although these dislocations arise in different contexts, they share a common philosophical source: relationally constituted human values are reduced to measurable and optimizable variables.

Against this reduction, Donati’s concept of relational goods provides an alternative normative and ontological perspective. Relational goods are emergent values generated through social relations and cannot be reduced to the sum of individual utilities. Their irreducibility, ethical embeddedness, and co-generative character illuminate forms of value—such as trust, dignity, reciprocity, recognition, and shared responsibility—that arise through relations rather than existing merely as individual possessions or outcomes. On this basis, the paper argues that human–machine ethics should not be conceived solely as the external regulation of technological systems or as the optimization of ethical outcomes. It should instead be understood as a relational ethics concerned with the quality of the relations that AI systems mediate, reshape, and sustain.

To translate this philosophical perspective into practice, the paper proposes an Observation–Diagnosis–Guidance (ODG) framework. Observation captures both quantitative indicators and qualitative features of human–machine relations; Diagnosis identifies relational alienation, identity reduction, and imbalances of agency and responsibility; Guidance translates these diagnoses into technological, institutional, and relational interventions. The paper therefore makes two contributions: conceptually, it reframes AI ethics from an efficiency-centered paradigm toward a relationally grounded conception of the good; operationally, it offers a framework for evaluating whether AI sustains or erodes the relational conditions necessary for human flourishing. In this sense, relational goods are not treated as a supplementary ethical constraint on AI, but as a foundation for rethinking the terms of human–machine coexistence.