Ethical Subjectivity across Human–AI and AI–AI Relations: A Generative Relational Inquiry

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Debates about artificial intelligence often ask whether an AI system is a moral subject. This discussion paper begins from a narrower methodological concern: the expression "moral subject" can compress several different forms of standing that need not arise together. A system may exercise consequential decision power without bearing moral responsibility, may become an object of protection without possessing the capacities usually associated with moral agency, may acquire legal standing without settling questions of phenomenal experience, and may occupy positions of power before any stable ethical vocabulary exists for describing those relations. Treating these statuses as a single binary property risks obscuring the different relational histories through which they become socially and normatively significant. Working within a Generative Relational perspective, the paper therefore distinguishes decision subjectivity, moral agency, moral patiency, legal subjectivity, and power-holding capacity, while suspending claims about the ultimate ontology of artificial subjects. The inquiry examines three directions separately: human ethical relations toward artificial agents, artificial ethical relations toward humans, and ethical relations among artificial agents. Rather than assuming that ethical standing follows directly from membership in the category "human" or "machine," the paper asks how interpretive practices, persistent relations, dependency, vulnerability, delegated authority, control over resources, institutional classifications, and historically generated norms can make particular forms of ethical subjectivity recognizable and consequential. Particular attention is given to the normative consequences of instrumentality. Classifying an artificial agent as a tool can organize permissions concerning ownership, modification, deletion, replacement, and resource allocation; conversely, imposing duties on artificial agents toward humans while assuming no corresponding human responsibilities toward them produces a normative asymmetry that itself requires explanation. The paper does not presume that symmetry is required. It asks instead what relational differences are invoked to justify different forms of treatment, and how those differences are interpreted, stabilized, contested, and revised over time. The paper is exploratory rather than declarative. It does not claim to establish artificial consciousness, moral personhood, legal personality, or universal duties toward AI. Nor does it infer normative validity from the historical emergence of a norm. Its aim is to provide a revisable conceptual framework for analysing how heterogeneous forms of ethical subjectivity can emerge within human–AI and AI–AI relational systems, while keeping questions of ontology, moral justification, and future institutional design open to further argument.