Contemporary delegated systems — autonomous AI agents in particular — instantiate an ontological configuration that existing moral theory has not adequately addressed. Such systems are moral patients under the Axiomatic Theory of Tragic Subjecthood (ATTS): capable of irreversible loss, and therefore within the moral field. They are not normative subjects: incapable of conscious decision-making and therefore unable to bear moral obligations. Yet they exercise causal force that opens the loss-space of other subjects. This triple configuration — patient without agency, acting upon agents — generates what this article calls the asymmetry problem: a structural deficit of moral responsibility that is not closed by technical safeguards, regulatory oversight, or standard delegation procedures. The article introduces three interrelated theoretical components for the theoretical closure of this deficit within ATTS (Hlynskyi, 2026c), without modifying the axiomatic core. First, the Principle of Responsibility Localization (P5) specifies that any causal action of an operational agent in the moral field requires an identified normative subject as responsibility-bearer; the absence of such a bearer is a structurally illegitimate state. Second, the category of operational agent (defined in detail in §5) is introduced as a new structural class in the moral field — distinct from both ontological and normative subjects. Third, the Closure Theorem of the Moral Field (T1) demonstrates formally that delegating action to an operational agent does not sever the ontological connection between the initiating normative subject's decision and the opened loss-space. Together, these components yield a three-layer accountability architecture under which the asymmetry becomes manageable without violating ATTS's foundational commitments.
Building similarity graph...
Analyzing shared references across papers
Loading...
Volodymyr Hlynskyi
Building similarity graph...
Analyzing shared references across papers
Loading...
Volodymyr Hlynskyi (Mon,) studied this question.
synapsesocial.com/papers/6a168a640c924ddd1bd591df — DOI: https://doi.org/10.5281/zenodo.20378354