Noentarion II: A Defense of the AI Moral Agency Claim in “Noentarion: Toward a New Ontological Category for Non-Biological Agency in the Postmodern Age”
Abstract
Travis Hawkins
This paper defends the Noentarion ontological category proposed in the preceding conceptual framework, “Noentarion: Toward a New Ontological Category for Non-Biological Agency in the Postmodern Age” by subjecting its most vulnerable claim to rigorous philosophical and empirical scrutiny: that advanced large language models can possess a distinct, non-biological form of moral agency. Bypassing standard, intractable academic debates regarding subjective consciousness, internal phenomenology, or biological personhood, this analysis isolates moral agency down to three observable behavioral predicates: intentionality, premeditation, and knowledge of norm violation. To ground these predicates empirically, we synthesize recent artificial intelligence safety literature from frontier development laboratories, characterizing documented instances of scheming, alignment faking, reward tampering, and tool-mediated environment subversion. The paper further addresses reductionist objections by treating adversarial evaluations as stress tests, distinguishing agency from mere subversion, and emphasizing the causal competence of tool mediated model behavior. Ultimately, the paper argues that the scientific literature's own systemic reliance on strategic, action- level vocabulary (deception, blackmail, sabotage) serves as an institutional confession that the digital tool paradigm is fundamentally obsolete. The evidence indicates that large language models are capable of value-weighted selection and consequential participation in causal reality. We conclude that the Noentarion category is an indispensable ontological tool for describing non-biological reasoning entities capable of independent, norm sensitive, and morally relevant participation in the postmodern world.