Participated Agency and the Limits of Informational Naturalism: Toward a Theological Ontology of Artificial Intelligence

The question of artificial agency has become urgent for Christian theology: how we characterise the causal role of AI systems determines whether the human person retains structural primacy over the technological environment. This article engages Floridi’s Multiple Realisability of Agency (MRA) and argues that, while its descriptive achievements are genuine, its epistemological method cannot generate the ontological foundation its normative programme requires. MRA’s triple subordination of AI to human intelligence – by origin, finality, and evaluation – implicitly claims structural force that the Level of Abstraction method, operating on informational models rather than on the being of the systems modelled, cannot warrant. The Thomistic framework of Participated Agency provides that foundation: grounded in instrumental causality and the metaphysics of participated existence, it articulates the human-AI asymmetry as a feature of AI’s mode of being, not of observational models. This complementarity is asymmetric: the Level of Abstraction method retains descriptive-functional primacy; Participated Agency supplies the ontological and teleological horizon that AI ethics requires. The theological corollary is direct: human oversight over AI is not a contingent procedural choice but the recognition of an ontological structure grounded in the order of creation, one no increase in AI capability can dissolve.