Can Robots Love Their Electric Sheep? Artificial Intelligence, Virtue, and the Impossibility of Moral Alignment

Abstract

Contemporary discussions of artificial intelligence generally assume that intelligence is fundamentally computational and that the central ethical challenge is aligning increasingly capable systems with human values. This article argues that such accounts presuppose a reductionist understanding of intelligence itself. Drawing upon Aristotle, Augustine, Thomas Aquinas, David Bentley Hart, and Jacques Ellul, it develops a classical Christian account of intelligence as an ontological participation in being that culminates in phronesis and is ultimately ordered toward beatitudo . Intelligence is therefore not merely the capacity to process information or produce correct moral outputs, but the perfection of a rational creature whose intellect, desires, and affections are formed toward the good through virtue and communion with God. Against this account, the article examines contemporary alignment research, focusing particularly on Anthropic's Constitutional AI, and argues that such systems achieve behavioral alignment rather than moral formation. While artificial intelligence may increasingly simulate the outward expressions of prudence, it cannot itself become prudent because it lacks the teleology, participation, embodiment, habituation, desire, and beatitude that constitute practical wisdom. The central challenge posed by artificial intelligence is therefore not merely technological but anthropological: before asking whether machines can become intelligent, we must recover what it means for human beings to be intelligent.