Artificial intelligence is increasingly involved in cognitive and decision-making processes, raising concerns about the atrophy of human capacities, the overreliance on technological systems, and the erosion of individual responsibility. As a result, contemporary debates on AI are often framed in predominantly dystopian terms. Against this backdrop, the present paper addresses a central question: whether and under which conditions artificial systems can be integrated into decision-making processes in ways that preserve—and potentially enhance—human autonomy and moral agency. Building on the distinction between genuine and apparent enhancement, I argue that AI-mediated support may improve moral deliberation only when it becomes integrated into the agent’s reflective process rather than merely producing externally plausible answers. Having argued for the desirability of moral enhancement in light of the limitations of human moral psychology, and for the superiority of AI-mediated enhancement over biomedical alternatives, I critically examine existing models of Artificial Moral Advisors—particularly those grounded in subjectivist and realist metaethical frameworks. In response to their limitations, I propose an original model— which I have termed Constructivist Moral Advisor. Grounded in a constructivist metaethical perspective inspired by the Kantian and Rawlsian traditions, the ethical assistant I have conceived does not seek either to mirror the agent’s pre-existing opinions or to replace their judgment with externally imposed prescriptions. Rather, its function is to support and structure the deliberative process by fostering a reflective equilibrium among intuitions, principles, and contextual considerations. The practical implications of this proposal are explored through DeliberApp—a prototype conceived as an applied implementation of the Constructivist Moral Advisor. By addressing objections concerning its alleged practical ineffectiveness, I argue that AI, under certain conditions, can be reconceived as an epistemic interlocutor capable of expanding the horizon of moral reasoning while preserving individual autonomy.
