A taste for what should not be wanted: generative AI and transgressive fantasy

Abstract With generative AI, everyone can materialize whatever fantasy they can conjure​ up in their imagination, including transgressive ones. The ‘orthodox’ view on the moral status of fantasies, even of materialized ones, is that such content is outside the scope of morality, since fantasies are quarantined from the agent’s real-world beliefs and desires. In this paper, we challenge this view by arguing that even if we grant that that is the case, repeated engagement with fantasies can contribute to the cultivation of various affective dispositions, or tastes, that can themselves be morally criticizable. We then argue that generative AI allows users to enter a ‘co-creative feedback loop’, a near-instantaneous cycle of prompting, viewing, and refining one’s fantasy, which makes sustained engagement with the transgressive representations very easy. Our claim is that such repeated and focused engagement with transgressive content can contribute to the cultivation of a taste or sensibility for what is morally problematic in those fantasies. If such tastes are morally criticizable, then engaging in practices that cultivate them is also open to moral criticism, even when no one is directly harmed.