Answers are grounded in this paper and cite their sources.
Contemporary discussions of artificial intelligence emphasize alignment with human intentions, preferences, norms, and values. This article argues that alignment is necessary but aesthetically insufficient for understanding human–AI creative interaction. In artistic practice, criteria of success may change through encounters with unforeseen results. I define productive misalignment as a bounded departure from an antecedent intention, expectation, or provisional aesthetic criterion that initially fails to satisfy that criterion yet becomes retrospectively affirmable through subsequent aesthetic judgment. Unlike the discovery of a hidden preference or satisfaction of a higher-order preference for surprise, the decisive case involves transformation of the evaluative criterion itself. Drawing on theories of aesthetic judgment, accident, resistance, and improvisation, the article distinguishes productive misalignment from technical failure and ethical misalignment. It also defends an asymmetry between distributed generative contribution and project-level evaluative authorization. I propose aesthetic negotiation as a model of bounded, revisable human–AI interaction: normative constraints establish the limits of permissible generation, while aesthetic judgment determines whether generated difference should be rejected, retained, or allowed to redirect the work.
No takes yet. Share an insight, caveat, or question.
Choi B* (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: