WHO ANSWERS FOR THE WORK? ADAPTIVE AGENCY, POLYPHONY, AND HUMAN RESPONSIBILITY IN AI-ASSISTED CINEMA

Abstract

The conference's central question — how authorship, creativity, and responsibility survive the spread of generative AI — appears in acute form in cinema, where a single operator can now coordinate functions once distributed across an entire production team: writing, direction, cinematography, performance, editing. Does compressing this chain into one human–AI workflow collapse the dialogical structure long taken to characterise authored work? This paper argues that it does not, but that the resulting polyphony is structurally thinner. It introduces adaptive agency — the capacity to adjust action, deliberation, and control under new information while still shaping a work toward meaning — and distinguishes three senses of polyphony routinely conflated: hosted polyphony, voices held within a single authorial consciousness that retain answerability (Bakhtin on Dostoevsky); ensemble polyphony, real co-present subjects who resist authorial totalisation; and generative polyphony, alterity produced by a medium that lacks answerability. The distinction turns on one criterion: whether the other voice could, in principle, answer for its utterance. The AI can speak, yet cannot be responsible for what it says — an asymmetry with direct ethical consequences wherever generative systems enter culture, education, and translation. Work can therefore still be authored, provided a human remains the answerable agent; what it cannot do is reproduce the meaning yielded by irreducible others. Cinema is the test case, but the criterion is portable: it locates precisely where, across AI-mediated practice, responsibility must stay human.