Who Shapes the Remembered Past? Narrative Authority and Representational Contestability in AI-Mediated Memory Art

Abstract

Generative AI can give photographic form to personal and familial pasts for which photographs are absent, incomplete, or irretrievable. Yet the memory that sets a generated image in motion does not necessarily determine the image that will eventually materialize. This article approaches such divergence by comparing Maria Mavropoulou's Imagined Images (2021-2023) with Weidi Zhang, Lijiaozi Cheng, and Jieliang Luo's ReCollection (2023). Drawing on Doreen Mattingly's idea of narrative authority, the article proposes four dimensions along which the control of representation may be distributed across generative processes: input control, elaborative control, revision capacity, and contextual control. It then develops representational contestability as a domain-specific elaboration of scholarship on contestable AI to ask whether those whose memories underlie generated representations can identify significant interpretive additions and have any meaningful capacity to respond to them. The contrast thus suggests that personal contribution, intimacy, emotional resonance, and narrative authority are interrelated but not identical. Therefore, memory art mediated by AI requires analysis not only of photographic truth or synthetic fabrication, but of how interpretations of personal memory are produced, attributed, revised, and open or resistant to contestation.