Artificial Extinction Archives: Machine Learning for Ecological Mourning

This article examines how algorithmic and machine-learning–based creative practices can offer new modes for narrating, and mourning, species extinction in the Anthropocene. To do so, I conduct a creative-practice–led analysis of the Artificial Extinction Archives (AEA): a new media artwork wherein a TensorFlow audio model has been trained to recognize the audio vocalizations of fifteen extinct or likely extinct bird species. I argue that computational media can make perceptible the haunting presence of lost species by staging encounters with fragmented translations of these vocalizations. The AEA produces partial, poetic textual apparitions when contemporary environmental sounds partially resemble archived recordings. This intentional incompleteness highlights the unnarratability of extinction, which is a vast, multispecies, multi-scalar process, and aligns the work with post-classical narratology, where narrative becomes fragmented and cognitively co-created by readers. Drawing on extinction studies, material ecocriticism, and computational narratology, I argue that fragmented, combinatory, algorithmic narrative and poetry forms can reflect the discontinuities and absences that structure ecological loss. I further argue that the act of making such an artwork constitutes a form of ecological mourning in itself.