The open work and AI: Co-creation and haunting in text scores and inherent scores

Abstract Drawing on Umberto Eco’s epistemological metaphor of the open work, this paper explores the intersection of two open forms of notation, inherent scores and text scores, with generative AI. Building on the notion of inherent scores, in which the interface merges with the notation, we introduce embodied sketching, a notational approach that streamlines composition and performance with real-time neural audio synthesis (NAS). We then examine text scores in text-to-audio NAS, presenting Mouja+, a work combining real-time NAS with embodied sketches and AI-generated audio from Fluxus scores. Based on the experience of composing and performing Mouja+, we show how AI’s statistical processing of language introduces interpretative gaps between the human understanding of the scores and the model’s output and propose prompting strategies to streamline the use of text scores with text-to-audio generative AI. We continue by discussing how NAS adds to the open work through algorithmic processes that coalesce into an elusive and deferring sense of presence. Through Derrida’s notion of hauntology, we thus extend the open work into what we term the ‘haunted work’, an epistemological metaphor encompassing a growing corpus of works engaging with the tension between presence and absence as a source of openness.