The Acousmatic Question and the Will to Datafy: Otter.ai, Low-Resource Languages, and the Politics of Machine Listening

Abstract

What happens when Nina Eidsheim’s acousmatic question—“Who is this?”—is delegated to machines? Machine listening processes turn sound and voices into data. This article explores the political stakes that accompany the automated extraction, processing, and analysis of human voices in machine listening, specifically speech recognition. While machine listening is promoted to users in the name of utility, inclusiveness, and access, it also serves corporate purposes: the expropriation and ownership of massive collections of data. This extractive will to datafy subtends commercial and state-based machine listening operations. We outline this problematic process though two case studies: the datafication of “low-resource” languages for speech recognition in India and the widespread adoption of Otter.ai transcription services in Canada and the United States during the COVID-19 pandemic. In both cases, noble aims—inclusion and access—are simultaneously coopted to serve corporations’ extractive projects, which are built on denying speakers the right to their own voices.