Sounding out voice biometrics: Comparing and contrasting how the state and the private sector determine identity through voice

Abstract

The voice biometrics industry is promised today as a new center of digital innovation. Tech companies and state agencies are massively investing in speech recognition and analysis systems, pushed by the belief that the acoustics of voice contain unique individual characteristics to convert into information and value through artificial intelligence. This article responds to this current development by exploring the under-researched datafication of the auditory realm to reveal how the sound of voice is emerging as a site for identity construction by both states and corporations. To do so, we look at two different case studies. First, we examine a patent granted to the streaming service Spotify, which aims to improve the platform's music recommendation system by analyzing users’ speech. Second, we discuss the use of voice biometrics in German asylum procedures, where the country of origin of undocumented asylum seekers is determined through accent analysis. Through these seemingly distinct case studies, we identify not only the common assumptions behind the rationale for adopting voice biometrics, but also important differences in the way the private sector and the State determine identity through the analysis of the sounding voice. These two entities are rarely examined together and are often conflated when addressing practices of auditory surveillance. Thus, our comparative and contrastive approach contributes to existing scholarship that questions the claimed efficiency and ethics of voice biometrics’ extractive practices, further defining the operations and assumptions of the private sector and the State.