Daniel Leix Palumbo | From Spoken Audio to Digital Identities: How AI impacts the interpretation of information communicated by the sound of voice

Daniel Leix Palumbo | From Spoken Audio to Digital Identities: How AI impacts the interpretation of information communicated by the sound of voice | University of Groningen, Department of Media and Journalism Studies | NWO PhDs in the Humanities | Supervisors: Prof. dr. Susan Aasman, Dr. Robert Prey, Dr. Matt Coler | 1 September 2022 – 31 August 2027 | d.f.g.leix.palumbo[at]rug.nl

Today’s digitised media context is undergoing an oral/aural renaissance, where voice technology increasingly permeates much of contemporary life. In particular, tech companies and states are massively investing in voice biometrics, making such technology a multi-billion global industry and a new centre of digital innovation. The increased market interest in voice biometrics is informed by a socio-technical imaginary that identifies in the soundings of voice a site to analyse and extract unique characteristics about an individual through AI. This imaginary is lately also shared by immigration agencies: voice biometrics have been incrementally introduced in asylum procedures to analyse the accent of undocumented asylum seekers and pinpoint their country of origin, starting with the European refugee crisis. Border authorities present voice biometrics as the newest secure and efficient techno-solution for decision-making, now added to the assemblage of technologies and databases that compose the surveillant knowledge infrastructure of the digital fortress of Europe.

The emergence of voice biometrics in asylum procedures indicates the soundings of voice as a new amenable site for identity construction and border control. I examine this data driven-practice of identity-making by looking at the use of voice biometrics by the German and Turkish immigration agencies to determine eligibility for asylum through accent recognition. Voice biometrics remain an under-researched top-down governmentality practice in asylum procedures, compared to longer-standing forms of biometric data collection such as fingerprinting and facial recognition. Whereas research that critically investigates AI voice technologies as surveillant systems has mainly focused on the literal translation into text of spoken audio, largely overlooking other information embedded in the sound of voices and its implications for digital identity management. This research project aims to contribute by bringing renewed critical attention to the auditory realm to analyse contemporary modes of governance and data surveillance, exploring how the datafication of sound can be instrumental in making the digital selves and the knowledge that composes them.