The Political Dimensions of Sound and Language
As the new guest artist of the Max Planck Institute for Empirical Aesthetics’ INHABIT Artist-in-Residence Program, Pedro Oliveira began his residency at the end of May—initially from home due to the coronavirus pandemic. During his three-month residency, Oliveira will pursue combined artistic and scientific research on the political dimensions of sound and language, focusing on the use of speech recognition software in the German asylum system.
Oliveira has lived in Germany since 2009, completing both an MA in Digital Media in Bremen and a PhD in Design Research in Berlin. In his doctoral thesis, he dealt with the very current topic of police violence, specifically, sonic articulations of police violence in post-2013 Brazil. He has since extended this work to the field of speech recognition technologies; and his current project examines the automated accent recognition software used by the German Federal Office for Migration and Refugees (BAMF) to determine the origin of asylum seekers. Oliveira considers the use of such software to be critical and problematizes above all the sound database from which it is fed.
“That these forms of congealment of given speech traits are then used as a matter of identification, supporting evidence, and subsequent deportation of undocumented migrants is an articulation of sonic violence I am particularly interested in exploring, challenging, and intervening in.”
For Oliveira, the deployment of accent recognition software in the asylum process is crucial for understanding how machine listening can be instrumentalized as a violent and dehumanizing device. During his residency at the Institute, Oliveira will examine how the sound database is annotated and parsed before and during the process of training and calibrating the software. He is particularly interested in how the software is trained, especially what it is trained to ignore. Paraverbal speech characteristics, such as hesitation marks, coughs or specific intonations, are to a certain extent labeled as “noise” by both human and machine ears, and are for this reason a promising entry point for understanding the affective and material properties of language. The results of his collaboration with the Institute will be presented and documented as a fusion of sound installation and performance.