The Talking Lure: Raciolinguistic Ideologies of Reception and the Listening Subjects of American Vocal Biomarker AI

Abstract

In the United States, vocal biomarker AI is on the rise. The discourses encircling this interdisciplinary subfield propose that AI can identify discrete biological indicators of mental disability supposedly expressed in the sounds of the voice and detached from the sociocultural dimensions of communication. Many supporters and makers of vocal biomarker AI claim that AI holds revolutionary potential for American mental health care because it can identify mentally distressed speaking subjects based on these biological-acoustical features alone, without “hearing” any other categories of difference, such as race. Drawing from fieldwork with a lab that crafted an anthropomorphic agent for their vocal biomarker–detecting prototype, I deploy the analytic of the “talking lure” to demonstrate that racialization can coexist alongside, and indeed even materially enable, ideologies of AI’s race-evasive listening. Members of the lab at once enacted, depended on, and somewhat underplayed the significance of various processes of racialization that shaped the foundations of their technology’s AI model. By tracing the racialized and gendered hierarchies of listening and auditory personhood that were assembled and disassembled across the lab’s technology development pipeline, I outline and challenge vocal biomarker AI’s implicit theorization of race as an internal auditory signal that lies in wait to be either identified or disentangled from disability.