Resistant Resonances: Vocal Biomarkers, Transductive Labor, and the Politics of Things Not Heard Available to Purchase

In the United States, engineering and mental health-care professionals increasingly collaborate to develop vocal biomarker artificial intelligence, technologies that can supposedly detect mental distress by analyzing the sounds of the voice alone. This article draws from ethnographic fieldwork with individuals typically excluded from dominant accounts of vocal biomarker AI’s promises and perils: technicians and human research subjects who engage in transductive labor, or the work of transferring sound across media. Transductive labor enables the conditions of possibility of vocal biomarker AI, knitting together the connection between sound and psyche that AI “finds.” Yet it also enables subversive and computationally intractable glitches, gaps, and care practices throughout the technology development pipeline. Thus, focusing on transductive labor can complicate both techno-optimistic and technopessimistic investments in the capacity of AI to fully capture mental distress through the voice, directing attention instead to the forms of relationality vocal biomarker AI fabricates, re-articulates, or disrupts.