Listening Like a Computer: Attentional Tensions and Mechanized Care in Psychiatric Digital Phenotyping

Abstract

This article explores negotiations over the humanistic versus mechanized components of care through an ethnographic account of digital phenotyping research. I focus on a US-based team of psychiatric and engineering professionals assembling a smartphone application that they hope will analyze minute changes in the sounds of speech during phone calls to predict when a user with bipolar disorder will have a manic or depressive episode. Contrary to conventional depictions of psychiatry as essentially humanistic, the discourse surrounding digital phenotyping positions the machine as a necessary addition to mental health care precisely because of its more-than-human sensory, attentional capacities. The bipolar research team likewise portrays their app as capable of pinpointing sonic signs of mental illness that humans, too distracted by semantic meaning, otherwise ignore. Nevertheless, the team members tasked with processing the team’s data (audio recordings of human research subject speech) must craft and perform a selectively attentive machinic subject position, which they call “listening like a computer”: a paradoxical mode of attention (to speech sound) and inattention (to speech meaning). By tracing the team’s discursive and on-the-ground enactments of care and attention as both humanistic and machinic, I tune a critical ear to the posthuman promises of digital phenotyping.