Listening/Not Listening: Educations of Inattention in the Automation of Psychiatric Assessment

• In Event: Attention! - II: Language, Expertise, and Labor

Abstract

This paper considers the ethical dimensions of attention, exploring a situation in which paying attention is not an elusive goal but an intractable, ever-present hurdle. I draw from ethnographic fieldwork with teams of psychiatric and engineering professionals working to develop technologies for conducting psychiatric assessment based on how a person sounds instead of the content of what they say.

Researchers invest hope in their algorithmic prototypes’ capacity to identify objective, “vocal biomarkers” of mental illness that are otherwise inaudible to humans too distracted by subjective, sociocultural meaning. On one hand, researchers position their technologies as superior to conventional psychiatric assessment because of what their technologies can attend to (acoustic qualities of speech thought to be directly tied to the biological mechanisms of psychopathology) and what they can ignore (the semantic significance of language). On the other hand, in order to build the data infrastructure foundational to their technologies, researchers themselves must first elicit, record, listen to, and label excerpts of research participants’ speech, defining the features that their algorithms should listen for while also honoring ethical protocols meant to protect the participants’ privacy.

I follow researchers as they grapple with the fraught task of “listening like a computer”: listening to (and quantifying) sound but not listening to (or becoming emotionally invested in) content. Taken together, their techniques offer a critique of Euro-American ideologies of language, machine autonomy, and professional codes of responsibility. They also suggest how inattention, rather than being opposed to care, can enact a form of care-ful listening.

Author

• Beth Semel, Massachusetts Institute of Technology (MIT)