How are technologies that travel under the sign of AI shaping the sensory practices, ethical sensibilities, and professional landscape of American mental healthcare? My current book project explores this question through an ethnographic investigation of efforts to use voice analysis technologies to augment how people are listened to and heard as mentally distressed in the United States.

Speech, Signal, Symptom (under contract, University of Chicago Press) shows how machine listening technologies retrench, rather than reorient, the American mental healthcare system’s dominant interpretive paradigms, further constricting the agency of patients, practitioners, and technologists across the full stack of the research and development pipeline. At the same time, it turns up the volume on the self-reflexive critiques and strategic maneuverings that people engaged in constructing these technologies—from data annotators to human research subjects—develop through their perspective as the ones most responsible for iteratively transducing speech into computationally tractable and clinically valuable signals.

Research from this project was funded by the National Science Foundation, the Wenner-Gren Foundation, and the Society for Psychological Anthropology. It has been published in Science, Technology, & Human Values, the edited volume Technocreep and the Politics of Things Not Seen, and will appear in a forthcoming special issue on AI and linguistic anthropology in The Journal of the Royal Anthropological Institute and a 2026 issue of Current Anthropology. You can also read about this work in Somatosphere and the AI Now Institute’s A New Lexicon of AI series.